Tom Ewing

Deliberately distracting survey participants isn’t generally good research practice. And people investigating shopper behaviour don’t often start by spraying a store with perfume But these are a couple of the techniques we’ve been using at BrainJuicer to overcome one of modern marketing’s greatest – and most stimulating – challenges: making market research reflect what psychologists now know about how people really make decisions.

In the case of the distracted participants the problem started with a client conundrum. This client – an FMCG giant who acted on these findings and so prefers to remain anonymous – was being outsold at the shelf by a competitor. The client blamed the packaging. The competitor’s products came in a pack with a prominent, appealing picture, keeping benefit claims to a minimum. Their own product, on the other hand, had a smaller, less compelling picture and a stack of information.

Racked side by side, it seemed common sense that our client was going to lose out to their sparkier, happier competition. But every time the client ran a simulated shelf test, their brand won – the opposite result to its real world performance. What was going on? Was the packaging really to blame – and if so, why wasn’t research picking up on it?

With the client we worked out a hypothesis. The duller pack won in tests simply because respondents had unlimited time to assess both packs. This meant they took in the information on each pack and weighed it up before choosing – and the most informative pack triumphed.

But this is absolutely not what happens in a supermarket. Depending on the category, studies have shown that decisions at shelf take between 3 and 10 seconds. A considered decision is very rare – a habitual or instinctive one is far more likely. And in those conditions, we theorised, the more directly appealing pack was winning out.

So to simulate those conditions, we tried different ways of putting people’s decisions under pressure, including limiting the time they had to answer, and introducing distractions in the form of a fake tannoy announcement while they ‘browsed’.

Both of these interventions reversed the brand preference in the pack test – the simpler competitor pack moved ahead of our client’s more complex design, providing a far closer match to real-world sales data. We’ve since repeated the test and have found that this effect – pressuring respondents makes choices more realistic – applies across a range of categories.

Researchers have known for a long time that results are context-dependent. If you can simulate the context of a decision, you’ll get better responses. But we’re only starting to take on board decades of psychology explaining exactly why this is – and understanding what we need to do about it.

The Elephant And The Rider
The reason simpler packaging often succeeds at shelf is the same reason 9 out of 10 patients don’t change their lifestyle after a heart bypass, and the same reason you’re more likely to leave your spouse than your bank. People very rarely make decisions – even difficult or complex ones – by carefully weighing information and options.

Psychologists, in fact, believe we have two linked decision-making processes. One intuitive and unconscious, one is conscious and deliberative. And the fast, intuitive one is far more powerful: what we believe to be reasoned thinking is usually post-facto justification for decisions we’ve already unconsciously taken.

Nobel prize winner Daniel Kahneman calls these modes of decision making System 1 (the fast one) and System 2 (the slow one) and the best summary of their implications is in his bestselling book, Thinking Fast And Slow. But your own intuitive System 1 is likely to grasp them better through metaphor. System 1, says Rory Sutherland, is the Oval Office – taking our decisions – and System 2 is the press office. Our favourite metaphor – because it captures the difficulty of behaviour change – is from psychologist Jonathan Haidt: System 1 is an elephant and System 2 its rider. You may identify with the rider, and imagine you’re directing the elephant, but it is far more powerful than you are! An elephant is very difficult to steer – much better to build a path it will naturally follow.

In the same way, marketers wanting to affect decisions led by System 1 need to build a path to their brand for the elephant, not persuade the rider. And this is what our client’s brand failed to do. In the real world, its competitors’ simpler and more emotionally engaging packs made decisions faster and easier. They didn’t try and persuade the rider, they simply made the elephant’s path smoother.

If almost all our decisions are led by unconscious and intuitive impulses – emotions, habits, and heuristics – and the reasons we later give are post-rationalisations, this changes how marketers think. But it changes our jobs as researchers even more – forcing us to rethink how we approach our respondents. Direct questioning is thrown into jeopardy – the research environments we build often seem precisely designed to force people into System 2 thinking, which isn’t the prime mover of their real decisions. As our use of distraction and pressure shows, you can change the research context in useful ways – but can you go further still?

The Smell Of Success?
People very rarely make decisions – even difficult or complex ones – by carefully weighing information and options
One set of techniques often held up as a solution is neuromarketing, which measures our unconscious decisions by measuring body and brain activity directly. This is a very useful step forward, but it isn’t the be-all and end-all of implicit measurement. For one thing, it still puts the focus squarely on the individual and their response to a specific stimulus – a brand, an advert, or a pack – but the decision context goes well beyond that. People make decisions relative to the social and environmental context they are in. – selecting popular or distinctive choices, or being swayed by default options or sensory cues, for instance. To really get at how choices are made, we need to focus on the behaviour, not the behavee.

We believe a new style of research is emerging that takes this into account: not wholly qualitative or quantitative, but hands-on and experimental. It’s this new research that we attempted at BrainJuicer when we worked with Belgian lingerie retailer Hunkemoller to create experimental approaches to improve sales across their stores.

Our initial work on this project fell well outside standard definitions of research. It involved a literature review of relevant consumer psychology to understand what might affect customers in such a store, and working in tandem with Hunkemoller to develop a number of “behavioural interventions”. These were small experimental changes to stores which we hypothesised might drive a sales increase. Introducing scent into the store, for instance, or offering customers chocolates at the door to get them in a good mood.

Hunkemoller nominated a series of their stories for us to run these interventions in, and we drew up an experimental schedule so that for every intervention, in any given week, each store would be paired with a control store where no experiments took place. Our ‘dataset’ would be the real purchase data from stores.

Did it work? Not every intervention succeeded – but this will always be the case with behavioural studies of this kind. When we looked at the data from the scent intervention, though, we found an exciting and positive trend. The average basket value in stores with scent was up 20% on stores without it – the perfume was making the store feel more emotionally appealing and luxurious, we hypothesised, and encouraging customers to buy more upscale items.

Taming The Elephant
This kind of research is a significant departure from the traditional role played by market researchers. Research, in general, rests on a model of decision making dominated by System 2 thinking. In its crudest form, this model assumes people make conscious, individual decisions, and can either bear direct witness to their motivations, or at least provide indirect answers which point us to them.

Research based on behavioural experiment, however, rejects this. Instead it assumes the capacity for individuals to explain their decisions is very limited, and looks toward researching the context for those decisions instead. It focuses on the elephant and its path, not the rider. So how do you do it effectively?

You need a good grounding in decision science and in the specific context of a decision, in order to create hypotheses and design interventions to test them. We find this is best done with a literature review and workshop stage. You need interventions that are practical and testable but which will also give meaningful results – so you need a close partnership with a client and an agile, constructive attitude. And you need to make sure you are using control samples and robust experimental designs.

These conditions are not easy to meet, and this sort of research – rooted in design and hands-on action – is in its infancy. But we believe the approaches outlined in this article – experimenting with real behaviour and with research context – are vital parts of any new research toolkit. Hacking the context of research – with time pressure, distraction, or other tools – can get us closer to System 1 decision making conditions and help us understand how decisions really work. Behavioural design – auditing contexts and building interventions – can go a step further. To paraphrase Karl Marx, the researcher’s job will no longer be to interpret the world. It will be to change it.

Tom Ewing is Digital Culture Officer at Brainjuicer