Decoding Decisions
Making sense of the messy middle
Authors
Jonny leads Google’s insights team in the UK. Prior to his 10years of
marketing research at Google, he plied his trade in the media agency
world. He is a regular speaker on evolving consumer behaviour, cross-
media measurement, and marketing effectiveness. He represents
Google in a range of forums in the UK marketing and research industries.
Jonny Protheroe
Head of Market Insights UK, Google
Ali began his career researching ad effectiveness in agency before
moving client-side to build and lead a team with an expanding remit
ranging from marketing and proposition development to innovation
and business strategy. Over the past 20 years he has developed a
highly strategic approach to insight development which he now uses to
explore consumer behaviour and decision-making at Google, where he
is a regular speaker on these topics.
Alistair Rennie
Research Lead, Market Insights UK, Google
Claire began her career on the agency side at 360i and relocated
to London ve years ago to head up the insights team at 360i’s
new European outpost. Upon joining Google, Claire specialised in
emerging and innovative research methodologies, from neuro and
physiological response research to trend identication through
text analytics. In2017, she was named one of Management Today’s
35 Women Under35. She is now working as a Product Manager in
Google’sworkshop forexperimental products.
Claire Charron
Product Manager, Google
Gerald’s team works with UK clients, bringing together Google data
analysis and research to help address their challenges. Before joining
Google almost 10 years ago, he worked on some of the UK’s biggest
brands at well-known creative agencies. He sits on the Institute Of
Practitioners in Advertising (IPA) Effectiveness Advisor y Board and is a
regular speaker on marketing effectiveness and strategy.
Gerald Breatnach
Head of Strategic Insights UK, Google
Chapter
Chapter
Chapter
Chapter
Chapter
Chapter
Contents
1
4
3
2
5
6
Introducing the messy middle ...............................04
Taking a walk down internet street .................................04
What does the consumer journey look like? ..........................06
Charting undiscovered territory ...................................07
Marketing in the messy middle ....................................09
Identifying the messy middle ................................10
Cheap, or best? .................................................10
Enter behavioural science ........................................12
Riders and elephants ............................................13
A brief histor y of the evolution of marketing models ...................14
Observed shopping behaviour ..................................... 15
Exploring and evaluating .......................................... 16
The science behind explore and evaluate ............................ 17
Show... me... the... model! ......................................... 18
E is for... .......................................................19
Investigating the messy middle .............................26
Searching for clues in Google Trends ...............................26
Are you not entertained? The slow demise of ‘free’ ....................28
OK Google, let’s go shopping ......................................31
‘Best’ and ‘cheap’ around the world ................................45
Influencing the messy middle ...............................47
Homo-not-so-economicus .......................................47
A summary of six biases ..........................................49
Testing the six biases ............................................51
The simulation .................................................55
The power of showing up .........................................56
Social proof: people respond to people .............................. 60
The low-hanging fruit of behavioural science ........................61
Cross-functional implementations .................................65
Supercharging the second-choice brand ............................68
Starting from nothing ............................................75
Simulation summary ............................................. 81
Implications of the messy middle .............................. 84
Implications for brands ..........................................84
Ensuring brand presence .........................................86
Intelligently (and responsibly) employing behavioural science ..........88
Closing the gap between trigger and purchase .......................91
The importance of measurement ..................................92
Organisational implications .......................................92
Inhabiting the messy middle ................................94
1
Taking a walk down internet street
Let’s take a walk down internet street. You might know it, and since the
coronavirus pandemic you might have spent more time there than usual. It’s
open 24 hours a day, seven days a week. It’s probably the biggest shopping
district you’ve ever seen, but somehow you can get to any part of it you want
in the time it takes to blink.
Whenever you’re ready, you can tell someone what you’re looking for and,
in addition to the brands that are already in your mind, you’ll immediately be
shown every possible option and variation.
Every shop that sells that item is somehow just a step away, whether it’s a
huge department store or a tiny boutique. And the shops you don’t need will
magically disappear from view until you want to see them again.
Maybe you don’t know what you want. If that’s the case, there are places you
can go that will show you every product available in every store. They’ll let
you rank and compare them in every imaginable way, sorting and ltering
until you see something you like.
And if you still can’t decide, you can ask a friend for advice. Or an expert. Or a
famous celebrity. They’re all here too, some of them hanging out at their own
places, while others will come and meet you at the store. In fact, there are
millions upon millions of people here, most of whom are only too happy to talk
to you about the things they decided to buy, and how that experience turned out.
Introducing
the messy middle
CHAPTER 15
INTRODUCING THE MESSY MIDDLE
There’s a lot happening here on internet street. Because it’s all so easy, you
might wind up making several visits before you get around to actually buying
anything, moving in and out of multiple stores, going back for a second and
then a third look, making full use of everything internet street has to offer.
That’s the reality of shopping on the internet today, but it hasn’t always been
like this. Before the internet, we shopped on a physical street, where we had
less choice and less information. What we ended up buying was restricted by
availability and proximity, and we relied on brands to reassure us that we were
making the right decisions. We even had to carry our own shopping baskets.
Our behaviour has fundamentally changed, but for the most part we revel in
it, as instincts formed by thousands of years of scarcity are supercharged
with a sudden wealth of options and opportunity. So much choice, so many
shops to visit and products to view. So much complexity that we’ve turned to
a range of coping mechanisms – mental shortcuts and techniques that help
us cut through to what matters.
Marketing has also evolved and developed new ways of cutting through.
Marketers have embraced new platforms, new technology, new data, and
new formats. And lately, innovations like machine learning and articial
intelligence are pushing all of this further and faster into the future.
Most of these developments have been good things. The expansive reach
ofdigital marketing has allowed new businesses to emerge and grow.
Butwhile this is ultimately a report about marketing, it is not a report about
that side of the equation.
Instincts formed by thousands of years of
scarcity are supercharged with a sudden
wealth of options and oppounity.
CHAPTER 16
INTRODUCING THE MESSY MIDDLE
1 ONS Retail Sales Index time series (DRSI), UK, May. 2020
Instead, this report is about the mental processes that have been activated
by the abundance of the web. It’s about how consumers deal with scale and
complexity using cognitive biases encoded deep in our pre-digital history.
If behaviour has evolved, as we believe it has, then it is crucial that marketers
understand how consumer decision-making has changed so that they can
continue to uncover new growth opportunities and defend existing brand share.
What does the consumer journey look like?
This is among the questions most frequently asked of Google’s insights
team. There are a couple of variations involving phrases like “purchase
funnel” and “path to purchase” but, for the most part, they’re all asking the
same thing. There’s a lot of value in questions like these, but we’ve come to
realise that there is another aspect of what shoppers are doing that needs
to be considered. The other question we need to answer is this: how do
consumers decide what they want to buy and who they want to buy it from?
It isn’t surprising that businesses are keen to outsource this question. It’s
probably the most important in all of advertising, but also the hardest to
answer. Often, research in this area will focus on the journey, resulting in a list
of touchpoints that people hit along the path to purchase. But while such lists
offer valuable insight into the places people go during their online journey,
they can’t address the equally important question of why a shopper ended up
making the decision they did.
We know more about advertising performance than ever before, and can
measure outcomes with amazing granularity. And yet, understanding
consumer decision-making is more difcult than it’s ever been. In 2020,
following the outbreak of coronavirus and subsequent restrictions on
physical retail, the proportion of purchases happening online has risen to
record levels. And while the majority of purchases are still made offline, the
media and information that inform those purchases are increasingly online,
and the complexity of potential decision-making pathways has grown
considerably. If we don’t update our thinking about consumer behaviour to
account for this huge expansion in choice and attendant complexity, we’ll be
trying to account for 21st century behaviour with 20th century models.
CHAPTER 17
INTRODUCING THE MESSY MIDDLE
2 We shared an early draft of the project with the well-known advertising strategist Vicki Holgate, and she played it back to us as “a kind of messy-middle”.
We tried various titles and names for presentations, but this was the phrase that stuck.
Charting undiscovered territory
So, over the course of the past two years, our team has embarked
on a multi-pronged project with the goal of trying to understand
how consumers on internet street interpret and manage increased information
and choice while buying online and offline. This research has led us to
identify a specic territory within the labyrinth of searches, ads, links, and
clicks involved in making a purchase. We call it the “messy middle”, a space
of abundant information and unlimited choice that shoppers have learned to
manage using a range of cognitive shortcuts.
2
Successfully learning how to
navigate its switchbacks, hairpin bends, and dead ends is going to be as crucial
to future marketing success as any investment in technology or platforms.
The ‘messy middle’, a space of abundant
information and unlimited choice that
shoppers have learned to manage using a
range of cognitive shocuts.
Once we discovered this territory, we set out to map it. In doing so, we
devised an updated model for how we believe people behave in this sphere of
abundance and uncertainty.
With the help of behavioural science expert The Behavioural Architects, we
recruited people to complete shopping tasks, captured their behaviour, and
listened in real time as they told us what they were thinking and doing, and
why they were doing it. As we watched, we began to notice how seamlessly
consumers switch between complementary states of “exploration” and
“evaluation”. We then applied behavioural science to help us cut through
the participants’ explanations and post-rationalisation to understand the
underlying cognitive processes at work.
CHAPTER 18
INTRODUCING THE MESSY MIDDLE
Unsurprisingly, it turns out that faced with all this complexity, people try to
keep things simple – an effort that in itself turns out to be quite complex.
To validate the existence of the exploratory and evaluative states, we also
looked through Google’s historic search data for clues. In several cases, we
found examples of changes in the way people search over time that illustrate
how these behaviours manifest in the real world.
Alongside this, we also undertook a thorough literature review to try and
isolate the specic cognitive processes at work while people are caught up
in the exploratory and evaluative whirl. We identied six of the most critical
biases, and then devised a large-scale experiment to test the effectiveness of
these shortcuts and heuristics in guiding shoppers out of the messy middle
and towards purchase.
Over the following chapters, we’ll explain why we started looking for the
messy middle, the tools we used to identify and codify it, and the discoveries
we made while exploring it. We’ll share some of the most surprising insights
from the process, including:
The power of showing up – how simply being present in
moments of deliberation can be enough to win or retain
consumer preference.
Several of the most powerful behavioural biases we
investigated can be easily addressed by marketers surfacing
and modifying existing assets.
Why addressing some of the most powerful behavioural
biases requires cross-functional cooperation from marketing, user
experience, product development, and nance.
Finally, we’ll wrap up with specic ideas for how marketers from both
established and challenger brands can adapt to this rich and complex space.
CHAPTER 19
INTRODUCING THE MESSY MIDDLE
Marketing in the messy middle
Access to media and information has led to the growth of important
influences that don’t necessarily t into traditional brand marketing or
performance marketing buckets. This has some big implications for
marketers from brands both large and small. If you don’t truly understand
why consumers make the purchase decisions that they do, you may not
achieve the full return on your brand investments, and could nd yourself
vulnerable to nimble competitors.
It seems then that “messy middle” might also be a good way to describe how
marketing has evolved over the past decade or two, with the polarisation
between branding and direct response creating a gap into which all sorts
of valuable consumer behaviour goes unrecognised and underserved.
Getting comfortable with the messy middle could ultimately help bridge
organisational divides that our research suggests mean more to marketing
departments than they do to consumers.
Of course, guring out what consumers think and how they behave is not a
new idea. It’s an aspiration that’s always been at the very heart of marketing.
But, as we’re about to nd out, the context within which marketers are trying
to achieve this goal has changed dramatically.
Geing comfoable with the messy
middle could ultimately help bridge
organisational divides that our research
suggests mean more to marketing
depaments than they do to consumers.
3 Source: https://www.blog.google/products/search/search-language-understanding-bert/
2
Cheap, or best?
The research project behind this report began with a hunch that there was
more to say about the evolution of choice, information, and decision-making
on the internet. The next step was to look for clues to support and expand our
initial hypothesis.
To kick off the investigation, we turned to one of our biggest resources
as researchers at Google: our search trends data. Google sees billions of
searches every day, and 15% of those queries are ones we haven’t seen
before.
3
Our freely accessible search query exploration tool, Google Trends,
represents a detailed history of how our curiosity and thirst for knowledge
has evolved throughout the digital age. Using Google Trends data you can
chart the fortunes of celebrities, politicians, and reality TV stars, observe the
rise and fall of a decade’s worth of memes and fads, and watch the iPhone
and Android create and dene a category.
But the names of people and objects aren’t the only data points in our Google
Trends dataset. When consumers search, they often modify the query with one
or more adjectives or other descriptors. You aren’t just looking for any laptop,
but for the right laptop – however you dene it. We call these additional words
modiers, and they describe what the user wants to know about the thing they
are searching for, or add precision to their search. Modiers provide a cognitive
and emotional snapshot, allowing researchers to see how our feelings and
needs have evolved through the lens of the things we all search for.
Identifying
the messy middle
CHAPTER 211
IDENTIFYING THE MESSY MIDDLE
4 UK Ofce for National Statistics – Living Costs and Food Survey
Data source: Google Trends, United Kingdom, 1st January 2004 - 1st January 2020, All categories, Web Search
Turning to our trends data, we immediately began to nd some tantalising
clues. Take the terms “cheap” and “best”. In the UK, interest in search queries
containing the word “cheap” has steadily declined over the past 15 years,
while interest in “best” has increased with an impressive degree of negative
correlation (gure 1).
Figure 1
2004 2006 2008 2010 2012 2014 2016 2018 2020
0
25
50
SEARCH INTEREST
75
100
cheap
Search term
best
Search term
This data suggests that at some point around 2009, consumer interest in
nding the cheapest item online was eclipsed by a desire to nd the best. One
hypothesis to explain this might be that as average incomes increase over
time, an appetite for signiers of wealth, such as having the “best”, might
increase too. However, when these two trends crossed over in 2009, the
world was in the grip of the worst nancial crisis since the Wall Street Crash
– following which median household incomes in the UK actually fell.
4
Looking more closely at “cheap” and “best”, it quickly becomes apparent that
these two modiers are very different in scope and application. “Cheap” is
quantiable and rational, “best” is more subjective and emotional. The precise
value of “cheap” may vary between individuals, but it still carries a singular
meaning. “Best”, on the other hand, can have a wide range of meanings, being
applicable to value, quality, performance, popularity, and more.
The trends for UK searches containing “cheap” and “best” have been in opposite directions.
CHAPTER 212
IDENTIFYING THE MESSY MIDDLE
As the internet has grown, ithas
transformed from a tool for
comparing prices to a tool for
comparing everything.
