Large data sets only really sing when combined with small data, in the form of specialist surveys and qualitative research
When I first started working as a researcher in a big advertising agency, one of my jobs was to ‘mine the data’. I always wanted to reply ‘what am I looking for?’ but assumed I should know.
Today’s companies, especially those operating in the digital arena, have access to increasing amounts of data – plus entire teams to ‘mine the data’.
But are the data analysts always getting the most out of what is at their fingertips – and coming to the right conclusions?
With more experience under my belt, I know that such large data sets only really sing when combined with small data, in the form of specialist quantitative surveys and qualitative research.
But why? What additional insights can the combination of big and small data bring you?
It is our belief that this combination will help by:
- Identifying the core behavioural challenge
Understanding the real opportunity for a brand comes from pinpointing the specific behaviour that needs to be changed.For example, in one study we found that many of those identified on the client’s database as ‘one-time users’, were actually ‘multi-time users’, driven to use different sign-ins by a generous sign-up incentive. This meant their Behavioural Challenge was not about getting users to come back as first thought. It was about rewarding repeat usage.
- Identifying and sizing the triggers and barriers to different behaviours
The beauty of most databases is they allow the identification of people adopting different behaviours (eg: ‘loyalists’ v ‘occasional users’ v ‘lapsed users’) – but not the reasons underlying that behaviour.Researchers can speak to groups of people in each category, and try to understand what might nudge them towards the desired behaviour.
- Minimising the risk of getting correlation confused with causation
In a large dataset it is easy to find correlations between different measures (i.e. different measures moving together) but statistics cannot tell you if these correlations are meaningful. Combining with qualitative research or perhaps a few experiments allows a deeper exploration of these relationships.
- Bringing the data to life By giving the stories behind the behaviour, people find it easier to remember and relate to the subtleties of what is going on. This helps them empathise and think of possible solutions.
For example, in one study the client could not understand why a competitor was doing better despite having a seemingly inferior product. Getting deep into the individual stories behind how people chose brands in the category suggested that while people said they wanted rational product benefits such as low interest rates etc, they actually chose the brand they had heard of before and they believed would quickly accept their application rather than a careful analysis of all the product’s benefits.
- Giving new hypotheses from which you can then go back and reanalyse the larger dataset
So in summary, big data and research need each other!
Big data should be looked at with a questioning hat, instead of a solution hat, providing patterns of what is going on – and sparking questions from which research can go in to do deeper explorations.