The Data Dialogue Disconnect

Here’s one of the things that keeps me awake at night.

I'm pretty sure that most data analysts see their end product as a table or a graph. Maybe sometimes a collection of tables and graphs (which we'll often call a dashboard). But regardless of whether this end product is one exhibit or multiple exhibits, it doesn’t have a lot of words in it. And when we do show multiple exhibits, the exhibits are rarely joined together.

This 'joining together' of data exhibits – along with the words (written or spoken) that are needed to join the data exhibits together into a narrative – is something that analysts tend to regard as someone else’s responsibility: it’s the responsibility of decision-makers. And - on the face of it - I think this is a sensible way to think about it. Our job as analysts is to put information in front of decision-makers in a way that helps them gain a better understanding of the issues. Our job is to present information in ways that make visible the patterns and trends that might otherwise remain invisible. Our job is to display data in ways that help decision-makers join the dots. But it’s not necessarily our job to put the individual data exhibits into the right order or to join them up with a clear, compelling, watertight narrative.

So yes, analysts can do the data. But decision-makers can do the joining-up of the data.

But the trouble with this approach is that if decision-makers are going to put together arguments with the disconnected data exhibits given to them by the analysts, they need to have the right data exhibits in front of them to start with. This doesn’t always happen!

Moreover, if they think something’s missing, they need to know what to ask for in order to plug the gap. In other words, they need to have a reasonably clear sense of what is possible. What data is collected that might help? And what is it possible to do with that data once it’s been found?

And here is the rub. The people best placed to answer those questions are data analysts. It's the analysts who know what data is available and what – technically and visually – can be done with that data. So if there's a cause-and-effect hypothesis that needs testing, it's the analysts who are best placed to do it. They know exactly what data to look for, how to slice and dice it, and how to visualize it to see if the hypothesis makes sense or not.

But analysts can only generate hypotheses if they have meaningful connections with the real word of operational and strategic decision-making. And those connections are often missing. Analysts tend to see themselves first and foremost as technical data experts rather than as business - domain knowledge - experts.

So analysts tend not to spend too much of their time testing hypotheses because they either can’t or won't generate the hypotheses in the first place. That’s the job of decision-makers. But decision-makers often don’t pursue their theories or hunches because they don’t realise the data is available or crunch-able in ways that will help them.

Many years ago, in order to try and illustrate this data Catch-22, I drew a picture of a grey pyramid and a black box. To this day I still sometimes use it on training courses to ask the question: "Do decision-makers need to find out more about the data or do data analysts need to find out more about the business?"

But I no longer think of this an either/or question; instead, I think it highlights the need for a dialogue between analysts and decision-makers. Hypotheses are best generated and tested in a collaborative environment. But my worry is that the ways of working that have become commonplace since the pandemic (remote working, Microsoft Teams) have made these dialogues and these collaborations much harder to initiate, build, and maintain.