It is this transition from simple to complex modiers that offers the rst
signicant clue to how consumer behaviour and decision-making have
changed. As the internet has grown, it has transformed from a tool for
comparing prices to a tool for comparing everything.
Enter behavioural science
To go beyond describing what consumers are doing on the internet to
understanding why that behaviour has changed, we needed to take a
different approach, grounded in cognitive science. Our partner from the
beginning of this project has been The Behavioural Architects, a global
consultancy specialising in the application of behavioural science to
marketing challenges.
We’re certainly not claiming to be the rst to apply behavioural science to
marketing. Influential marketers have long emphasised the importance
of using mental shortcuts to build brand salience and create messages
that generate a response, so the use of behavioural insights will not be a
new concept for advanced practitioners. However, with the help of The
Behavioural Architects we’ve been able to comprehensively review a
signicant proportion of the available scientic and marketing-related
literature, and to use it as the foundation of a series of large-scale
experiments exploring the impact of behavioural biases that we’ll review in
Chapter Four.
CHAPTER 213
IDENTIFYING THE MESSY MIDDLE
Riders and elephants
There’s a famous analogy used to describe how reason and emotion interact
when we’re making decisions. Jonathan Haidt, psychologist and Professor
of Ethical Leadership at New York University, likens the relationship to that
between an elephant and its rider. The rider is notionally in charge of where
the pair are going, but as soon as some stimulus or other catches the
elephant’s attention, the rider quickly nds out how little control they really
have. The signal of the reins is soon drowned out by the noise of a trumpeting
giant charging towards the fullment of one of its primal needs.
Inevitably, the elephant’s motives are something of a mystery to the rider. If
you ask them to explain what happened, they’ll be able to tell you where they
wanted to go, but not why they ended up where they did. Answers about the
elephant will be mostly guesswork and post-rationalisation. The mechanism
that often causes emotion to overhaul reason remains hidden to us.
Many attempts have been made over the years to isolate the signals and
cues most likely to make the elephant take control and, in a sense, the project
we embarked upon had a similar goal. After all, anywhere that has recently
been visited by an elephant tends to end up a little messy.
The mechanism that oen
causes emotion to overhaul
reason remains hidden to us.
CHAPTER 214
IDENTIFYING THE MESSY MIDDLE
The Behavioural Architects eventually proposed the above list as representative of the way
thinking in this space has evolved. It isn’t exhaustive – we chose not to include any model that
seemed more focused on organisational concerns than consumer perspectives – but what
this list does show is a general direction of travel and a tendency towards increasing detail.
1898 1997
1924
2005
1961
2009
1986
2011
1. AIDA
Elmo Lewis’ theoretical customer
journey from the moment a brand or
product attracts consumer attention
to the point of action or purchase.
2. The Funnel
William Townsend’s adaptation
of AIDA. Introduced the funnel
concept.
3. DAGMAR
Not intended as a decision-making
model, but Russell Colley adds an
important pre-awareness stage to
the funnel.
4. Moment of Truth
Jan Carlzon’s model, captured in
his claim that: “Any time a customer
comes into contact with a business,
however remote, they have an
opportunity to form an impression”.
5. ATR-N
Ehrenberg’s model emphasises
the importance of post-purchase
experience and interaction (nudges).
6. First and Second
Moments of Truth
A.G. Lafley builds on Carlzon’s moment
of truth, distinguishing between looking
at the product and then using it with the
rst and second moments of truth.
7. The McKinsey consumer
decision journey
McKinsey’s “active evaluation” stage
updates decision-making to reflect
a less linear, active process and
introduces the “loyalty loop”.
8. ZMOT
Google extends Carlzon’s and Lafley’s
moments of truth with the “zero moment
of truth” - when you start to learn about a
product or service for the rst time.
A brief history of the evolution of marketing models
One of the ways that marketers have tried to describe (and to some extent prescribe) the
paths elephant and rider take towards purchase is to map them in marketing models.
To give us some historical context, The Behavioural Architects kicked things off with
an extensive investigation of marketing model white papers, starting with Elmo Lewis’
famous AIDA, and covering several of the influential models that have emerged over the
intervening century and a bit.
CHAPTER 215
IDENTIFYING THE MESSY MIDDLE
Observed shopping behaviour
Google started out as a postgraduate research project, so we have a healthy
respect for the perspectives of academics and experts. However, our
ultimate source of truth is always the consumer, and we knew we wanted to
nd a way to get back to their perspective.
Our method for doing this was to observe several hundred hours of shopping
tasks, covering 310 different journeys across 31 categories. In these tasks,
shoppers were asked to research a product for which they were currently
in-market. Journeys were recorded using screen capture audio and video,
while the shoppers talked us through what they were doing. The Behavioural
Architects then analysed the journeys through the lens of behavioural
science, annotating the video playback with the different cognitive biases
they observed.
After watching the recordings, we made an initial attempt at describing what
we’d seen. On a Post-It Note we drew the purchase trigger at the top and the
purchase itself at the bottom, and in the middle we drew this (below).
CHAPTER 216
IDENTIFYING THE MESSY MIDDLE
search engines, review sites, video sharing sites, portals,
social media, comparison sites, forums, interest groups /
clubs, retailer sites, aggregators, blogging sites, voucher /
coupon sites, branded sites, publishers, noticeboards
Shopping touchpoints observed
Figure 2
Having arrived at these sites, of which there are multiple to choose from,
many of the shoppers spent signicant amounts of time navigating back
and forth, switching between sites across multiple browser tabs and apps. In
fact, in some of the sessions we observed, the product under consideration
actually changed mid-search, as a new option became preferred.
Exploring and evaluating
Taken together, the literature reviews and observed shopping tasks started
to reveal some of the core characteristics of the new reality of consumer
decision-making.
We began the rst chapter as shoppers browsing an innite high street,
moving effortlessly between vendors until something catches our eye. If we
like what we see on closer inspection, we can check out immediately, but if not
it hardly matters – there are plenty of other stores to visit on internet street.
In between those two points there is a winding, scrawled squiggle, which
seemed a reasonable way to represent our rst signicant discovery: there
are no typical journeys. Instead there is a confusing web of touchpoints that
we likened to spaghetti, not least because it was clear that this would be a
real mess to clean up.
The different sites and touchpoints visited by the shoppers who took part in
the shopping observations included, but weren’t limited to, the items listed in
gure 2.
CHAPTER 217
IDENTIFYING THE MESSY MIDDLE
5 McKinsey (2009), https://www.mckinsey.com/business-functions/marketing-and-sales/our-insights/the-consumer-decision-journey#
6 Pirolli, P., & Card, S. (1999). Information foraging. Psychological Review, 106(4), 643–675. https://doi.org/10.1037/0033-295X.106.4.643
This sequence of looking for products and then weighing options equates
to two different mental modes: exploration and evaluation. And, as it turns
out, they are the key to understanding the messy middle. Exploration is an
expansive activity, while evaluation is inherently reductive. When exploring,
we add brands, products, and category information to mental portfolios or
“consideration sets”. When evaluating, we narrow down those options.
In McKinsey’s consumer decision-making model
5
(one of our favourite recent
models), these modes are combined into a single “active evaluation” phase.
However, our research suggests that they are cognitively distinct with different
reward systems and, as such, different tactics are required to connect with
consumers depending on whether they are exploring or evaluating.
The difference between giving a consumer information about a category or
product and actively closing a sale is subtle but important. In any transaction
choice is power, and consumers are now more powerful than ever before.
Sending the wrong signal at the wrong moment could be highly disruptive,
with the result that the offending brand is jettisoned from the shopper’s
consideration set.
The science behind explore and evaluate
Next we wanted to validate that explore and evaluate t within the existing
body of behavioural science. So, we went back to our stack of books and
periodicals to see if anyone else had identied a similar pattern of behaviour.
One theory that closely matched our hypothesis is “information foraging”,
6
which describes behaviours humans exhibit to reduce energy expenditure
whilst maximising gain. Historically, this theory was derived from a food
foraging theory which helped biologists understand animals’ feeding
strategies – in the case of a predator: how much energy is required to hunt
prey versus the energy that will be gained from eating it? Applying these
theories to online behaviour could explain how we explore and evaluate: how
easy is it to nd the information we need and how useful will that information
be? If it’s useful, we tend to exhaust the information at that location before
proceeding to the next. If not, we rapidly switch sources before we expend
too much energy.
CHAPTER 218
IDENTIFYING THE MESSY MIDDLE
Show... me... the… model!
We didn’t set out to build a new marketing model, but after sifting through
hundreds of white papers and spending as many hours observing online
shopping journeys, we realised that only a new structure would allow us to
piece together everything we’d learned.
Between the twin poles of trigger and
purchase sits the messy middle.
CHAPTER 219
IDENTIFYING THE MESSY MIDDLE
7 Binet, L., & Field, P. (2013). The Long and the Short of It: Balancing Short and Long-Term Marketing Strategies. IPA
If you recognise a few of the marketing models mentioned earlier, there’s a
chance that our model will feel familiar, sharing common elements with the
McKinsey model and others. This is intentional – our brief history shows how
each generation builds on the models that came before, stretching all the
way back to AIDA. However, we do believe that we’ve identied several novel
elements that reflect nuances of decision-making that previous models don’t
fully capture.
In our model, between the twin poles of trigger and purchase, sits the messy
middle, in which consumers loop between exploring and evaluating the
options available to them until they are ready to purchase. This process takes
place against an ever-present backdrop of exposure – effectively a substrate
representing all of the thoughts, feelings, and perceptions the shopper has
about the categories, brands, products, and retailers. After purchase comes
experience with both brand and product, all of which feeds back into the sum
total of exposure.
That’s the simple version – over the rest of this chapter we’ll look at each
component of the model in more detail.
E is for...
Alliteration is a good aide-memoire and, in a happy coincidence, all of the
novel components in our model beyond trigger and purchase begin with the
same letter.
Exposure
Describing the effect of brand advertising in a marketing model is tricky.
Brands can inspire powerful emotional responses and, as Binet and Field
have shown,
7
their impact can be felt throughout the decision-making
process. Moreover, their power doesn’t only derive from advertising. Brands
have a presence beyond marketing: our associations with them may
be life-long in some cases and everything we know about them, from a
newspaper article to a conversation overheard on the street, can influence
our perceptions.
CHAPTER 220
IDENTIFYING THE MESSY MIDDLE
To capture this broad spectrum of interaction and influence, we propose
“exposure”. Exposure is your awareness of the brands and products in a
category. Exposure is the sum total of all the advertising emanating from a
category that you’ve seen or heard. It’s the things you’ve learned through
word of mouth, the things you’ve read in the press and online. It can be
passively assimilated prior to a purchase trigger, part of the trigger itself,
actively sought or experienced post-trigger, and it can be a deciding factor in
the nal purchase.
But crucially, exposure is not a stage, or a phase, or a step. It’s an always-on,
constantly changing backdrop that remains present throughout the duration
of the decision-making process.
And it’s not just made up of branding and brand perceptions. Broader
category exposure and related category exposure are also components of
the backdrop. This too is a vast territory, but these types of exposure are
often complicit in triggering a purchase.
CHAPTER 221
IDENTIFYING THE MESSY MIDDLE
The exploration and evaluation loop
This seemingly innite construct is the dening characteristic of the messy
middle (the design we’ve chosen for the loop isn’t an accident). Consumers
explore their options and expand their knowledge and consideration sets,
then – either sequentially or simultaneously – they evaluate the options
and narrow down their choices. For certain categories, only a brief time
might be required moving between these modes, while habitual and impulse
purchases may bypass the loop altogether. But other purchases, typically
more complex, encourage or even oblige us to engage in lengthy exploration,
generating a healthy number of options to evaluate.
CHAPTER 222
IDENTIFYING THE MESSY MIDDLE
The loop is our best attempt at describing the non-linear nature of the messy
middle, with its back and forth between destination sites and mental modes
until one lucky brand emerges victorious. For marketers the challenge
is simple: how do you ensure that when the shopper stops flip-flopping
between states, it’s your product or service that wins? In other words, how
do you persuade someone to stop shopping around and actually buy what
you’re selling?
But while the endless circulation of the exploration/evaluation loop might
frustrate advertisers, it’s important to remember that it often delights
consumers. The goal is not to stymie the customer or force them out of the
activity they have chosen to pursue, but to provide them with everything they
need to feel comfortable making a decision.
Consumers explore their options and
expand their knowledge and consideration
sets, then – either sequentially or
simultaneously – they evaluate the options
and narrow down their choices.
CHAPTER 223
IDENTIFYING THE MESSY MIDDLE
Experience
This last component of our model occurs outside of the messy middle, so
we’ll only touch on it briefly. The experience a customer has with the product
or service they’ve purchased feeds directly into their background exposure.
A brand that provides a good experience has a head start here, and a brand
that delivers an amazing experience might even become a trigger itself,
potentially increasing the frequency of purchases.
But with so much choice available in the messy middle, a brand that delivers
a poor experience will probably have to work extremely hard to do business
with that customer again. If it’s a complete disaster, that experience might
push that customer out of the category entirely, and risks their dissatisfaction
being discoverable to other potential customers in the form of negative
reviews or comments on social media.
CHAPTER 224
IDENTIFYING THE MESSY MIDDLE
Triggers and purchase
It might seem odd to cover these critical points in our model as an
afterthought, but as they strictly occur beyond the boundaries of the messy
middle, our research doesn’t touch on them directly.
Sufce to say that triggers are responsible for moving consumers from a
passive state into an active purchase state. We’ve made them plural in our
model to account for the fact that it is often not just one inciting factor that
prompts the desire to purchase. In many cases an interconnected set of
internal and external factors – feelings and memories, ads, and reminders –
are responsible for triggering an active purchase state.
CHAPTER 225
IDENTIFYING THE MESSY MIDDLE
The messy middle identified
In this chapter, we’ve gone from a vague hunch to an updated model of
consumer decision-making, via some intriguing hints from Google Trends
and a long reading list of behavioural science. In the next, we’re going to
nd out what else search data can tell us about the evolution of behaviour
on the internet.
Q: What exactly is the purpose of the model?
A: It labels the specic cognitive inputs and mental modes that consumers
engage when processing vast amounts of information and managing choice.
It also illustrates the relationship between those inputs and modes. In short, it
helps to make sense of what’s going on in the messy middle.
Q: How is this model new?
