We spend a lot of time “looking under the bonnet” of survey research. That means getting our hands dirty with the data, and looking at it from the point of view of each respondent one at a time.
Rather like our qualitative research colleagues we want to know: “Why does this individual do what he does? Has particular preferences? Said what he did?” Starting from a set of principles we develop methods which enable us to process the data for each individual in turn to build an approximate model of the drivers for the individual. Now usually we will have only limited information for that individual, and so the model we have is only a partial model.
By bringing together these partial models across the whole sample, we obtain a robust picture of the market as a whole. At the same time we see how much variation there is across the population and can identify groups of people who are similar within each group but very different from group to group.
Another important activity that Marketing Sciences drive is linking research data to actual market data. Claimed purchasing behaviour in a survey is one thing, but the real volume of sales leaving store shelves may well be quite different. There are many reasons for these differences and over the years we have come to understand the mechanisms well, and have ways to calibrate the results.
If our research project shows that the impact of making changes to a brand’s line-up, then it is important that those results are expressed in terms that the client can understand. So we ensure that the base situation matches the brand volumes or shares that are recognised within the client’s company.
As we work closer with our clients, we often build simulators that show the impact on their business of different strategies and decisions, expressing this in terms of margin.