Keeping up with their promise of a release every quarter, Tableau’s 2019.3 version has just been released, and there are lots of new features to be excited about. From the Tableau Catalogue, available with the Data Management add-on for Tableau Server to Prep Builder’s new integration with R and Python, there’s a lot to explore.
In this post, we’re focussing on what’s new in Tableau Desktop. There’s particularly exciting news regarding the addition of AI in this release to help the everyday Tableau user perform more meaningful analysis with their data. We’re talking about Explain Data.
I started using Tableau earlier this year, but I have been working with data on a daily basis for the better part of the last three and a half years. And there is clearly a pre and post Tableau way of experiencing data exploration. Most – if not all – of you can relate to how tedious and inefficient data exploration can be when using a tool like Excel.
One of the aspects of data exploration that was made easier by Tableau was outlier analysis. By allowing to create views with only a few clicks, outlier detection becomes incredibly fast, thus leaving more time for proper analysis.
Enter Explain Data.
This new feature makes use of the power of AI to help uncover what’s behind outliers in data. We got a glimpse of how powerful it was when we were first introduced to Ask Data. Now, with a simple click, Tableau allows us to understand better the why behind unexpected values in the data.
Let’s use Superstore’s dataset to show how this feature works.
A simple scatterplot of sales versus profit for each customer shows some clear outliers in the data. At this point, we ask ourselves ‘What is it about these customers that makes them so profitable? ‘. The answer might uncover insights that could potentially be used to tap into a niche of customers. It may also lead to helping increase their profits as well.
Tableau suggests some explanations behind this customer’s high profit ratio by comparing that customer with all other customers in this set of data.
The first dimension for which Tableau finds a distinctive behaviour is State. About 90% of this customers’ purchases are in the State of New York (whereas the average of the other customers is around 10%).
Notice how in our first view we were only showing two measures (profit and Sales for each customer) and one dimension (the customer’s name). But Tableau’s Explain Data picked up on other dimensions not present on the view to explain the outliers, like state and status of the purchase.
Interestingly, a significant percentage of this customer’s purchases are being returned, which means that the profit ratio for this customer is far from what the scatterplot initially suggested. If you’re working with data that you understand very well, you’d probably be able to find this insight very quickly. That’s because you knew where to look.
Whether it’s an old or a brand-new dataset, a lot of what we do in exploratory data analysis is driven by our perceptions of the data. That makes it easy to miss essential hidden insights in the data.
But what’s behind the scenes, you ask?
According to Tableau’s website,
Explain Data uses powerful Bayesian methods to surface statistically significant explanations behind data points. Behind the scenes, hundreds of potential explanations are checked and the most likely ones are surfaced.
Besides Explain Data, there’s more to discover in this new release of Tableau Desktop, such as improvements to parameter actions, new spatial calculations, enhanced cross-database join control and much more.
I hope to get to these features in the coming weeks and I’ll share my experiences and thoughts with you.
In the meantime, happy to hear your comments about Explain Data.