A: Given that it brings together various elements of previous models and
theories, it isn’t entirely new. But it does effectively illustrate the non-linear reality of
decision-making – such as the constant backdrop of exposure and the innitely
looping relationship between explore and evaluate.
Q: Do the older models still have value?
A: Not all models are built with the same purpose in mind. We wanted to focus
specically on delineating consumer behaviour, while other models give greater
focus to branding, loyalty, and the role of habit and impulse.
Q: Is the funnel dead?
A: Our model is designed to reflect the complex way that people make decisions.
As such, it is tightly focused on the consumer, rather than on marketing or sales
processes. As a tool for formulating marketing objectives, the funnel is still very much
alive. In fact, at 120 years old and counting, the funnel is quite possibly immortal.
Model FAQs
3
Searching for clues in Google Trends
With a hypothesis now in place, our next step was to return to the Google
Trends data to see if we could nd real-world evidence of behaviour changing
over time on the web.
At a high level, people use search to look for information about a particular
subject or object. But because the amount of available information is so vast,
searches are often modied with an additional word or phrase that describes
what it is the searcher wants to know about the thing they’re searching for. In
our search data, fads, trends, and memes blip in and out of popularity, but the
way people use search has slowly increased in range and complexity over time.
If you’re looking for a laptop, you might prefer to narrow down your search by
modifying it to “best laptop” or “cheap laptop”, or even “laptops near me”. The
modiers people use can’t always be neatly broken down into exploration or
evaluation, but if we trend the use of these sorts of terms, we can nd clues
that illuminate how behaviour has changed.
For those not familiar with Google Trends, here’s a quick primer on how it works,
and a few clarifying notes on what the charts used in this chapter represent.
Investigating the
messy middle
CHAPTER 327
INVESTIGATING THE MESSY MIDDLE
All charts in this chapter have been generated in Google Trends and,
because it is a publicly available tool, all the charts can be easily recreated.
To make comparisons between terms easier, Google Trends normalises
search data by time period, location, and topic. It therefore displays the
relative popularity of a term over time, not absolute.
Numbers represent search interest relative to the highest point on the
chart for the given region and time. A value of 100 is the peak popularity for
the term.
Due to a change in the methodology of Google Trends on 1st January 2011,
relating to improved geographical assignment, the majority of the charts
we feature in this chapter begin on this date.
By using double quotation marks around search terms, for example
“giftideas”, the results include the exact phrase, possibly with words
before or after, like “birthday gift ideas”.
Search Tips for Trends is a must-visit for anyone wanting to have a play
around in Google Trends.
The way people use search has
slowly increased in range and
complexity over time.
Google Trends
CHAPTER 328
INVESTIGATING THE MESSY MIDDLE
Are you not enteained? The slow demise of ‘free’
In the last chapter, we discussed the strange intersection of “best”and
“cheap” in our search data. But these two modiers weren’t all we
were searching for back then. For the rst decade and a half of the new
millennium, it seems that we were keen to get something for nothing.
Evenmore than “best” and “cheap”, in the 2000s “free” was king (gure 1).
Figure 1
0
25
50
75
100
2004 2006 2008 2010 2012 2014 2016 2018 2020
SEARCH INTEREST
cheap
Search term
best
Search term
free
Search term
Data source: Google Trends, United Kingdom, 1st January 2004–1st January 2020, All categories, Web Search
The proportion of UK searches containing “free” or “cheap” has been in decline, but the proportion containing “best”
has been increasing.
CHAPTER 329
INVESTIGATING THE MESSY MIDDLE
As we fast forward through the past 16 years, the relative volume of
entertainment searches containing “free” gradually diminishes and today,
in2020, the frequency of these expressions is far lower in relative terms.
The demise of “free” is partly a story about our changing search behaviour
but, of course, we can’t forget that it’s also a reflection of how new platforms
and streaming services have changed the entertainment industry. In 2004
there was no YouTube (founded 2005), no Spotify (founded 2006), Netflix was
still a DVD sales and rentals business (it didn’t offer streaming until 2007),
and there was no App Store (launched 2008).
Figure 2
0
25
50
75
100
2004 2006 2008 2010 2012 2014 2016 2018 2020
SEARCH INTEREST
free games
Search term
free music
Search term
free movies
Search term
However, appearances can be deceiving. When we look at the kinds of
queries that contain these modiers, we begin to see some revealing
patterns. In the 2000s we didn’t search for free everything. For the most part,
we wanted free entertainment: games, music, and movies (gure 2).
Data source: Google Trends, United Kingdom, 1st January 2004–1st January 2020, All categories, Web Search
Declining UK search interest for entertainment queries containing “free”.
CHAPTER 330
INVESTIGATING THE MESSY MIDDLE
The demise of “free” and the category-specic details that add nuance to
that narrative, serve as a cautionary tale for the rest of this chapter. As we
look at other search modiers, in each case we have to bear in mind that the
same word can have different meanings in different categories, countries,
and languages.
Figure 3
0
25
50
75
100
2004 2006 2008 2010 2012 2014 2016 2018 2020
SEARCH INTEREST
free
Search term
gluten free
Search term
sugar free
Search term
dairy free
Search term
meat free
Search term
The lesson of ‘free’ – a warning to the curious
This isn’t to say that “free” no longer features in searches today. It still
represents signicant volume, but the types of free things we are looking
for have evolved, and the composition of search queries containing “free”
helps us to understand that evolution. For example, if we limit our analysis
to the food and drink category, we see that when people use the word “free”
in a search they tend to be looking for items that are “free from” a specic
ingredient or allergen (gure 3).
Data source: Google Trends, United Kingdom, 1st January 2004–1st January 2020, Food & Drink category, Web Search
Rising UK search interest for the term “free” in the food and drink category.
CHAPTER 331
INVESTIGATING THE MESSY MIDDLE
OK Google, let’s go shopping
Now let’s take a look at modiers and categories within a more
commercialcontext.
We’ll look at seven main search modiers: “ideas”, “best”, “difference
between”, “cheap”, “deals”, “reviews”, and “discount codes”.
The order in which these seven modiers are listed is intentional. While it
might not be possible to strictly classify a search query as either exploratory
or evaluative, we can at least hypothesise that some searches have a more
expansive, information-gathering intention, and others are more reductive
and clarifying.
best
ideas reviews
deals
cheap
difference between
discount codes
CHAPTER 332
INVESTIGATING THE MESSY MIDDLE
OK Google, inspire me with ideas
We begin with the modier “ideas”, which has been gradually increasing its
share of UK search over time (gure 4).
“Ideas” is arguably the most expansive of the seven modiers on our list. Located
rmly within exploration territory, searchers employ this term when seeking new
information, inspiration, and brands to add to their consideration sets.
Figure 4
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
0
25
50
75
100
SEARCH INTEREST
ideas
Search term
Data source: Google Trends, United Kingdom, 1st January 2011–1st January 2020, All categories, Web Search
An upward trend in UK searches containing “ideas”, spiking each year at Halloween.
CHAPTER 333
INVESTIGATING THE MESSY MIDDLE
It is likely that searches containing “ideas” will often follow on quickly from
one or more triggers, such as a pressing need to identify and buy a gift for
somebody. While “Christmas gift ideas” is the largest phrase by volume, as
seen in gure 5, we also turn to Google to help us with birthday gift ideas
(afairly flat pattern given birthdays occur all year round) and anniversary gift
ideas (on close inspection peaking in May-September each year, aligned with
the fact more couples in the UK marry in summer months than in winter).
8
Figure 5
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
0
25
50
75
100
SEARCH INTEREST
christmas gift ideas
Search term
birthday gift ideas
Search term
anniversary gift ideas
Search term
8 UK Ofce of National Statistics, https://www.ons.gov.uk/visualisations/dvc360/index.html
Data source: Google Trends, United Kingdom, 1st January 2011–1st January 2020, All categories, Web Search
Popular UK search queries containing “gift ideas”.
CHAPTER 334
INVESTIGATING THE MESSY MIDDLE
Figure 7
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
0
25
50
75
100
SEARCH INTEREST
bedroom ideas
Search term
bathroom ideas
Search term
kitchen ideas
Search term
living room ideas
Search term
When it comes to our homes, there isn’t a room in the house where we’re not
seeking ideas, inspiration, and new additions (gure 7).
Data source: Google Trends, United Kingdom, 1st January 2011–1st January 2020, Food & Drink category, Web Search
Data source: Google Trends, United Kingdom, 1st January 2011–1st January 2020, Home & Garden category, Web Search
Figure 6
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
0
25
50
75
100
SEARCH INTEREST
breakfast ideas
Search term
lunch ideas
Search term
dinner ideas
Search term
We also increasingly turn to Google to give us ideas of what to rustle up for our
next meal (gure 6).
Rising mealtime “ideas” searches in the food and drink category in the UK.
Rising popularity of “ideas” searches in the home and garden category in the UK.
CHAPTER 335
INVESTIGATING THE MESSY MIDDLE
OK Google, what’s best?
Of all our seven modiers, “best” has the widest footprint across categories.
People use “best” as a modier in searches for everything from ironing
boards to insurance, from TV sets to travel destinations.
Since we’ve already spent some time thinking about “best” in earlier sections,
we won’t repeat those lessons here. However, there’s one further insight
worth bearing in mind: it can be challenging to denitively label “best” as a
signier of exploration or evaluation. At rst we might assume it to be strictly
evaluative – after all, to ask what’s “best” implies a side-by-side comparison.
But on closer inspection, it turns out that “best” is also being used to explore
categories in conjunction with more generic search terms.
People use ‘best’ as a modier
in searches for everything from
ironing boards to insurance, from
TV sets to travel destinations.
An interesting feature of searches containing “ideas” is that, compared with
other modiers, the term they appear alongside is rarely the name of a brand
or retailer. In 2019, less than 5% of UK searches for gift ideas, meal ideas, and
room ideas also contained the name of a brand or retailer.
9
This supports the
hypothesis that “ideas” searches are associated with an exploration mindset
– people are adding information, products, and brands into their thinking, not
evaluating between a shortlist of known products, brands, and retailers.
‘Ideas’ searches are associated
with an exploration mindset.
9 Google internal data, UK, 2019
CHAPTER 336
INVESTIGATING THE MESSY MIDDLE
For example, when we look at overall searches for the “best” restaurants and
places to eat, as well as the best bars and pubs, we can see that these are all
consistently growing over time (gure 8).
And in the example from the travel and tourism category, all of these “best”
searches are recognisably exploratory in nature (gure 9).
Data source: Google Trends, United Kingdom, 1st January 2011–1st January 2020, All categories, Web Search
Data source: Google Trends, United Kingdom, 1st January 2011 - 1st January 2020, Travel category, Web Search
Figure 9
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
0
25
50
75
100
SEARCH INTEREST
best places
Search term
best beaches
Search term
best hotels in
Search term
best things to do
Search term
Rising UK search interest in the travel category for the “best” places to visit.
Figure 8
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
0
25
50
75
100
SEARCH INTEREST
best places to eat
Search term
best restaurants
Search term
best bars
Search term
best pubs
Search term
Rising UK search interest in the “best” places to eat and drink.
CHAPTER 337
INVESTIGATING THE MESSY MIDDLE
OK Google, what’s the difference?
Food – and new food trends in particular – can often be a source of
confusion for consumers. So it’s not surprising that we often ask Google to
explain the “difference between” two products (gure 10).
For example, we ask for help understanding the differences between pairs
of related items: cappuccino and latte, lager and beer, gelato and ice cream,
baking powder and baking soda, sultanas and raisins, fromage frais and
creme fraiche, whisky and bourbon, vegetarian and vegan, champagne and
prosecco, cacao and cocoa, paella and risotto.
This trend in particular is suggestive of both expanding choice in the messy
middle, and of consumer desire for information that claries and reassures.
We oen ask Google to explain the
‘dierence between’ two products.
Data source: Google Trends, United Kingdom, 1st January 2011–1st January 2020, Food & Drink category, Web Search
Figure 10
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
0
25
50
75
100
SEARCH INTEREST
difference between
Search term
Rising UK search interest in the food and drink category to understand the “difference between” two items.
CHAPTER 338
INVESTIGATING THE MESSY MIDDLE
OK Google, I want the best trip, but I want it cheap too
The story of “cheap” is largely a story about travel and tourism, even in 2020
when these categories were severely disrupted by the coronavirus pandemic.
Of the top 10 UK search queries since 2011 including or related to “cheap”,
seven were denitively from the travel and tourism category (gure 11).
Top 10 searches related to the term ‘cheap’
cheap ights, ights, cheap holidays, cheap hotels, cheap
tickets, cheap holiday, cheap insurance, cheap ight, cheap
train tickets, cheap cars
Figure 11
Looking at “best” and “cheap” travel searches side by side, they mirror the
same pattern visible at the aggregate level (gure 12).
Data source: Google Trends, United Kingdom, 1st January 2011 - 1st January 2020, All categories, Web Search
Data source: Google Trends, United Kingdom, 1st January 2011–1st January 2020, Travel Category, Web Search
Figure 12
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
0
25
50
75
100
SEARCH INTEREST
cheap
Search term
best
Search term
Travel category searches in the UK more often contain “cheap” than “best” but these trends have been converging.
cheap
CHAPTER 339
INVESTIGATING THE MESSY MIDDLE
And as we saw earlier, with “best” sometimes occupying an exploratory
function when used alongside a generic search term, these two modiers
capture both sides of the exploration and evaluation loop.
On the one hand, we perform exploratory searches to determine the most
appealing destinations, eateries, and activities. For example, the upward
trends we noted for “best places”, “best beaches”, “best hotels in”, and “best
things to do”.
On the other hand, we appear determined to pay as little as possible for our
transport to get there and our accommodation once we arrive, frequently
modifying our travel searches with “cheap” (albeit with “cheap” featuring in a
decreasing proportion of travel searches over time, gure 13).
Figure 13
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
0
25
50
75
100
SEARCH INTEREST
cheap flights
Search term
cheap holidays
Search term
cheap hotels
Search term
cheap train tickets
Search term
cheap all inclusive
Search term
It’s notable that travel searches including “cheap” rarely contain the names
of brands. While it might be tempting to assume that “cheap” searches are
purely evaluative, the absence of brands shows us that these might often
also be exploratory.
Data source: Google Trends, United Kingdom, 1st January 2011–1st January 2020, Travel category, Web Search
The term “cheap” appears in a range of popular searches in the travel category in the UK.
