Andrew Wiseman reviews the big data science debate from the recent MRS conference.
Before I start, I suppose I should make an admission. Neither my colleagues at Honeycomb or myself attended the MRS conference last week – we were a bit busy launching something new and exciting! However, in the aftermath of both the launch and the conference, I spent a bit of time reviewing what had been spoken about, and was drawn to this article. The article offers a summary of a panel debate on data science featuring contributors from the Office for National Statistics, Sky, InterContinental Hotels, Diageo and GfK.
The headline “It’s so hyped” piqued my interest immediately and got me thinking about whether data science is all hype. Can it really have heroic qualities that can really accelerate growth for those brands who use it effectively.
Make no mistake, data science isn’t going away any time soon. Whether this is defined as regulation research analytics, machine learning, Artificial Intelligence or Deep Learning, there isn’t a day that goes by when another plethora of articles appear online. These articles either evangelise or play down the importance of data science in helping brands make smarter choices.
One of the key takeouts from the panel debate was that ‘big data’ (cue shrugs and sighs) is unruly and integrating these rich seams of data is not without peril. That’s not to say it’s not worthwhile. Our view here at Honeycomb is very much ‘just because you can integrate data together, it doesn’t mean you should’. The key questions that we’d always ask before looking to integrate datasets:
- What are the outcomes we’re looking to achieve by bringing these sets of data together?
- What added value will this combined data bring to the story we want to tell?
- Is there any causation between the datasets we’re looking to combine, or just spurious correlation?
In basic terms, if we can’t answer these questions, the use of data science becomes akin to looking for a needle in a haystack. Of course, you may find some nuggets of information – but will these be of interest, and more importantly, of value?
What this shows is that in practice, many of the principles that we’ve picked up as researchers and as analysts stand the test of time in this new, data-rich world: What do we want to find out? What are our hypotheses? What data might be useful to us? Can we bring this data to life to create compelling recommendations on which clients can act? For many, the lack of focus on these principles inevitably lead to ‘sailing the data lake with one oar’ – travelling around in circles trying to make sense of the data that has been collected obsessively.
For me though, the thought that rang true the most came from Diageo’s Andrew Geoghegan, who stated “data science is just another tool, like research and behavioural economics” – and he is absolutely correct in stating this. In short, data science will get you to a desired destination in the same way that a good programme of quant or qual research will. Our experience so far has been that bringing the relevant bits of these approaches together can add an additional lens of clarity – but like any programme of work, it needs to be guided by an expert pair of hands. If you can achieve this, data science need no longer be frightening, and you can start to benefit from the incremental gains that it can offer.