Epidemiologists, who study viruses, measure two key things (among many others) to give them a sense of how a virus is spreading and behaving. Today we will learn the first one:
(Missed the first five lessons? Start here.)
As a designer, there is a good chance that you have never used the word incidence like this before, and there is a good chance that even your boss and colleagues haven’t either.
Viruses don’t just happen in a single moment. They spread over time. More people are constantly being exposed, and if we only look at conversion then we don’t get the whole picture.
Incidence is similar to conversion, which we measure often in UX, but it also considers time, and the groups of people “exposed” to the virus.
Let’s start with a simple example and build up.
Incidence Rate includes three things:
1) Number of new cases
Let’s say 1 million people downloaded your app, and 50,000 people have paid for premium features. That would be a conversion of 5% to premium features.
With me so far? Great.
Incidence also includes time.
We aren’t just asking how many people have converted in total, we need to know when.
We might compare the number of upgrades this month, with last month. Or this year and last year. Or whatever. Any time period will do.
This is different from basic conversion because when we count the upgrades this month, we’re not including the upgrades from last month.
Maybe this month we converted 4% and last month we converted 6%.
Still with me, right?
Incidence also includes population.
This is where it gets interesting. Analytics usually treat people as numbers, not people. When we do that, sometimes we lose some important information.
Imagine 10,000 people downloaded your app last month (Group A), and another 10,000 people downloaded it this month (Group B).
This month, 500 people upgraded.
There is no way to know whether those 500 people are in Group A or Group B unless we track it.
Maybe all 500 upgrades were from Group A.
Maybe they were all from Group B.
Maybe it was some of both.
Maybe some users canceled and then upgraded again.
Group A and Group B are “populations”. A virus spreads quickly, so the difference between Group A and Group B might be big, and the numbers get tricky if you’re not sure who is who.
We could also track Group A across several months to see how the incidence rate changes over time. But only if we don’t mix in other populations.
Evernote, the note-taking app, did an analysis like this. They learned that only about 1% of new users upgraded, but 25% of users who had used the app for several years had upgraded.
Overall their conversion rate was about 5%.
If that was your app, would you want to ignore information like that?
My goal for this lesson was just introduce you to the idea of incidence and some factors that can influence your numbers. There is a more scientific explanation of calculating incidence on Wikipedia.
Tomorrow we will finish UX Virality Week with another key measurement: Prevalence.