CHAPTER 340
INVESTIGATING THE MESSY MIDDLE
OK Google, show me the bargains
Three modiers used in similar categories and for similar purposes are
“deals”, “offers”, and “sale”.
“Deals” is especially common in the internet and telecom sector. We use this
modier to seek value when both exploring and evaluating broadband, phone
contracts, and TV subscriptions (gure 14).
Figure 14
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
0
25
50
75
100
SEARCH INTEREST
broadband deals
Search term
sim only deals
Search term
tv deals
Search term
It’s much harder to broadly characterise searches containing “deals” as
explore or evaluate based on the presence of brands. Both are commonly
used with and without brand names, implying that people are looking
expansively for information and new brands, as well as critically evaluating
the deals on offer from the brands they are considering.
Data source: Google Trends, United Kingdom, 1st January 2011–1st January 2020, Internet & Telecom category, Web Search
A rising proportion of UK searches in the internet and telecoms category contains the term “deals”.
CHAPTER 341
INVESTIGATING THE MESSY MIDDLE
Our other modiers of this nature exhibit similar trends over time, albeit
with different category afnities. As “deals” is to telecoms, “offers” is to
grocery retail, especially with relatively high unit price purchases, such as
alcohol. Conversely, the word “sale” tends to be more associated with retail
categories such as clothing and furniture.
10
OK Google, does it have good reviews?
We’ve seen how the search modier “best” can help people nd out what
others consider to be worth doing or buying. An even more explicit way of
expressing a wish to investigate the views of others – be they peers, previous
buyers, vloggers, or category experts – is to include the modier “reviews”
in a search. However, there’s an interesting distinction between searches
containing “best” and “reviews”, with searches containing “best” rarely
containing the name of a brand, while searches containing “reviews” often do.
Top 15 searches related to ‘sim only deals’
ee sim only deals, ee sim only, sim only vodafone, vodafone
sim only deals, o2 sim only, o2 sim only deals, sim only deals
virgin, virgin sim only, tesco sim only, sim only deals tesco,
best sim only deals uk, phone sim only deals, sim only deals
uk, mobile sim only deals, best sim only deals
Figure 15
For example, the top 15 “sim only deals” searches in the UK in 2019
comprised 11 with a brand name (in red) and 4 without (in pink, gure 15).
10 Note that “sale” and “for sale” are search modiers with very different meaning and usage. The term “sale” is associated with discounts
and price reductions offered by a business, whereas “for sale” typically denotes second-hand or private sales.
Data source: Top 15 queries related to “sim only deals”, Google Trends, United Kingdom, 2019, All categories, Web Search. Excludes related
queries not containing “sim only”
sim only deals
CHAPTER 342
INVESTIGATING THE MESSY MIDDLE
Reviews are also a prevalent feature of the automotive category, with a
similar trend visible in the combination of modiers and brands (along with a
few authoritative websites and magazines). The modier “review” provides
us with a clear example of shoppers actively seeking out authoritative
viewpoints to boost condence during decision-making.
Unlike “best”, the fact that “review” searches typically contain the names of
specic brands and products hints that review searches on the whole are
more evaluative than exploratory. In many cases it appears that people have
one or more potential brands and models in mind, and are looking for further
information to help evaluate which would be the better choice.
As these and similar examples from this chapter show, the presence of
brands is often evaluative, especially in conjunction with a specic product
name. However, it is important to note that by itself, the presence of a brand
term in a query is insufcient to signal the shopper’s mental mode.
Top 10 queries related to ‘review’
laptop review, headphones
review, ipad review, ps4 review,
kindlereview, kindle re review,
lenovo yoga review, galaxy note
review, microsoft surface review,
kindle re hd review
Top 10 queries related to ‘best’
best apps, best camera,
bestlaptop, best tv, best pc,
bestheadphones, besttablet,
bestipad, best laptops,
bestspeakers
Figure 16
For example, if we look at the top UK search queries of the past 10 years in
the computers and electronics category, of the top 10 related to “best”, only
one contains the name of a manufacturer or product brand (“best ipad”). In
contrast, all but two of the top 10 search queries related to “review” contain
the name of a company or product brand, with the exceptions being the top
two results (“laptop review” and “headphones review”, gure 16).
Data source: Google Trends, United Kingdom, 1st January 2011–1st January 2020, Computers & Electronics category, Web Search.
Negative keyword “-buy” applied to searches containing “best”.
review best
CHAPTER 343
INVESTIGATING THE MESSY MIDDLE
OK Google, can I get any money off?
Our nal search modier is “discount code”, although we’ll group this
together with its sibling, “promo code”.
These modiers have grown as a proportion of search over the past 10 years,
spiking in November and December (gure 17).
Figure 17
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
0
25
50
75
100
SEARCH INTEREST
discount code
Search term
promo code
Search term
Data source: Google Trends, United Kingdom, 1st January 2011 - 1st January 2020, All categories, Web Search
A rising proportion of UK searches contains “discount code” or “promo code”.
CHAPTER 344
INVESTIGATING THE MESSY MIDDLE
The majority of searches containing “discount code” also contain the name
of a retailer. For example, of the top 15 related queries for these terms in the
UK in 2019, 11 (inred) contained the name of a retailer (gure 18).
Data source: Top 15 queries related to “discount code”, Google Trends, United Kingdom, 2019, All categories, Web Search
Top 15 queries related to ‘discount code’
asos discount code, argos discount code, amazon discount
code, jd discount code, debenhams discount code, boots
discount code, boohoo discount code, currys discount
code, ebay discount code, john lewis discount code, next
discountcode, just eat discount code, tui discount code,
travelodge discount code, nike discount code
Figure 18
The presence of a named retailer in these searches implies that little
exploration is happening here, and that the evaluative phase may be nearing
an end too. As such, these modiers place us as close as search gets to the
moment of purchase.
Modifying the messy middle
As these examples show, our relationship with the things we search for is
complex and mutable. But the search modiers people use are a rich source
of insight into how our thinking and behaviour have evolved over time, and
can even offer clues about the underlying cognitive biases at work. In the
next chapter, we’ll take some of these insights and attempt to quantify the
impact that specic biases can have on decision-making.
discount code
CHAPTER 345
INVESTIGATING THE MESSY MIDDLE
‘Best’ and ‘cheap’ around the world
Although we need to account for how the meaning of words can differ
between categories, many of the trends we identied in UK data are also
visible in other geographies and languages.
This is how the pattern of “best”, “cheap”, and “free” looks for all searches
globally in English (below).
Figure 19
2004 2006 2008 2010 2012 2014 2016 2018 2020
0
25
50
75
100
SEARCH INTEREST
cheap
Search term
best
Search term
free
Search term
Data sources: Google Trends, Worldwide, 1st January 2004–1st January 2020, All categories, Web Search
Similar long-term trends for “best” and “cheap” are also visible when we
translate those terms into the native languages of many other countries:
The proportion of worldwide searches containing “free” or “cheap” has been in decline, but the proportion containing
“best” has been increasing.
CHAPTER 346
INVESTIGATING THE MESSY MIDDLE
The lines for “best” and “cheap” don’t always cross as they do in the UK: for
example in the US, searches including “best” have been more frequent from the
beginning. However “best” has still seen a steady rise in the US, and “cheap” a very
gradual decline, making it broadly consistent with the UK and other countries.
cheap
Search term
best
Search term
2004 2012 2020
0
25
50
75
100
2004 2012 2020
0
25
50
75
100
2004 2012 2020
0
25
50
75
100
2004 2012 2020
0
25
50
75
100
2004 2012 2020
0
25
50
75
100
2004 2012 2020
0
25
50
75
100
2004 2012 2020
0
25
50
75
100
2004 2012 2020
0
25
50
75
100
2004 2012 2020
0
25
50
75
100
SEARCH INTEREST
SEARCH INTEREST
SEARCH INTEREST
India Germany Spain
France Italy Brazil
United Kingdom United States Australia
Data source: Google Trends, various countries, 1st January 2004–1st January 2020, All categories, Web Search
Figure 20
Search trends for “cheap” and “best” around the world, translated into local languages.
4
The next stage of our research involved taking what we’d learned from our
literature reviews, Google Trends data, and shopping observations, and applying
it in an experimental setting. Over the course of 310,000 simulated purchase
scenarios, we tested the impact that various behavioural biases can have on
shoppers’ brand preferences.
Homo-not-so-economicus
As theories about “economic man” have given way to metaphors about ridersand
elephants, it would seem that most behavioural scientists now agree that, in
reality, our decision-making apparatus comprises both reason and emotion.
Inuencing the
messy middle
Even a seemingly functional, low‑cost
purchase like buying a favourite
shampoo can be prompted by emotional
or rational considerations.
CHAPTER 448
INFLUENCING THE MESSY MIDDLE
In the context of shopping decisions, we might be tempted to propose that
the degree of rationality increases with the size and importance of the
purchase. But as anyone who has ever bought a car, a house, or an expensive
holiday knows, the moment the deal closes can still be fraught with complex
emotion. And at the other end of the scale, even a seemingly functional,
low-cost purchase like buying a favourite shampoo can be prompted by
emotional or rational considerations, depending on the individual.
And of course, muddying the water of reason and emotion further is
advertising – particularly branding. Brands often seek to cultivate an emotional
connection with consumers – in fact, many people will openly describe
themselves as loving or hating a particular brand. These associations, often
bound up in our sense of ourselves and our aspirations for who we want to be,
are a powerful source of behavioural change in themselves.
To design an experiment looking at how behaviour is influenced during the
crucial explore and evaluate phases of our model, we needed to draw up a
list of behavioural science biases to test. For this, The Behavioural Architects
returned to the literature of academic behavioural science. Over the course of
more than 50 years, the discipline has codied some 300 principles that explain
the conscious and unconscious workings of the human mind. Of course, not all
of the 300 are relevant to the kind of decision-making we’re exploring here, so
during a thorough review, the team whittled down the list to six biases that are
closely associated with the explore and evaluate phases of our model.
CHAPTER 449
INFLUENCING THE MESSY MIDDLE
A summary of six biases
The names we’re using for these biases may or may not be familiar to you,
but the underlying denitions are congruent with those used in academic
behavioural science. Of course, you may well have used some of them in
your own campaigns, or recognise them at work in the ads of one of your
competitors or favourite brands.
Category heuristics are shortcuts or rules of thumb that aid us in making a
quick and satisfactory decision within a given category. An example would be
focusing on how many megapixels (MP) the camera has when purchasing
a smartphone or how many gigabytes (GB) of data are included in a mobile
phone contract.
Princeton psychologists, Shah and Oppenheimer,
11
found heuristics reduce
cognitive effort through the following impacts on decision-making:
Examining fewer pieces of information
Relying on easy-to-access pieces of information
Simplifying the weighting of information
Integrating less information in a decision process
Considering fewer alternatives overall
Authority bias describes the tendency to alter our opinions or behaviours to
match those of someone we consider to be an authority on a subject. When
we’re unsure, we tend to follow the lead of people we believe to be credible
and knowledgeable experts, and therefore may use an authority view as a
mental shortcut. In one experiment, the brains of 24 college students were
scanned while making nancial decisions. If students received advice from
a renowned economist, the scans showed that the decision-making parts of
students’ brains showed less activity as the students “offloaded” the burden
of the decision process to the expert.
12
2.
11
Shah, A. K., & Oppenheimer, D. M. (2008). Heuristics Made Easy: An Effort-Reduction Framework. Psychological Bulletin, 134(2), 207–222
12 Engelmann J. B., Capra C. M., Noussair, C., & Berns G. S. (2009). Expert Financial Advice Neurobiologically “Offloads” Financial Decision-
Making under Risk. PLoS ONE 4(3): e4957. https://doi.org/10.1371/journal.pone.0004957
1.
CHAPTER 450
INFLUENCING THE MESSY MIDDLE
Social proof posited by psychologist Robert Cialdini
13
describes the tendency
to copy the behaviour and actions of other people in situations of ambiguity
or uncertainty. The internet has digitised word-of-mouth reviews and
recommendations, making it much easier for people to rely on social proof
as a shortcut for decision-making. Sometimes we’re conscious of this, for
example if we take the time to read consumer reviews, but often we’re influenced
unconsciously. For example, without thinking, we might click on an ad that
includes a four- or ve-star rating, drawn to what appears to be a popular choice.
Power of now describes the fact that we tend to want things now rather
than later. Humans are wired to live in the present – our evolutionary survival
hinged on our ability to deal with the problems of the here and now rather
than our ability to plan for the future. This explains why people often nd it a
challenge to save for their future.
14
“Power of now” also explains the success of
instant downloads or 24-hour delivery versus having to wait to get a product.
15
Scarcity bias is based on the economic principle that rare or limited resources
are more desirable. As Robert Cialdini states: “The scarcity principle trades on
our weakness for shortcuts”.
16
Scarcity typically takes one of three forms:
Time limited: when there is a time limit to a product’s
availability, it creates a deadline that makes people act
before the time is up.
Quantity limited: limited or rare supplies are perceived by
people as a threat to their freedom of choice, triggering a
reaction to ght the threat and maintain their access to the
resource.
Access limited: meaning limited access to features like
information, groups, or spaces. Censorship makes people
place a higher value on restricted features because
exclusivity makes them feel special.
4.
5.
13 Caldini, R. B. (1984). Influence – The Psychology of Persuasion. Collins. 14 Thaler, R. T. (1991). “Some Empirical Evidence on Dynamic Inconsistency” in Richard H. Thaler,
ed.. Quasi Rational Economics. New York: Russell Sage Foundation, 127–33. 15 The scientic name for “power of now” is discounting the future, which the economist Richard
Strotz explored in 1955 with his work on hyperbolic discounting and time inconsistent preferences. 16 Caldini, R. B. (1984). Influence – The Psychology of Persuasion. Collins.
3.
CHAPTER 451
INFLUENCING THE MESSY MIDDLE
Power of free describes the fact that there is something special about the
price of zero. The demand for a product or service is signicantly greater at
a price of exactly zero compared to a price even slightly greater than zero.
In his book “Predictably Irrational”, behavioural economist Dan Ariely writes
about a study in which people were given the option to choose between two
offers. One was a free $10 Amazon gift card, the other a $20 gift card that
could be bought for only $7. More people chose the $10 gift card, despite the
other option offering superior value.
17
The power of free can be thought of
as an emotional hot button – a source of irrational excitement that can be
critical in persuading a consumer to make a purchase decision.
