What Advanced Analytics can learn from Behavioural Economics

Much has been made of the importance of measuring customers’ ‘System 1’ thinking – the fast, intuitive thought process via which most decisions are made. On paper, advanced analytical techniques such as conjoint analysis and max diff have much to offer – their outputs producing ‘revealed’ rather than ‘stated’ factors of importance. However, in practice these methods are guilty of many of the crimes of traditional questioning – asking too much of customers and resulting in either over-rationalised or inaccurate responses.

We’ve teamed up with our statistical partners, Bonamy Finch, to discover what such techniques can learn from the world of Behavioural Economics and, ultimately, get closer to the phenomena that really shape how people shop.


Keep it visual

When out shopping, customers rarely assess a product by evaluating it across a range of dimensions – so why do we probe for this in surveys? Not only do visuals lead to greater engagement with questions, they also reduce the amount of reading and, hence, the amount of effort each customer needs to exert. The less effort exerted, the more likely their response will be reflective of their System 1.

Keep attribute lists short, and the number of dimensions tight

The greater number of attributes and dimensions, and the more rotations needed, the longer each respondent will be required to answer questions; risking burnout. Whilst it’s always tempting to find out more, we need to balance this against the danger of overtaxing customers and driving them into that unrepresentative, over-rationalised mode. 

Protecting System 1 by giving customers time to reset

There are many possible ways of helping feedback to remain intuitive. For example, barrages of trade-off questions can be onerous, so why not split them into 2 or more chunks? Employing quiz questions or other engaging content can disrupt the customer and give them a chance to reset their brains; reducing fatigue and keeping answers accurate. In the analysis stage, early responses (where the participant is still carefully considering their options and learning how to answer the questions) and late responses (where they’re more likely to have given up and be rushing through) can be excluded – leading to even more telling results.

Use adaptive choice-based conjoint (ACBC)

Traditional conjoint asks customers to make decisions in the light of a range of dimensions and variables – operating on the assumption that all choices are ‘in bounds’. In reality, to help them decide, consumers place constraints on themselves (“I definitely need at least 2GBs of data”, “I only need a meal for 2”) – automatically ruling out many available options. ACBC takes this into account – tailoring the choices respondents are asked to make, and making the barrage more representative of the factors they genuinely consider when they make decisions.

For more information, get in touch with us