Sales of Widget Z are up 20 percent!
That’s what the headline figure on a chart in a meeting I was in the other week said. The presenter was thrilled and speculating about what could have caused such a jump in sales.
But that 20% tells you very little. 20% of what? Compared to when? I was unfamiliar with the client’s products and sales figures so I asked.
As it turns out, Widget Z had sold 10 units the previous week and 12 this week. Combined sales were in the tens of thousands of units and best selling products were selling in the thousands of units.
So an extra 2 sales of Widget Z was totally irrelevant. In fact it’s average weekly sales rate for the previous 6 months was just over 15 units, so it was actually performing below average.
This got me thinking about the importance of providing context when presenting figures. You might know the numbers inside out and have an intuitive understanding of how they all relate and change over time but don’t assume the person you’re giving the numbers to does. Even if they do, work on the assumption that they will give your figures to someone who doesn’t.
Dashboard design is something I’m doing a lot of at the moment, using Tableau and Google Data Studio so I got thinking about relating this idea to designing data dashboards.
Let’s say you’re doing a simple dashboard showing some performance metrics for a website, you’ll probably want to show it’s conversion rate:
Ok, so far so good, but is that good or bad? Lets add a comparison:
Better, but what’s the timeframe?
That’s much better, but it’s not quite enough, this figure could be skewed by a couple of extreme values, so lets add a time series:
The scorecard and time series work well together. We can tell it’s increased fairly significantly and if we look over at the time series we can tell that there was a dip but that we’ve now recovered and there seems to be a slight upwards trend.
Now we could drill deeper in to this, looking at a longer timeline, breaking out other dimensions like traffic sources etc but we can see the headline figure at a glance. Whether or not you need to drill deeper depends on who the dashboard is for, I plan to cover this in another post