While certainly not a denitive list of every bias in play, our set of six
represents several of the most powerful principles identied in the literature,
all of which are suitable for testing at scale. It also has the advantage of
covering implementations that range from simple copy changes to more
complex merchandising and logistical decisions.
Testing the six biases
The biases identied by The Behavioural Architects have been thoroughly
examined in an academic context, but to gauge their importance to
marketers we knew we would have to place them within a purchase-making
context to see how they affect the emotional weight of competing brands.
The previous experimental results we reviewed were often from relatively
small samples, without a purchase or brand aspect, and not systematically
applied across different products and categories. So we set out to build
a method that would address these challenges: a shopping simulation
purpose-built to provide the insights marketers need.
As the basis of our experiment we chose to apply conjoint analysis – a statistical
technique much used and well understood in marketing to quantify the relative
importance people place on the different attributes of a product or service.
6.
17 Ariely, D. (2008). Predictably Irrational: The Hidden Forces That Shape Our Decisions. Harper.
CHAPTER 452
INFLUENCING THE MESSY MIDDLE
Typically a conjoint analysis compares the importance of a range of tangible
features or benets to a proposition, but the points of variance in our test
would be the presence and relative strength of the cognitive biases. Conjoint
studies can be delivered in a range of formats, but for our purposes we chose
to create a generic, unbranded website which would situate participants’
decision-making within a familiar context.
Before the simulation began, shoppers were asked to share their rst and
second favourite brands from a selection within a specic category. These
preferences then became the basis of the simulation, with the shoppers
asked to choose between pairs of brands to which some or all of our six
biases had been applied. Using this method, we were able to measure the
preference of brand versus brand on a level playing eld, and test the power
of each bias to switch preference from favoured to less-favoured brands.
We were able to measure the preference of
brand vs. brand on a level playing eld, and test
the power of each bias to switch preference
from favoured to less‑favoured brands.
A few limitations
There are, of course, a couple of real-world variables that our simulation
can’t account for. Price is often a determining factor in purchase decisions,
especially where there is a large degree of difference between options.
As such, the shoppers who participated in our research were told that the
products and services they were considering were priced at the current
expected market value, eliminating price as a variable.
The second complicating variable in the simulation is to do with brand
building. Once in the simulation, shoppers were exposed to full-colour
graphical logos of their preferred brands. Any pre-existing associations
between our shoppers and those brands (what our marketing model terms
“exposure”) remained active throughout the simulation.
CHAPTER 453
INFLUENCING THE MESSY MIDDLE
It’s a jungle out there
We argued at the start of this chapter that any purchase decision, from
choosing a mortgage to buying your favourite shampoo, can contain both
rational and emotional elements. In certain circumstances, rider and elephant
might eventually reach the same destination, but having made the journey via
very different routes.
To test whether the impact of brand preference and cognitive bias remains
stable across categories, we selected 31 products representing a broad
range of risk, complexity, and emotional and nancial investment, covering
several major verticals and sectors, including travel, nancial services,
consumer packaged goods, retail, and utilities (gure 1).
MORTGAGE
ENERGY PROVIDER
CAR INSURANCE
DETERGENT
MOISTURISER
&
CAR HIRE
SHAMPOO
WHISKY
PACKAGE HOLIDAY
FITTED KITCHEN
HOTEL
TV
CEREAL
CAT FOOD
ISA
(Individual Savings
Account)
CLOTHES
SHORT HAUL
FLIGHTS
LONG HAUL FLIGHTS
CHILDREN’S TOYS
CINEMA TICKETS
BROADBAND
SOFA
MOUNTAIN
BIKE
SUV (Sport Utility Vehicle)
LAPTOP
BATHROOM SUITE
MAKE-UP
MOBILE
PHONE
MOBILE
NETWORK
POWER
DRILL
CREDIT CARD
HIGH COMPLEXITY
LOW COMPLEXITY
LESS ENJOYABLE MORE ENJOYABLE
Figure 1
Matrix of product categories, showing perceptions of enjoyability and complexity.
Given the online purchase format of our conjoint experiment, we established
some broad criteria to qualify our shopper sample.
Source: Google / The Behavioural Architects. n=31,000 category buyers, online shoppers, aged 18–65 (31 categories, 1,000 respondents in each). Participants answered the
following questions (1–7 scale). Results were then grouped by factor analysis (questions 1 and 5 for “enjoyment”, question 2–4 for “complexity”) and plotted accordingly. 1. How
enjoyable do you nd looking for [relevant ‘product]? 2. How complex/difcult is it to nd the right [relevant ‘product]? 3. How much effort does it take to nd the [relevant ‘product]
you want? 4. How worried are you about making the wrong choice of [relevant ‘product]? 5. How experienced/knowledgeable do you feel about [relevant ‘product] in general?
CHAPTER 454
INFLUENCING THE MESSY MIDDLE
We wanted participants who were familiar with online shopping, so to
control for this we selected people who said they had shopped at the UK’s
largest online retailer. Likewise, we wanted shoppers who were familiar
with searching for products online, so we selected people who had used
the UK’s most popular search engine for that purpose. Together, these two
characteristics provided a broad, qualied sample of participants familiar
with the parameters and conventions of online shopping.
18
The nal and most important qualication was that every participant had
to be in-market for the product featured in their simulation, and intend to
purchase it within a timeframe appropriate for that category (in other words,
for car shoppers the applicable window would be longer than for someone
buying shampoo). We also excluded anyone who said they had already made
their mind up about exactly which product they were going to buy, to exclude
the possibility that participants might have already exhausted their capacity
for exploration and evaluation.
To ensure a robust sample size for each product, we recruited 1,000
shoppers in every category. This equated to several thousand shoppers per
sector, and a total sample of 31,000 in-market shoppers for macro-level,
cross-category analysis. Participation was remote, with each shopper
completing 10 purchase simulations within a given category, giving a total
of 310,000 purchase scenarios within which to analyse our six cognitive
biases. Because of the prejudicial effect of measuring the presence of a bias
againstthe absence of the same bias, we paired different levels of execution
ranging from strong to weak (for instance next-day versus seven-day
delivery, or ve-star versus three-star reviews).
We believe that our tests amply – and with statistical validity – demonstrate
the fluidity of preference between trigger and purchase. However, the results
of a simulation can only ever be indicative, and as such we don’t suggest that
anyone should treat our results or recommendations as a substitute for their
own rigorous, in-market testing.
So, with caveats and methodology taken care of, on to the experiments.
18
Respondents who never use Google Search or never use Amazon (2% of category buyers aged 18–65) were screened out before participating in the research.
CHAPTER 455
INFLUENCING THE MESSY MIDDLE
The simulation
The objective of these purchase simulations is to understand how marketing
effectiveness can be improved in the messy middle, using behavioural
science principles to either avoid or create disrupted brand preferences.
This translates into a threefold research objective:
Quantify and measure the importance of brand preference in
the messy middle
Quantify and measure the susceptibility of those preferences to
disruption through the application of cognitive biases
Understand how the above varies across different product
categories and verticals
Before the simulation began, each of our 31,000 shoppers were asked
for their rst-choice and second-choice brands. These preferences then
appeared on screen as in the example on the previous page (gure 2).
PREFER A PREFER APREFER B PREFER B
Examples of the simulation interface, taken from the laptops and package holidays categories.
Figure 2
The navigational conventions and layout of the site were modelled on
familiar retailers, but without any specic branding in the user interface.
The only brand signals the shopper received were those exposed within the
experimental frame itself.
1.
2.
3.
CHAPTER 456
INFLUENCING THE MESSY MIDDLE
Within the frame, shoppers were presented with two boxes, Prefer A and
Prefer B. During the simulation, these boxes contained eight smaller boxes,
which displayed the logos of the brands being tested and information about
the product that the shopper might nd during exploration. In our simulation,
all of this information was contained on one screen rather than being
revealed over the course of several sequential clicks and screens.
It was this supplementary product information to which our behavioural
science principles were applied during testing. For example, star ratings were
varied to test different applications of the social proof principle, or different
recommendation types to measure the importance of authority bias. Each of
the expressions featured in these information boxes had up to three levels of
intensity (for example three-star, four-star, and ve-star reviews) for comparison.
The expressions of our biases were modelled on real-world instances, but were
quite basic in their execution, lacking any sort of creative gloss.
With both brand logos and all relevant information in place, the shopper was
asked to choose which they preferred. They were instructed not to overthink
the decision, but to follow the same process of discernment they would when
making a real-life purchase. From the collated results, we’re able to measure
the impact of any single element or combination of elements, quantifying the
impact of each change as an increased or decreased share of preference for
the respective brand.
The power of showing up
Implicit in the structure of our experiment (and marketing in general for that
matter) is the idea that to take preference share away from a competitor
brand, you have to be present when consumers are deliberating.
This might seem obvious, but it’s such a fundamental point that we don’t
want its importance to be mistaken. And as we’ll see, there is surprising
power in just showing up at the right moment.
In our rst analysis of the simulation data, we compared rst- and second-
preference brands, with all other expressions of our biases statistically
controlled to remain neutral.
CHAPTER 457
INFLUENCING THE MESSY MIDDLE
In this chart we can see that when a second favourite brand was introduced
as an option, 30% of shoppers changed away from their rst preference.
Of course, for many shoppers the second choice brand also might be
positively associated with many of the factors mentioned above.
Figure 3
0
25
50
75
100
100
30
70
Stated 1st choice brand Introduction of 2nd choice brand
1st choice brand
2nd choice brand
Transfer of preference from rst choice to second choice brand after introduction of second choice brand,
car (SUV) category.
This example (gure 3) simulated a car purchase (for an SUV specically)
– a decision into which several considerations, such as safety, reliability,
efciency, and performance might reasonably intrude.
Source: Google / The Behavioural Architects. 10,000 simulated car (SUV) purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18–65.
Simply giving the shopper the option
to choose their second choice brand
was enough to entice 30% away from
their initial choice.
CHAPTER 458
INFLUENCING THE MESSY MIDDLE
It turns out that car insurance is also far from immune to the power of
showing up. In fact, the effect is even larger than that witnessed in the car
purchase simulation, with only two of the 31 categories in our experiment
being more prone to switching than car insurance.
Figure 4
0
25
50
75
100
100 40
60
Stated 1st choice brand Introduction of 2nd choice brand
1st choice brand
2nd choice brand
Transfer of preference from rst choice to second choice brand after introduction of second choice brand,
car insurance category.
But even bearing that caveat in mind, it is remarkable that, despite their
stated preference, and statistically controlling for the differences in other
variables, simply giving the shopper the option to choose their second choice
brand was enough to entice 30% away from their initial choice.
The car category is full of recognisable brands, so this result may in part
simply be down to two sets of powerful associations doing battle in the
shopper’s mind. But what if we look at another category, no less hotly
contested but with very different associated values and brand attributes?
Buying a car sits at one end of the spectrum of purchase complexity on our
product matrix, so let’s look at a related but less complex purchase - car
insurance (gure 4).
Source: Google / The Behavioural Architects. 10,000 simulated car insurance purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18–65.
CHAPTER 459
INFLUENCING THE MESSY MIDDLE
The extent of the impact on share of preference ranges considerably. On the
far left of the chart, just showing up delivered a relatively weaker share of
preference for second choice brands in the smartphone category (18%) than
those who were willing to switch their preference of bathroom suite brand (44%).
Figure 5
0
25
50
75
100
18
21
24
25
25
28
28
29
29
29
30
30
30
31
31
32
33
33
34
34
35
36
36
36
37
39
39
39
40
42
44
82
79
76
75
75
72
72
71
71
71
70
70
70
69
69
68
67
67
66
66
65
64
64
64
63
62
61
61
60
58
56
Smartphone
Children’s toy
Detergent
Cinema
Shampoo
Mobile network provider
Cat food
Whisky
Energy provider
Laptop
TV
Broadband provider
Car (SUV)
Face moisturiser
Flights - long haul
Hotel
Mortgages
Cereal
Make-up
Package holiday
Credit card
Clothing
Flights - short haul
Power drill
Fitted kitchen
ISA
Mountain bike
Sofa
Car insurance
Car hire
Bathroom suite
1st choice brand
2nd choice brand
Transfer of preference from rst choice to second choice brand after introduction of second choice brand,
all categories.
According to our product matrix, the purchase of car insurance is not just less
complicated than a car purchase, it is also less enjoyable. These characteristics
might partly explain the increased impact of the introduction of the second
choice brand, as it suggests the purchase requires lower levels of engagement
and therefore is more prone to switching. Nevertheless, the results are stark.
Below is a chart showing all of the products in our experiment (gure 5),
ordered according to the size of the impact on share of preference when
shoppers were offered the choice of a second brand (the yellow portion
shows the share seized by the second favourite brand when exposed).
Source: Google / The Behavioural Architects. 310,000 simulated purchases. n=31,000 category buyers, online shoppers, aged 18–65
(31 categories, 1,000 respondents in each). Numbers may not add to 100 due to rounding.
CHAPTER 460
INFLUENCING THE MESSY MIDDLE
What this chart shows is the likelihood across categories that shoppers
will switch from their stated rst choice brand to their second choice, when
presented with both as options. However, since each brand within a category
will have a different level of resilience, the chart cannot be used to predict
the extent to which any individual brand will be susceptible to transfer of
preference to a competitor.
Looking at performance across verticals reveals a couple of interesting
patterns. The favourite consumer packaged goods brands were broadly less
susceptible to the presence of another brand in our simulations than utilities
like mobile network, broadband, and energy supplier. General retail products
such as children’s toys, laptops, TV, clothing, and sofas are scattered
throughout, while nancial services products (mortgage, credit card, ISA,
car insurance) generally sit towards the right-hand side, with a greater
susceptibility to preference switching.
Social proof: people respond to people
Having established a baseline for switching preference without variation in
any of the cognitive biases, we wanted to see what degree of preference shift
could be achieved by applying the principles of behavioural science identied
in our literature review.
In nearly every case, social proof (expressed as three-star versus ve-star
reviews) proved to be the most powerful behavioural bias, having either
the largest or second-largest effect in 28 of the 31 categories we tested.
19
Therefore we’re going to state this upfront, and then quickly move beyond it
to look at some of the more nuanced, category-specic examples.
Giving people evidence that other shoppers have already had a positive
experience with a brand, product, or service is extremely persuasive. The
gold standard of social proof – reviews and comments – can be difcult for
marketers to create out of nothing, as it relies on customers sharing their
post-purchase experience. However social proof, when it exists, can also be
evoked simply and powerfully through claims in copy, such as “the nation’s
favourite” or “the popular choice”.
