A little precognition is a handy skill to have – too bad it’s not readily available. What is available? Data. Lots and lots of data.
Your company sits on a goldmine of data – which you can use to predict user behavior. How? By finding meaningful patterns that allow you to see which customers are most likely to churn.
Let us gaze into our crystal data-sets…
Maybe you only log 5 types of user events: views, clicks, purchases, tutorial views and logins. Based on a user’s activity in the first week (or month) after downloading the app, can we predict whether that user will still be active in 6 weeks? Considering some statistics show that 80% of users only use a new app once – once! – we can easily predict a high attrition rate, but that’s just a guess. However, by looking at the views, clicks, purchases, tutorial views and login data, we can develop an early warning system for attrition.
Take your event log data for the first month, or first week, whatever the most useful timeframe might be, then look at which users were still active during week 6.
Let’s say the crew of Serenity is interested in your new navigation app. Their user events might look like this during the first week.
|Captain Mal||Oct 1||Week 1||Viewed a deal|
|H. Washburne||Oct 2||Week 1||Viewed a deal|
|Simon Tam||Oct 3||Week 1||Viewed a deal|
|Kaylee Frye||Oct 1||Week 1||Viewed a deal|
|Kaylee Frye||Oct 2||Week 1||Clicked a deal|
|H. Washburn||Oct 2||Week 1||Viewed a deal|
|H. Washburn||Oct 3||Week 1||Clicked a deal|
|H. Washburn||Oct 3||Week 1||Bought a deal|
|Kaylee Frye||Oct 3||Week 1||Bought a deal|
|H. Washburn||Oct 4||Week 1||Viewed tutorial|
|Kaylee Frye||Oct 4||Week 1||Viewed tutorial|
When you review the data from 6 weeks later, you see this:
|H. Washburn||Nov 6||Week 6||Logged in|
|H. Washburn||Nov 7||Week 6||Logged in|
|Kaylee Frye||Nov 8||Week 6||Logged in|
Now you can go back to that first week and look for patterns between who did, and did not, log in during week 6. (Note: logins alone don’t matter.) Mal and Tam were clearly just looking, but Kaylee and Wash both viewed deals, clicked, bought and viewed the tutorial. Of those four, viewing the tutorial might be the best indicator of engagement. And, if I were to take this chart further, I’d look for other engagement indicators like re-viewing the tutorial or clickthroughs of all the app pages and options.
You don’t need a lot of data to find informative patterns. Even if you only track views, clicks and purchases, you can see whether users who click more than X times in Week 1 are more likely to be active in Week 6.
No, you don’t need a Data-wizard, there’s an app for that.
You don’t have to hire a data-wiz or find these patterns on your own (which is great, because you probably have a lot more data than just four characters from Firefly). Tools like BigML can look at your data and find useful patterns to predict churn.
With the help of a churn-prediction tool, you can see which segments of users are more likely to churn in a given period of time, and which segments are more likely to keep coming back. Most importantly, you can find the specific behaviors that indicate someone may leave soon – and that’s your chance to swoop in with your Customer Success team to find out why your product isn’t helping that user meet his or her desired outcome. If your product is a good fit for that user, you stand a solid chance of not only re-engaging the user, but really impressing them!
But, you might find that the user segment most likely to leave is leaving for good reason – they thought your product solved a problem it doesn’t. This is a cue that your marketing messages may be off, and by changing them, you can bring in more of your ideal customers who are more likely to remain active and engaged for weeks, months, and years.
For more insights on SaaS marketing and customer success, sign up for our free community. The place where SaaS leaders share posts and information all SaaS related.