Why Premature Viz might be bad for you

Category : Article

Written with my colleague Alex Lea @alex_lea_8

So we all know the key to seeing and understanding your data is to be able to explore it fast. Being creative, chasing ideas, and feeling your way to the insight. That’s why we love Tableau. It doesn’t matter if something doesn’t work. You didn’t spend much time creating it in the first place – let’s ‘fail fast’ is the cry. Or as Beckett would have it:

Ever tried.
Ever failed.
No matter.
Try again.
Fail again.
Fail better.

Along the way you might create lots of ugly looking charts. But you know each one helps provide the full picture of your data. We’re not trying to build the final dashboard. You’re in data discovery mode. Or to quote Andy Cotgreave, you’re happy to build,

 “100 ugly visualisation rather than one beautiful one”

Or rather, 100 visualisations that get me closer to my final dashboard.

But therein lies a problem, and I’d like to talk about something serious for a moment. Something that may have affected a lot of people…

Premature Visualisation

Building too much, too soon. The downside of fast visualisation is swamping users with a tidal wave of your ideas. Ideas you expect users to immediately understand. Just because we can produce a lot of content, it doesn’t mean we perhaps should. Or, if we do, then we should understand when to show it to someone else.

So one danger of this fail fast approach is that we could be leading to more viz, but not necessarily more useful viz or better analysis.

To paraphrase Don Lucchesi in Godfather III,










For one recent customer, we created over 170 views. A blizzard of dots on maps, heat maps, a calculation of catchment areas, breakdowns by urban/rural and, finally, a cartogram. Now as much as I love them, I can’t help wondering whether cartograms are the GPS of graphics. They are there to tell you that you have gone too far in the data discovery process.

After 170 views requiring various Tableau pyrotechnics. Plus a plethora of meetings with the user(s). This work went all out of the window, and this is what provided the evidence and ended up in the first report:







Yep, a bar chart. Simple as that.

What went wrong?

The users didn’t know what they wanted. So we gave them 170 views. We tried to sell them Tableau before we had sold them data viz. We ended up telling them ‘you could visualise it this way’ when we really needed to create the right conditions in which they said ‘can you visualise it this way’? With Tableau the answer is invariably, ‘yes’. We tried to sell them 170 ideas. We failed. Now, we’re not salespeople, we’re analysts. Researchers. But maybe we need to get better at pitching.

Or maybe we need to start by giving viz consumers less choice.

If you look the next time you’re at the supermarket, there’s an entire aisle of jam. I don’t want one jam to choose from, that would be boring. Maybe I don’t like pips. Or raspberries. But when I see the jam aisle, my decision-making part of my brain just throws its arms up in the air and says, “this is madness, just grab any old one.”

Too much choice can impede our decision-making process. Perhaps there’s a sweet spot (excuse the pun) in the amount of choice you should present a user. (Worth two minutes of your time is a short, but slightly sweary, sketch by Fry & Laurie on the proliferation of choice https://www.youtube.com/watch?v=6T2zUEiVQU4 . For more examples of how consumers deal badly with choice, read Philip Graves excellent book Consumology: The market research myth. )

Did the 170 views achieve anything?

Yes. It helped us to hone and refine our understanding of the data. It helped us understand what was interesting, what was important and the difference between the two.

We needed to run through a lot of different visualisations to make sense of it. We showed the business that the data was noisy and messy. We dispelled myths and preconceptions that existed. It helped refine the questions that were important for the business. But it also showed that we need to be better at selling our ideas. Like a film director, we need to know what to show our audience and what to leave on the cutting room floor. We need to construct a convincing story.

So don’t show the full-length trailer. Start by showing a teaser. Don’t then show the director’s cut. Let them see the theatrical version. Then sell them the extended edition Blu-ray with blooper reel at a later date.

It’s best to keep your audience wanting more. To keep them guessing and get them to ask their own questions and fill in the blanks themselves.

As in the best Hollywood movie, there was a twist. As our project continued, views dismissed out of hand at the beginning started to become more relevant. The users finally worked out that they could also edit the script.

In short, being a good analyst is a necessity, but you also need to be a good director.

Leave a Reply

Your email address will not be published. Required fields are marked *