19 In each instance, different average review scores between three stars and ve stars were compared with the total number of reviews for each brand remaining equal.
CHAPTER 461
INFLUENCING THE MESSY MIDDLE
The low‑hanging fruit of behavioural science
Many of the biases we tested are even easier to execute, requiring neither
large volumes of customer ratings, nor a memorable way with words.
In fact, several can be implemented through basic copy and design
modifications alone.
Category heuristics
Category heuristics are powerful and relatively simple to implement. In our
simulation, they achieved the largest or second-largest effect in 14 of 31
categories. In the scientic literature, category heuristics are dened as
shortcuts or rules of thumb that help people make decisions – vital pieces of
information that help clarify our options, such as the amount of memory in a
laptop or the number of carats in a diamond.
Many of the biases we tested
are even easier to execute,
requiring neither large volumes
of customer ratings, nor a
memorable way with words.
CHAPTER 462
INFLUENCING THE MESSY MIDDLE
To make effective use of category heuristics, marketers need to understand
which characteristics consumers most associate with a given product or
service. This is often also the characteristic they value most. For example,
when we looked at broadband, we found that highlighting data allowances
achieved the largest transfer in share of preference away from the initial
favourite brand (gure 6).
Figure 6
0
25
50
75
100
30
70
57
43
Stated 1st choice brand Introduction of 2nd choice brand 2nd choice + stronger category heuristic
100
1st choice brand
2nd choice brand
Category heuristics tested: “unlimited monthly usage” and “dedicated customer service”. Transfer of preference
from rst choice to second choice brand – category heuristics analysis, broadband provider category.
Source: Google / The Behavioural Architects. 10,000 simulated broadband provider purchase scenarios. n=1,000 category buyers,
UK online shoppers, aged 18–65.
CHAPTER 463
INFLUENCING THE MESSY MIDDLE
Figure 7
Figure 8
0
25
50
75
100
40
60
68
32
Stated 1st choice brand Introduction of 2nd choice brand 2nd choice + stronger category heuristic
100
0
25
50
75
100
33
67
49
51
Stated 1st choice brand Introduction of 2nd choice brand 2nd choice + stronger category heuristic
100
1st choice brand
2nd choice brand
Category heuristics tested: “no claims protection” and “autorenewal not required” (car insurance). Transfer of
preference from rst choice to second choice brand – category heuristics analysis, car insurance category.
Category heuristics tested: “28 month xed rate” and “5% deposit” (mortgages). Transfer of preference from rst
choice to second choice brand – category heuristics analysis, mortgage category.
Category heuristics also proved to be a decisive factor in the nance vertical,
achieving the greatest transfer in share of preference for both mortgages
and car insurance categories. In these highly structured products, our
simulations show that consumers are particularly attuned to look for
characteristics such as the duration of a xed rate or the treatment of
no-claims status (gures 7 and 8).
Source: Google / The Behavioural Architects. 10,000 simulated mortgage purchase scenarios, 10,000 simulated car insurance purchase scenarios. n=1,000
mortgage buyers / 1,000 car insurance buyers, UK online shoppers, aged 18-65.
CHAPTER 464
INFLUENCING THE MESSY MIDDLE
Scarcity bias
Scarcity messaging is perhaps one of the more immediately recognisable
executions of behavioural science in our list. However, in our simulations
it was most often the least effective bias. While it can be effective as a
clinching factor during nal evaluation, for exploring shoppers scarcity could
feel restrictive and provoke a negative reaction.
Figure 9
0
25
50
75
100
30
70
43
57
Stated 1st choice brand Introduction of 2nd choice brand 2nd choice + stronger authority bias
100
1st choice brand
2nd choice brand
Sources of authority tested: Which? and TechRadar. Transfer of preference from rst choice to second choice
brand – authority bias analysis, TV category.
Authority bias
Although our simulation shows it to be less powerful than its close cousin,
social proof, authority bias is still a very effective way to reassure shoppers
through citation of awards and expert reviews. This proved particularly
effective in categories where consumers might feel at a disadvantage
through lack of domain-specic knowledge, such as home furnishing, home
improvement, and electronics. Unsurprisingly, our simulation also found
that when it comes to authority, the endorsement of a publication known
to be impartial tended to carry more weight than a review from an industry
publication (gure 9).
Source: Google / The Behavioural Architects. 10,000 simulated TV purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18–65.
CHAPTER 465
INFLUENCING THE MESSY MIDDLE
Cross‑functional implementations
We also tested a selection of more involved biases. Implementing these will
require collaboration across functions, particularly where increased costs
are likely to be incurred.
The power of free
Giving something away isn’t always the sole discretion of the marketing
department, so capitalising on the power of free will probably involve buy-in
from other departments such as nance and merchandising. However, the
effort is likely to be rewarded, as our simulation ndings show that the power of
free can be a major influence on behaviour, having either the largest or second-
largest effect on transfer of preference in 18 out of 31 categories.
In the car hire category, we tested the power of free by boosting the
shopper’s favourite brand with a free car clean, while the second favourite
brand offereda free extra day’s hire. This effect turned out to be the third
most powerful of all the biases we tested, with a transfer of 70% away from
the favourite brand (gure 10).
Figure 10
0
25
50
75
100
42
58
70
30
Stated 1st choice brand Introduction of 2nd choice brand 2nd choice + stronger expression of power of free
100
1st choice brand
2nd choice brand
Power of free executions tested: “free day - 3 days for the price of 2” and “free car clean”. Transfer of
preference from rst choice to second choice brand – power of free analysis, car hire category.
Source: Google / The Behavioural Architects.10,000 simulated car hire purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18–65.
CHAPTER 466
INFLUENCING THE MESSY MIDDLE
While handing out freebies and upgrades worked well with expensive
transactions, the power of free also proved itself with lower-cost, everyday
purchases. A buy-one-get-one-free (BOGOF) offer was the second most
effective expression of a bias in transferring brand preference in the detergent
category, while free popcorn at the cinema also achieved a second-place result.
In the short-haul flights category, we see an interesting example of how
biases sometimes combine, with a free checked bags offer both expressing
the power of free and an important category heuristic (gure 11). This
is an issue to which our simulation participants obviously brought a lot
of baggage, as the offer proved the most powerful expression of any
behavioural bias in the category.
Figure 11
0
25
50
75
100
36
64
63
37
Stated 1st choice brand Introduction of 2nd choice brand 2nd choice + stronger expression of power of free
100
1st choice brand
2nd choice brand
Power of free executions tested: “free checked luggage” and “free hot drink”. Transfer of preference from rst
choice to second choice brand – “power of free” analysis, short-haul ight category.
Source: Google / The Behavioural Architects. 10,000 simulated short-haul flight purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18–65.
CHAPTER 467
INFLUENCING THE MESSY MIDDLE
The power of now
The immediate gratication of rapid delivery wasn’t a huge difference-maker
in our simulation, but it still had a meaningful effect on a handful of categories.
In fast-moving consumer goods (FMCG), products like detergent, moisturiser,
cereal, and cat food all saw consumers responding positively to offers of next-
day delivery (gure 12).
Figure 12
0
25
50
75
100
28
72
42
59
Stated 1st choice brand Introduction of 2nd choice brand 2nd choice + stronger power of now expression
100
1st choice brand
2nd choice brand
Power of now executions tested: “24 hour delivery” and “7 day delivery”. Transfer of preference from rst choice
to second choice brand – “power of now” analysis, cat food category.
Same-day delivery also had an appreciable effect in the clothing and
children’s toy categories, where the convenience of this option serves to
de-risk a highly individual purchase. It may also be more effective when
deployed during evaluation, when it could help to differentiate between
competing propositions. However, whether the additional costs of free
delivery would be justied from a business perspective can’t be discerned
from behavioural data alone.
Source: Google / The Behavioural Architects. 10,000 simulated cat food purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18–65.
Numbers may not add to 100 due to rounding.
CHAPTER 468
INFLUENCING THE MESSY MIDDLE
Supercharging the second‑choice brand
Having explored a variety of behavioural biases across a range of categories,
we next wanted to see how much more brand preference could be won
if second favourite brands were “supercharged” with strong expressions
across all six biases.
The shampoo category is an interesting case in point. First choice shampoo
brands proved surprisingly resistant when the second choice brand was
introduced, losing only 25% – less than were prepared to switch in high-cost
categories like cars and mortgages (gure 13).
Figure 13
0
25
50
75
100
25
75
Stated 1st choice brand Introduction of 2nd choice brand
100
1st choice brand
2nd choice brand
Transfer of preference from rst choice to second choice brand after introduction of second choice brand,
shampoo category.
Source: Google / The Behavioural Architects. 10,000 simulated shampoo purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18-65.
CHAPTER 469
INFLUENCING THE MESSY MIDDLE
We can speculate that the reason for this resilience might be that shampoo
is a product where, once a trusted brand has been identied, people tend
not to switch. So, if we take that hypothesis as a starting point, how much
preference share can we take away from the favourite brand if we use all the
biases at our disposal?
The result is impressive (or alarming, depending on your point of view) with
the second choice brand able to take a full 90% of preference away from the
rst choice brand when supercharged with all six biases (gure 14).
Figure 14
0
25
50
75
100
25
75
90
10
Stated 1st choice brand Introduction of 2nd choice brand 2nd choice brand “supercharged”
100
1st choice brand
2nd choice brand
Transfer of preference from rst choice to second choice brand – bias supercharging analysis, shampoo category.
Source: Google / The Behavioural Architects. 10,000 simulated shampoo purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18-65.
The second choice brand [was] able
to take a full 90% of preference away
from the rst choice brand when
supercharged with all six biases.
CHAPTER 470
INFLUENCING THE MESSY MIDDLE
We see a similar result in the detergent category, where the power of habit,
familiarity, and huge FMCG marketing budgets make initial brand preferences
impressively sticky in the presence of a challenger (gure 15).
In fact, of all the product categories we examined, only smartphone and
children’s toy preferences proved more resilient than detergent.
Figure 15
0
25
50
75
100
24
76
Stated 1st choice brand Introduction of 2nd choice brand
100
1st choice brand
2nd choice brand
Transfer of preference from rst choice to second choice brand after introduction of second choice brand,
detergent category.
Source: Google / The Behavioural Architects. 10,000 simulated detergent purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18-65.
CHAPTER 471
INFLUENCING THE MESSY MIDDLE
However, when we supercharged the second favourite detergent brand with
a range of powerful expressions aimed at our cognitive biases, such as a
BOGOF offer, five-star reviews, and an endorsement from Which? (aUK
brand that provides impartial testing, reviews, and advice), the impact was
profound. Boosted with everything we could throw at it, the second choice
won 78% of shopper preferences, in a category where the first choice
brandshad proven relatively resilient to the mere introduction of the second
choice (figure 16).
Figure 16
0
25
50
75
100
24
76
78
22
Stated 1st choice brand Introduction of 2nd choice brand 2nd choice brand “supercharged”
100
1st choice brand
2nd choice brand
Transfer of preference from rst choice to second choice brand – bias supercharging analysis, detergent category.
Source: Google / The Behavioural Architects. 10,000 simulated detergent purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18-65.
Shampoo and detergent are both fairly low-cost, low-complexity purchases
that we make several times a year at a minimum. But what about a big-ticket
purchase we only make once a year?
CHAPTER 472
INFLUENCING THE MESSY MIDDLE
Package holiday rst choices proved more susceptible to shoppers switching
preference than either shampoo or detergent, with 34% immediately willing to
switch to their second favourite brand when given a choice, holding the other
biases constant (gure 17).
Figure 17
0
25
50
75
100
34
66
Stated 1st choice brand Introduction of 2nd choice brand
100
1st choice brand
2nd choice brand
Transfer of preference from rst choice to second choice brand after introduction of second choice brand,
package holiday category.
Source: Google / The Behavioural Architects. 10,000 simulated package holiday purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18-65.
CHAPTER 473
INFLUENCING THE MESSY MIDDLE
With a much higher impact from the introduction of the second choice brand,
it’s perhaps unsurprising that when supercharged with powerful expressions
of all six biases, the favourite package holiday brand found itself unable to
hold on to much of its preference share. In total, the supercharged second
favourite brand managed to draw away 88% of shoppers, attracted by limited
availability, positive reviews, and similarly boosted expressions across the
board (gure 18).
Figure 18
0
25
50
75
100
34
66
88
12
Stated 1st choice brand Introduction of 2nd choice brand 2nd choice brand “supercharged”
100
1st choice brand
2nd choice brand
Transfer of preference from rst choice to second choice brand – bias supercharging analysis, package
holiday category.
Source: Google / The Behavioural Architects. 10,000 simulated package holiday purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18-65.
Across our 31 categories, when second favourite brands were supercharged
with all six cognitive biases, the result was a profound shift away from the
favourite. Even the stickiest category, mobile network provider, retained less
than a third of rst choice preference.
CHAPTER 474
INFLUENCING THE MESSY MIDDLE
As we saw in the earlier cross-category summary, financial products
such as car insurance, ISAs, and credit cards prove to be among the most
susceptible to a transfer of preference away from favoured brands, while
FMCG products such as moisturiser and breakfast cereal were among the
most resilient (figure 19).
Figure 19
0
25
50
75
100
72
73
74
75
75
76
76
77
77
77
78
79
83
83
83
85
85
87
87
87
87
88
88
89
89
89
89
90
90
91
94
28
27
26
25
25
24
24
23
23
23
22
21
17
17
17
15
15
13
13
13
13
12
12
11
11
11
11
10
10
9
6
Mobile network provider
Cereal
Car
Face moisturiser
Cinema
Whisky
Mortgages
Cat food
Smartphone
Clothing
Detergent
Flights - long haul
Fitted kitchen
Power drill
Car hire
TV
Flights - short haul
Children’s toy
Broadband provider
Laptop
Sofa
Package holiday
Mountain bike
Credit card
Energy provider
Hotel
Make-up
Shampoo
Bathroom suite
ISA
Car insurance
1st choice brand
2nd choice brand
Transfer of preference from rst choice to second choice – bias supercharging analysis, all categories.
Across our 31 categories, when second‑
favourite brands were supercharged with
all six cognitive biases, the result was a
profound shi away from the favourite.
Source: Google / The Behavioural Architects. 310,000 simulated purchases. n=31,000 category buyers, online shoppers, aged 18-65 (31 categories, 1,000
respondents in each). Numbers may not add to 100 due to rounding.
