Why prediction is essential for reporting

In the last post I discussed the concept of answering the obvious next question when showing data. This one extends that theme to show why reporting frequently needs to include forecasting or prediction.

Most maturity models of HR analytics show that reporting is the first step and needs to be achieved before moving on to more advanced techniques. In reality it’s far more intertwined than that.

In our usability testing one thing shows up time and time again. Contrary to what all the data visualisation purists suggest, viewers don’t read the data as it’s shown on the display, even if that data is visualised in the manner they suggest.

The data about what has or is happening that may be clearly displayed isn’t actually what is needed to make the decisions facing the viewer. To make a decision you need to have an understanding not of what has happened but what is likely to happen.

Users typically take the data that they see and project it mentally into the future. If a trend line shows an increase, the viewer is likely to believe that this will continue. If they see that Region A has a higher turnover to the rest of the firm they will interpret this as Region A having a problem (it could just be chance or a factor of the underlying population).

These behaviours and understood well in several disciplines. It’s the prime reason that fund managers show historic performance graphs and likewise the reason why regulators often insist that they put statements such as ‘historic performance is not a reliable indicator of future growth’.

We firmly believe that report designers need to address this issue, and that adding prediction to reports is often needed to provide this clarity.

What we want to do is to display two things:

  • A predicition of what the future holds based on a simple prediction model
  • A level of uncertainty associated with that prediction.

Again, we can turn to finance for a good example of how that is done. The Bank of England is noted for its use of the fan chart. This shows not only the most likely prediction but a range of likely values. It does so in a way that most lay readers can understand.

I wrote about a simpler example in an earlier article about prediction. In that instance we provided simply a percentage showing the probability of hitting the target. Hitting targets either will or won’t happen so it’s appropriate in these instances to show just the probability of hitting the target.

Whichever route is appropriate one thing should be clear – whenever you’re trying to present information that will be used for making decisions you should be providing guidance visually to help your readers make effective decisions.