As a researcher, the idea of using behavioural data instead of declarative data, i.e. survey responses, is extremely enticing. Data, which captures actual consumer behaviours, brings with it the ‘promise’ of greater truth and accuracy. But does it tell the whole story? Can declarative survey data be equally as accurate with the necessary narrative? Let’s examine the pros, cons and best practice use cases for each.
Put simply behavioural data is the record of an action and declarative data is words, in conversation or in this case turned into data points in a survey. Some examples include:
Actions - Purchase records, transactional data, search data, app usage
Declarations - Survey responses, vox pops as well as interviews, focus groups
Any declarative method is limited by the bounds of conscious experience. In a survey, interview, focus group or forum a consumer can discuss what they know about a purchasing decision, but they cannot tell you about parts of the process that they are not aware of, or that they haven’t logged in their memory bank.
We humans have a tendency to post-rationalise and over-rationalise. We need to store information and recall it in a sensible order and we do not have the capacity to store all of the information that we have observed. This means that when we describe a purchasing decision we will draw on what we know, we will talk about price and quality in depth for example. But we won’t necessarily provide the required brand and advertising feedback because we aren’t able to describe how these factors influenced our purchase decision.
For example, the purchase of new life insurance policy is a big investment/commitment and most people are very likely to consciously consider the product options available to them. In turn they are likely to be able to describe, report, discuss and ‘declare’ this experience with some accuracy. A person ‘popping out’ to buy a coffee however is unlikely to give this decision the same amount of thought. They are far more likely to walk into the first coffee shop they see and select their favourite type of coffee! If we invited that coffee buyer to an interview to describe how and why they decided to buy that coffee they probably wouldn’t have very much to ‘declare’.
Referring back to the example above, it is not to say that some people don’t buy life insurance on impulse or that others don’t plan their coffee purchasing very carefully either, but rather that the majority would tend to consider the bigger spend to a greater degree. Therefore when looking at coffee purchasing behaviour, a decision more often than not made in the moment, it’s really important to understand the contextual factors at play in that moment; shop layout, outside signage, the weather, the traffic, etc.
A short declarative survey would cover all of this for you and if you could deploy it right there in the moment it would probably offer you the best, most efficient summary of behaviour. However if you are not in a position to deploy a survey in real time you will be subject to the issues of customer recall, i.e. when you ask your customer to describe the weather, the signage they saw, how the shop looked, it is highly likely they will miss-remember some part of it.
In this instance the customer’s behavioural data, the purchase data, weather data and advertisement logs, may be a better bet for accurate shopper insight. You will be able to learn a lot from the data patterns and make some connections between the three. The behavioural data here is vast compared to the survey data however and will require more advanced software and more time invested in analysis to deliver the results.
When you are considering whether a quick declarative survey will give you the answers you need ask yourself:
1. Does my customer think this through?
2. Can they accurately describe it to me?
3. Can I overcome the rationalisation barriers, i.e. deploy the survey ‘in the moment’ at point of sale
If no, you either need to use techniques to reveal the historic thought processes without biases; facilitate accurate recall, i.e. implicit research methods, or consider some behavioural insight.
Let’s look at another example where the insight required relates to lunchtime food advertisements. If you ask people in a food court why they are there they will ‘declare’, because they are hungry! They won’t say they saw an advert upstairs and that made them feel hungry without some serious prompting and probing, because seeing the ad was a fleeting moment.
In this scenario the answer to questions 1. and 2. is ‘no’ and I would suggest (regardless of the answer to questions 3.) if you have transactional data that will tell you what people bought and when they went to the food court you don’t need to deploy an in the moment survey to be able to correlate ad type with sales uplift.
Key Rule of Thumb: How Much Thinking Does the Customer Do?
A classic problem with declarative data is ‘overclaim’. If you ask someone if they are concerned about an issue they will likely say yes because they think that they ought to be. For example, people SAY that data privacy is important to them and they will exercise their right to be forgotten, but what they often DO is blindly agree yes, yes, yes to sharing their data to get at products they want.
Declarative survey data is not necessarily inaccurate in this case, it still reflects the values that the customer aspires to very well, but it doesn’t accurately reflect their actions within the context of offering product benefits. Behavioural data would of course but it is historic. When developing as opposed to evaluating, projective techniques would be best employed.
How do you know if what you are seeing is the right data to inform your decision? Behavioural data has a lot to offer a less experienced researcher in this regard. If it is readily available providing a complete picture of influencing variables firm conclusions can be drawn without so much need for an appreciation of the nuances of human behaviour, i.e. effective declarative survey design.
At a recent MRS Data Science conference, Unilever said there is still an important role for both behavioural data and declarative survey data but the truth is that the use of behavioural data is growing and this naturally declines the need for surveys. Ultimately, surveys need to be smarter, to offer more unique insights if they are to compete.