CHAPTER 475
INFLUENCING THE MESSY MIDDLE
As a testament to the power of our six behavioural biases, this was
impressive enough. But there was still one more scenario we wanted to try.
Starting from nothing
Finally, to explore the most extreme implications of our ndings, we
introduced a complete wildcard. We decided to create a ctional test brand
to assess how much preference share an unknown challenger might take if it
was able to hit all of the biases we’d identied.
We decided to create a
ctional test brand to assess
how much preference share
an unknown challenger might
take if it was able to hit all of
the biases we’d identied.
CHAPTER 476
INFLUENCING THE MESSY MIDDLE
Source: Google / The Behavioural Architects. 10,000 simulated mobile network provider purchase scenarios. n=1,000 category buyers, UK
online shoppers, aged 18-65.
And even with everything we’d learned so far about the power of these
behavioural principles, the results came as a surprise. For example, in the
mobile network category, our ctional brand, Gem Mobile, was able to take
almost 50% of preference from the favourite brand (gure 20).
Figure 20
0
25
50
75
100
28 72 50
72
Stated 1st choice brand Introduction of 2nd
choice brand
2nd choice brand
“supercharged”
Fictional brand
“supercharged”
100
28
50
1st choice brand
2nd choice brand
Fictional brand
Transfer of preference from rst choice to ctional brand – bias supercharging analysis, mobile network category.
Entering new markets is a challenge. Even if we skip over the operational
barriers to entry, in many of the categories we simulated, incumbent
marketing budgets and brand associations are considerable, presenting yet
another hurdle for challengers. Agile, intelligent use of behavioural science
might give newcomers a vital advantage.
CHAPTER 477
INFLUENCING THE MESSY MIDDLE
Following on from the example of Gem Mobile, we also created Intergo, a new
broadband provider to test against the established competition. Similarly we
threw every advantage behind this newcomer, and the effect turned out to be
even more eye-catching. In this case, Intergo was able to claim 73% of brand
preference away from the original favourite (gure 21).
Figure 21
0
25
50
75
100
30 87 73
70
Stated 1st choice brand Introduction of 2nd
choice brand
2nd choice brand
“supercharged”
Fictional brand
“supercharged”
100
13
27
1st choice brand
2nd choice brand
Fictional brand
Transfer of preference from rst choice to ctional brand – bias supercharging analysis, broadband category.
Before we get too carried away, it’s worth noting that to achieve these
signicant shares of preference, the two challenger brands needed far
superior propositions. And indeed, some aspects of those enhanced
propositions are probably out of reach even for a well-funded challenger.
This is particularly true of the volume of positive reviews necessary to
constitute persuasive social proof, which must be earned over time as
consumers experience a product or service.
Source: Google / The Behavioural Architects. 10,000 simulated broadband provider purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18-65.
CHAPTER 478
INFLUENCING THE MESSY MIDDLE
And it’s also worth noting that established brands still exert quite a pull. Even
with a vastly superior proposition, half of mobile network service shoppers
still rejected Gem Mobile and opted for an inferior, less appealing proposition,
because it came from their favourite brand.
The comparison between these two verticals might indicate that broadband
provision in the UK, operating as it does on largely the same network
infrastructure, is more commoditised than mobile network provision. But
even with that being the case, over a quarter of shoppers rejected the
challenger and chose to stick with their tried-and-tested favourites.
And in both simulations, the second choice brands outperformed their
ctitious counterparts by a substantial margin when both were supercharged
to the same degree against the rst choice.
But we weren’t just limited to comparing these two product types.
Wecreated and tested ctional brands in each of our 31 categories. Allof
the brands we invented loosely followed the conventions of their category,
with logos and typefaces derived from their real-world counterparts. And
yet, despite their surface plausibility, the fact remains that none of our
participants had any awareness or investment in any of these brands before
the moment they rst encountered them. In terms of our marketing model,
their “exposure” level was zero.
CHAPTER 479
INFLUENCING THE MESSY MIDDLE
Fictional brands
Despite their suace plausibility,
the fact remains that none of our
paicipants had any awareness or
investment in any of these brands.
CHAPTER 480
INFLUENCING THE MESSY MIDDLE
Interestingly, the only category where shoppers showed relative hesitancy
over switching was with our ctional breakfast cereal brand, Honey C’s. Just
over a quarter were willing to switch from their favouredbrand, even when
the ctitious proposition was fully supercharged(gure 22).
Figure 22
0
25
50
75
100
33 73 28
67
Stated 1st choice brand Introduction of 2nd
choice brand
2nd choice brand
“supercharged”
Fictional brand
“supercharged”
100
27
72
1st choice brand
2nd choice brand
Fictional brand
Transfer of preference from rst choice to ctional brand – bias supercharging analysis, cereal category.
Source: Google / The Behavioural Architects. 10,000 simulated cereal purchase scenarios. n=1,000 category buyers, UK online shoppers, aged 18-65.
CHAPTER 481
INFLUENCING THE MESSY MIDDLE
Of course, as much as we’ve tried to keep the design of our ctional brands faithful
to the conventions of their sector, it is possible that these shoppers detected a
subconscious hint that signalled our deception. Or alternatively, breakfast cereals,
particularly sweet varieties, may just enjoy strong brand loyalty (gure 23).
Figure 23
0
25
50
75
100
87
86
82
79
79
78
76
75
75
73
73
71
65
64
63
63
63
63
62
60
60
53
51
50
50
50
49
48
44
43
28
72
57
56
52
51
51
50
50
49
47
40
40
38
37
37
37
37
36
35
30
27
27
25
25
25
22
21
21
18
14
13
Cereal
Flights - long haul
Car (SUV)
Whisky
Detergent
Smartphone
Face moisturiser
Mobile network provider
Cat food
Clothing
Children’s toy
Laptop
Cinema
Shampoo
TV
Fitted kitchen
Mortgages
Power drill
Flights - short haul
Make-up
Energy provider
Broadband provider
Package holiday
Credit card
Car hire
Hotel
Sofa
ISA
Bathroom suite
Mountain bike
Car insurance
1st choice brand
Fictional brand
Transfer of preference from rst choice to ctional brand – bias supercharging analysis, all categories.
Looking at the ctional brand scenario across all categories, once again
the yellow portion of the stack represents the share of preference for
the shoppers’ rst choice brands and the orange portion is the share of
preference for the supercharged ctional challenger.
The product reordering throws up some intriguing cross-vertical patterns:
FMCG predominantly featuring on the left of the chart, nancial services,
travel, and utilities towards the right, retail scattered throughout. However,
we should reiterate that these patterns are only suggestive, and certainly
shouldn’t be used to quantify market-entry opportunity, as there remain
uncontrollable aspects of each brand and product relative to their category
that will also have influenced the degree of preference shift.
Source: Google / The Behavioural Architects. 310,000 simulated purchases. n=31,000 category buyers, online shoppers,
aged 18-65 (31 categories, 1,000 respondents in each). Numbers may not add to 100 due to rounding.
CHAPTER 482
INFLUENCING THE MESSY MIDDLE
Simulation summary
Our simulations offer a framework for decoding how decisions are made in
the messy middle. Over the course of some 310,000 simulated decisions,
we’ve seen how the behavioural biases identied in our literature review can
have a powerful effect on shopper preferences.
Before we draw conclusions, we have to bear in mind that not all of our biases
are as effective across every category. And it’s worth repeating that none
of our executions used anything other than basic copy and design, so these
results don’t speak to the power of creative to harness and enhance cognitive
biases.
But with those caveats in mind, three broad conclusions can be drawn:
Even a brand you’ve never heard of can disrupt
preferences in the messy middle
There’s no doubt that the results of our ctional brand tests will be surprising
to many readers. Some may even nd themselves sceptical of the endeavour.
However, the results of the experiment are consistent with the premise that
behavioural biases have powerful effects on purchase decisions. In the world
of the simulation, these brands existed, supercharged with the best possible
expression of our behavioural biases. Shoppers made a choice, and while
established brands still exerted a powerful pull, the biases had the effect that
behavioural science theory said they would.
1.
Marketing history is liered with
stories of sta‑up challenger brands
who came out of nowhere to seize
substantial market share.
CHAPTER 483
INFLUENCING THE MESSY MIDDLE
After all, marketing history is littered with stories of start-up challenger
brands that came out of nowhere to seize substantial market share. Many of
those brands will have made extensive use of behavioural science to boost
the impact of their market entry. If you want a recent example, you only need
to look at the growth of direct-to-consumer mattress brands, which all make
use of powerful cues like free delivery, free returns, extensive user reviews,
and expert endorsements.
Our simulations have revealed some biases so powerful that every brand
should be aware of their influence, if only to be able to defend against
competitors leveraging biases such as social proof and the power of free. But
for the most part, brands would not want to approach this area piecemeal.
Each of the biases we explored addresses a cognitive need, and as our
supercharging results show, brands that know how to help consumers
navigate and simplify decision-making are often richly rewarded.
Many shoppers remained loyal to their
favourite brand even when the alternative
oered a vastly superior proposition.
2.
The overdog effect – brands (still) matter
Despite our best efforts to swing things in favour of the ctional brands, in
every category, many shoppers remained loyal to their favourite brand even
when the alternative offered a vastly superior proposition. In several cases,
more than half of all category participants were uninterested in shifting
away from their favourite, and in the majority of categories more than a third
ignored the challenger and stuck with their rst choice.
Everyone loves an underdog story – unfavoured brands shifting preferences
just by showing up and out-marketing their rival with some clever tactics. It’s
certainly an appealing tale, but as our data shows it isn’t the whole story.
CHAPTER 484
INFLUENCING THE MESSY MIDDLE
Presence can be all it takes to shift preferences in the
messy middle
There’s power in just showing up at the right moment. And this effect is
visible across every category we tested. This is such a fundamental truth that
we’re making it the last of our three insights.
Even in a complex world, sometimes all it takes to make a big impact is to
show up at the right time.
20
Even in a complex world, sometimes
all it takes to make a big impact is to
show up at the right time.
3.
Getting comfortable in the messy middle
The messy middle isn’t always an easy place for marketers to navigate.
But as our experiments show, with a few powerful behavioural cues to act
as signposts, brands can show up at the right moment and win consumer
preference, whatever their category.
In the next chapter, we’re going to be building on these results by looking
at the wider implications of our experiments for both established and
challenger brands marketing in the messy middle.
20 The “mere exposure” effect also suggests that continued presence should have a long-term impact on consumer afnity and preference,
as repeat exposure to something engenders an increase in positive feelings about it.
Implications for brands
With both theory and experimental evidence in hand, in this chapter we’re
going to start looking at what our research ndings mean in practice for
marketers. We’ll explore how both established and challenger brands can
adapt to the new consumer reality, and identify the key implications for both
types of business.
For established brands
As our shopping experiments show, even established brands can nd
themselves vulnerable within the swirl of the messy middle.
Heavyweight brands can’t afford to be complacent: understanding the
behaviour and mindset of consumers is now a vital part of protecting market share.
Established brands represent a signicant historical and ongoing investment.
This research suggests that these businesses may not be getting the optimal
return on that investment if they aren’t conscious of the disruptive potential of
the messy middle. Just being present during initial consideration isn’t enough.
With shoppers happy to loop through multiple phases of exploration and
evaluation, even the biggest brands need to ensure that they are present and
meeting consumer expectations throughout the decision-making process.
5
Implications of
the messy middle
CHAPTER 586
IMPLICATIONS OF THE MESSY MIDDLE
For challenger brands
For less-established brands, our shopping experiments demonstrate that the
messy middle offers rich prospecting for nimble and resourceful marketers.
Challengers should see the messy middle as a window of opportunity:
consumers are willing to explore and evaluate alternatives, and even entirely
new brands have the chance to change mindsets, disrupt established
preferences, and win new customers.
Our research reveals that far from being an insurmountable obstacle to
market entry for newcomers, consumer brand preferences can be fragile
across many categories. Our insights into the impact of biases such as social
proof, the power of now, and the importance of visibility at key moments of
consideration can help level the playing eld against even established brands.
The good news is that for both well-established and challenger brands, the
right approach to marketing in the messy middle is identical. We’ve identied
three key actions, which we’ll explore in detail over the rest of this chapter:
Ensuring brand presence, so that your product or service is
strategically front of mind while your customers explore.
Intelligently (and responsibly) employing behavioural science
principles, so that your assets and messages become more
compelling as customers evaluate their options.
Closing the gap between trigger and purchase, so that your
existing and potential customers spend less time exposed to
competitor brands.
1.
2.
3.
Challengers should see the messy
middle as a window of oppounity.
CHAPTER 587
IMPLICATIONS OF THE MESSY MIDDLE
Ensuring brand presence
Put simply, none of the other tactics explored in this report are possible if you
don’t rst show up and make a claim for the consumer’s attention. Being
present from the rst moment of deliberation is table stakes for any brand
hoping to emerge triumphant from the messy middle.
As we’ve seen, simply being presented with a choice can lead to signicant
changes in consumer preference. Consumers instinctively favour those brands
that enable exploration and help them to make sense of the messy middle,
especially when they rst enter the space. Ensuring brand presence creates (or
retains, in the case of repeat customers) mental availability for your products
and services, which would otherwise be ceded to competitor brands.
To cut through in the messy middle and make swift, effective connections
with customers in “explore” mode, you should:
Use available data to qualify and categorise shoppers who
are exploring – data-driven algorithms should eventually
make this identication possible at scale.
Provide a great user experience that makes exploring your
offerings as easy as possible.
Present all the relevant information potential customers
need to make a rapid transition into evaluation and then on
towards purchase.
Brands are long-term strategic assets, expensive to build and maintain.
This research is not intended to dene a comprehensive brand strategy,
nor to give insight into how the exposure phase contributes to the enduring
associations and attachments that branding activity seeks to foster.
1.
CHAPTER 588
IMPLICATIONS OF THE MESSY MIDDLE
However, as we’ve seen in our shopping experiments, simple behavioural
biases can powerfully undermine even strong brand preferences.
So, whether you’re seeking to maintain the preferred status of an established
brand, or looking to introduce a new contender to the market, you need to
show up and deploy the behavioural biases most relevant for your category.
Although we believe that a comprehensive search strategy is essential,
showing up isn’t just a question of keywords and ads. Depending on your
category, price comparison engines, social media platforms, video, news,
and niche content such as gaming or technology sites may be equally
important when maintaining parity of brand presence. Comprehensiveness
is key – any gaps in your media plan could see you locked out of the loop as
consumers begin exploring their options.
