I hate the phrase “data driven”, as in “data-driven initiative”. I know hate is a strong word, but fear not. This is not a lecture on data science.
It’s about demystifying what we do in ‘data’ – that part of the office that many think of as being protected by a Star Gate level force field; or, at the very least, ‘Beware All Ye Who Enter’ signs. What the inhabitants of that part of the office keep secret will all be revealed here.
What is data science, really?
At its core, data science seeks to answer deep questions. We don’t ask “How?”; we ask “Why?” Because this gives us the scope and information to churn out the valuable analysis that inform our insights and initiatives.
It’s what allows us to present a set of possibilities to interpret actions taken, then merge that with other peripheral data sources – like the weather on that particular day – or any other number of factors that may have contributed to the action taken or choice made.
Data science in action
To show how this works in practice, let’s compare the effect when we phrase a question differently. For example: “How did the driver change lanes?”
The answer is a function of their driving ability, learnt through months of K53 lessons, combined with the inherent capabilities of the vehicle they were driving, to complete that goal. In its most simplified form, it would look something like this: They used the car’s indicators and side mirrors and turned the wheel at an appropriate time to change lanes.
Now let’s consider: “Why did the driver change lanes?”
This is a richer, more nuanced question that requires further inquiry as to the initial goals when the driver set off, and the conditions under which they found themselves prior to the act of changing lanes. Was this a result of their predetermined plan, or was it a reaction to an incident or something they spotted up ahead in real time?
The answer to ‘why?’ is less direct and has a number of variables that contribute to it. And that’s where it starts getting interesting.
The complexity of ‘why’
As is evident from the example above, “why?” is a more complex, nuanced question than its “who”, “what”, “when”, “where” and “how” and counterparts.
We’ve all seen this play out. We’ve seen how advertising has gone from LSM segmentation to behaviourally segmented audience modelling, how lookalikes exist as a predefined category and not a poorly groomed man posing as George Clooney with a five o’clock shadow (begging the question ‘why?’ again).
So, here’s why I think of data science as ‘experimental baking’ rather than alchemy. Alchemy implies turning something mundane into something valuable using some secret formula. There’s no secret formula to data science; like baking, it’s about tweaking variables within a known formula to produce variations of the result.
As with baking, we know the basic ingredients that go into it, and we understand the outcome. And we can experiment with different ingredient combinations and processes until we reach a result we would be proud to present to the client (or Bake Off judge).
Agency environment
Data science in the agency environment rely on the ability to take readily available information, assumptions and peripheral information streams into account when seeking the data output required.
That output provides a reason – or different possible reasons – for an individual choice or action at a given time. Once we understand those reasons for those actions, we can use that information to influence behaviour.
This is why I believe we need to replace the phrase “data driven” from our vocabulary with “reason driven” initiatives instead. After all, the reasoning behind why you do something is far more valuable than the data that represents the outcome of that choice.
Ryan Campher is head of data strategy for VMLY&R South Africa.