Any gaps in your media plan could see
you locked out of the loop as consumers
begin exploring their options.
Intelligently (and responsibly)
employing behavioural science
In his 2019 book “Alchemy”, Rory Sutherland references a theory attributed
to former Ogilvy & Mather ad executive Joel Raphaelson. The theory states
that: “people do not choose Brand A over Brand B because they think Brand
A is better, but because they are more certain that it is good.” This is a subtle
distinction, but one we think is borne out by the results of our research.
In particular, consumers are looking for reassurance to buttress their
purchase decisions during the evaluation stage of our model and as they
move on to purchase.
2.
CHAPTER 589
IMPLICATIONS OF THE MESSY MIDDLE
Brands themselves provide this reassurance – in our shopping simulations,
even when ctional or non-preferred brands were supercharged to address
all six biases, the preferred brand still invariably retained some loyalty. This is
just one example of how a better understanding of the cognitive biases that
underpin decision-making can help to create a compelling proposition that
appeals to shoppers at an instinctive level.
A beer understanding of the
cognitive biases that underpin
decision-making can help to create a
compelling proposition that appeals to
shoppers at an instinctive level.
Employing behavioural science intelligently
Although pre-existing brand afnity and price are undoubted drivers of
purchase decisions, we have seen that purchase outcomes can also
be strongly influenced by the messages, propositions, and tactics that
competing brands bring into play. Behavioural science principles can be
applied at several points within the messy middle:
Use available data to qualify and categorise shoppers who
are evaluating – data-driven algorithms should eventually
make this identication possible at scale.
Ensure that your ad messaging is tailored to the needs of
evaluative shoppers, containing behavioural biases relevant
to your category.
CHAPTER 590
IMPLICATIONS OF THE MESSY MIDDLE
When shoppers visit your site, the user experience should
make the evaluation process as simple as possible, with
appropriate detail and functionality.
Use tactics such as retargeting and basket-abandonment
messaging to engage with evaluative shoppers who are in
danger of exiting back into explore mode.
Shoppers don’t engage with brands in a vacuum once they enter the messy
middle – the process of exploration and evaluation is inherently comparative.
With that in mind, it’s a good idea to regularly review how your offering and
messaging compare with that of the competition.
While many brands will audit their competitors for price and product feature
parity, the messy middle suggests that businesses now need to be aware of
the behavioural science being employed by their rivals.
To take the example of social proof – a bias that had signicant impact on
choices across all the products we researched – how do your consumer
ratings and reviews match up to those of your competitors? Are you utilising
positive user feedback about your brand and products in your marketing
activity? Likewise, are you building your brand authority by seeking out and
promoting expert endorsements and industry awards?
Employing behavioural science responsibly
Economist Richard Thaler has written extensively about “nudges” – small
cues that direct people towards positive behavioural change but aren’t
bribes, and don’t prevent them from making an alternative choice if they want
to. More recently, Thaler has introduced the notion of “sludge” – behavioural
cues that, unlike nudges, don’t have the customer or end-user’s best
interests at heart.
21
As the name suggests, sludge serves only to obscure and distort the
decision-making process, making the middle even messier. It’s a foundational
principle of everything we do at Google that if you put the user rst, all else will
follow. So it’s safe to say that we’re not big fans of sludge, and don’t want the
ndings of our research to be misunderstood or misapplied.
21 Thaler, R. H. (2018). Nudge, not sludge. Science Vol. 361, Issue 6401, pp. 431.
CHAPTER 591
IMPLICATIONS OF THE MESSY MIDDLE
Fortunately, a growing body of guidelines around the use of behavioural
science is already in the process of being established. At the category level,
there are codes of practice for marketing in nancial services, health, and
other regulated markets that set out how these kinds of tactics can be used
responsibly and sustainably. At the platform level, advertising services such
as Google Ads and its counterparts all have terms of use that govern the
kinds of claims and tactics that advertisers can implement. Ideally, each
brand’s own marketing policies should also contain guidance as to how its
messaging can make responsible use of behavioural science.
Finally, it’s worth remembering that the potential cost of doling out sludge
isn’t just the burden of additional regulatory oversight. At the heart of this
report is the realisation that consumer behaviour is constantly evolving,
and that over the past two decades it has started to change faster than
ever. Pressure tactics, such as scarcity bias and the power of now, can even
edge over from nudge to sludge if applied at the wrong moment or used
too regularly. Consumers soon grow wise to the tricks that unscrupulous
businesses play on them, and the cost to a brand of having its marketing
tactics recognised as sludge could be huge. Once lost, credibility and trust
are very hard to regain.
Closing the gap between trigger and purchase
The ultimate aim of this approach is to reduce the cognitive burden
experienced by consumers as they explore and evaluate your proposition. It is
particularly relevant for existing customers, who expect that their familiarity
with your products and services should be reflected in a simple, pain-free
purchase process. In short, once the shopping trigger has been pulled, the
goal is to marshal all your design, usability, and user experience resources to
ensure that your ad copy and website don’t shoot you in the foot.
After all, not every customer needs to explore and evaluate new brands.
If someone has bought from you before and they were satised with the
experience, they are likely to turn to you again to answer the same need. If
you don’t place any unexpected impediments or barriers in their way, there is
a good chance they’ll make a repeat purchase.
3.
CHAPTER 592
IMPLICATIONS OF THE MESSY MIDDLE
So, what might these barriers look like in practice?
Poor site speed, particularly on mobile.
Inconsistent or unclear messaging, particularly between
adcopy and landing page.
Inadequate information, such as missing product details.
User experience issues, such as unclear navigation,
pop-ups, and limited payment options.
The cost of getting these basic user experience considerations wrong can be
considerable. In a study looking at the importance of mobile speed, we saw
that while 95% of users said they would return to a site they perceived as being
fast, only 62% said they would revisit a site they perceived as slow.
22
In another
study, we saw that a 0.1 second improvement in mobile site speed increased
conversion rates by 8.4% for retail sites, and 10.1% for travel sites.
23
Lowering
the drag that factors such as speed and design have on interactions with your
brand increases cognitive ease, making shoppers less likely to be motivated to
dive back into another cycle of exploration and evaluation.
The importance of measurement
While the evidence in our research was based on a simulated
environment, the only way to understand the difference that these
changes can make to your business is to test them in the wild. And
given the difculty of attributing subtle causal effects to relatively blunt
metrics like sales and revenue, we believe that constructing robust,
controlled experiments is necessary to understand the impact of behavioural
biases on your bottom line.
22 Think With Google (2017). The need for speed: Evaluating the perception and reality of speed on the mobile web. Google. https://www.
thinkwithgoogle.com/intl/en-gb/advertising-channels/mobile/need-speed-evaluating-perception-and-reality-speed-mobile-web,
23 Think With Google (2020). How speeding up your mobile site can improve your bottom line.
Google. https://www.thinkwithgoogle.com/
marketing-resources/experience-design/mobile-page-speed-data/
CHAPTER 593
IMPLICATIONS OF THE MESSY MIDDLE
Measuring advertising effectiveness is a large topic, beyond the scope of
this report. Last year our research team published a report that looked at the
state of the art and the opportunities for improvement.
24
The rst section examines how controlled experiments and causal inference
can be used to increase the accuracy of measured improvements in
marketing performance, and contains a number of useful recommendations.
Organisational implications
Over the course of the past year, we’ve taken our ndings out on the
road, presenting them at large events and to individual marketing teams.
Something we’ve heard over and over is that many marketers feel our “messy
middle” metaphor actually serves as an apt description of the way marketing
departments have evolved over time.
Many marketers feel our ‘messy middle’
metaphor actually serves as an apt
description of the way marketing
depaments have evolved over time.
24 Taylor, M. et al. (2019). Measuring Effectiveness: Three Grand Challenges. Google. https://www.thinkwithgoogle.com/intl/en-gb/consumer-
insights/measuring-effectiveness-three-grand-challenges
Before the web it was easier for marketing to own the entire customer
experience. But over the past 20 years or so, the sudden rush of information
and complexity has led to organisational fragmentation, with different
departments owning web, mobile, data operations, and user experience. This
was understandable, as launching onto internet street required skills that
typical marketing departments of the time didn’t possess.
CHAPTER 594
IMPLICATIONS OF THE MESSY MIDDLE
Fast forward to the present day, and marketing has become a lot more
technical. Those fragmented responsibilities are starting to be reintegrated
into a singular function with ownership of the full customer experience.
Our research suggests that this is the right direction of travel, with many
of the executions considered by our research requiring cross-functional
collaboration to implement.
There’s also the question of the traditional separation between branding and
performance, developed in the days when television and direct mail were
paramount. This too, has mostly been ported directly into the digital age.
And while there’s still plenty of tactical value in both of these approaches, the
exploration and evaluation that takes place in the messy middle straddles
many traditional divides. It’s now clear that a signicant amount of potential
could be going untapped, falling into the gaps between silos.
Marketing departments have changed considerably over the past couple
of decades. Those that keep the needs of the messy middle in mind as they
grow and evolve should nd themselves with everything they need to keep
pace with whatever consumers do next.
Thriving in the messy middle
The messy middle changes things for marketers but, as we’ve seen,
while consumer behaviour is becoming more complex, many of the
approaches needed to address it are still reassuringly familiar. With a
better understanding of consumer thinking and a clear set of actions, both
challenger and established brands will have the tools they need to win and
protect their share of consumer preference.
6
Before we part ways, why don’t we take a last walk along internet street? Over
the course of this report we’ve seen how the growth of the web has brought
with it abundant choice and limitless information, transforming consumer
behaviour in the process.
While examining this transformation, we’ve identied a new model for
how people make decisions online. In our model, the sum total of a
shopper’s experiences and impressions creates a backdrop of exposure,
encompassing brands, products, and more. Against this backdrop, purchase
triggers prompt consumers to enter a cycle of exploration and evaluation,
gathering information and then narrowing it down. If the rst cycle doesn’t
yield a denite choice, they loop back, repeating as many times as necessary.
Finally, all options evaluated, they make a purchase. Or they don’t. Either way,
the whole experience feeds back into their background exposure.
The messy middle is a complex space for
marketers, where customers are won
orlost but, from the consumer perspective,
people are doing what they’ve always done.
Inhabiting the
messy middle
CHAPTER 696
INHABITING THE MESSY MIDDLE
Sounds complicated, right? And yet, here on internet street it doesn’t really
feel that way. The messy middle is a complex space for marketers, where
customers are won or lost but, from the consumer perspective, people are
doing what they’ve always done – perceiving a need and trying to answer
it with a purchase. The fundamental mechanics of shopping may have
changed beyond recognition on the web, but we’ve adapted. Mental modes
and behavioural biases that served our early ancestors turn out to be just as
useful for cutting through the complexity of shopping on the internet.
For marketers the story is a little different. Branding and performance,
traditionally divided in many marketing organisations, actually overlap in the
messy middle, but that doesn’t mean potential customers aren’t falling into
a gap. Fortunately, the messy middle itself can be a template for brands to
build empowered and integrated marketing organisations, flexible enough to
adapt to consumer behaviour now and in the future.
And those marketing teams that set out to tackle the messy middle will hopefully
nd in our research a valuable set of hints for where to direct their energies:
Show up at key moments of exploration and evaluation to
win or protect your share of consumer preference.
Apply behavioural biases to give shoppers the information
and reassurance they need to exit the messy middle and
complete a purchase.
Optimise site speed, user experience, and onsite messaging
to shorten the distance between trigger and purchase.
CHAPTER 697
INHABITING THE MESSY MIDDLE
We’ve also sounded a gentle note of caution. Behavioural science is a
powerful tool, and marketers not using it responsibly could nd themselves
doing long-lasting damage to the brands they represent. Humans tend to be
quite good at remembering grievances, after all.
But what if everyone who reads this report takes our research and applies
each of the cognitive biases we’ve identied? Won’t that just create an
elevated-but-level playing eld, leaving the middle just as messy?
From the consumer perspective, the answer is a denite no. The better
brands get at anticipating shoppers’ needs for information and guidance,
the better customer experience will become overall. Exploration will be more
efcient and evaluation will be simpler – the shopping journey will shorten
and result in better outcomes and experiences.
Fortunately, for marketers the answer is also likely to be no. In our simulation,
each bias was given a basic execution, with no attention lavished on design
or copy. In the real world, brands will hopefully take our indicative examples
and test them in-market, bringing their own ingenuity and insight to bear.
Ultimately, our research provides not just a framework for decoding
decisions and navigating the messy middle, but also a springboard for
creativity. Brands that are able to integrate the lessons of behavioural science
into their marketing toolset will have everything they need to flourish.
The beer brands get at anticipating
shoppers’ needs for information
and guidance, the beer customer
experience will become overall.
This work was influenced by many, and we thank
them for their advice, support, and contributions.
From Google:
Rohan Gifford, Debadeep Bandyopadhyay, Megan
Bowden, LucySinclair, Janusz Moneta, Asia Perek,
Celena Baggio, Nikki Borkovic, WillWhalley, Charlotte
Morton, Waqar Ali, Laura Smith-Roberts, Zarina de Ruiter,
Natalie Zmuda, Brianne Janacek Reeber, Casey Fictum,
Amy Shemesh, GeorgieAltman, Josh Ayto, Justine
L’Estrange, Liz Cracknell, Nisha Mathews, Paul Guerrieria
From outside Google:
Alya Hazell, Sian Davies, Sarah Davies, Jake Greaves,
Kate Pilling, Eimear MacGarty, Oliver Carey, Eleanor
Heather, Josa Taylor, Lisa Rabnor, SamPearce, and
Charlotte Brown from TheBehavioural Architects
Ben Wordsworth and MartinaBrillante from
TogetherCreative.
The team at Redwood BBDO.
Rory Sutherland, Pete Dyson and Mike Hughes at Ogilvy.
Neil Godber, Philip Walford, Reuben Milne, Richard
Shotton, Simon Gill and Vicki Holgate, for their
contributions as the project developed.
About The Behavioural Architects:
The Behavioural Architects are an award-winning global
insight, research, and strategic consultancy. Their work is
underpinned by the latest thinking from the behavioural
sciences which they leverage to help organisations better
understand and influence consumer behaviour. They apply
behavioural inspired frameworks and methods to help
address key strategic marketing and social challenges.
Acknowledgements
