The worlds of churn prediction and its close neighbor product analytics are populated by manual event tracking. The most prominent event tracking software today is Google Analytics 4 (GA4), which enables users to track page views, user demographics and some very basic user interactions with a website. GA4 is very handy for that sort of thing, and should be an entry-level staple in every web developers toolbox.
At Kirnu, we have taken the manual out of the equation and replaced it with automatic and care-free. Read below what I mean by this and why automatic & care-free beats manual!
Manual tracking requires you to define the events to track in advance. Forgot to define a tracker for capturing the number of people clicking on "Contact support" in your app? Too bad, that information is lost. Kirnu's autocollect for user interactions collects all essential data for you, no questions asked. With no questions asked, there are no wrong answers and thus no possibility for error in capturing the required events.
Is there a thing as too much data? No, not really, as long as the data is adequately processed. While today's machine learning models can easily handle huge amounts of data, the data still needs to be processed. Ensuring a consistent and meaningful data processing pipeline is where the magic happens.
Manual tracking is a resource intensive task and requires the attention and co-operation of developers and product analysts / managers. Perfecting the what to collect (see shortcoming #1) might take several iterations and by then, the results of the analysis could become obsolete. To make matters worse, setting up manual tracking is only the beginning of the ride. Manual tracking requires maintenance and if you make changes to your product (app or website) the trackers need to be updated accordingly, which gets us to my third point.
From idea to execution, c. 2020
The word manual is engraved in manual tracking and by definition, the context for manual trackers is defined in advance. What if the context changes? The trackers will have to go as well, as they stop functioning as originally intended. Back to the beginning it is. This means that if you alter the structure / contents of your application, you will have to adjust the trackers accordingly, making manual tracking non-adaptive. Trackers that do not adapt make gaps into the data, potentially endangering prediction accuracy.
Kirnu's algorithm is built to endure change and evolves with the underlying context, so you do not have to worry about change. How? That is a story for another time.
It is true that the level of detail in data collected by automated event tracking is higher, possibly to the extent that any correlations are lost for a human interpreter (is my clicking the picture of a cat meme 3 times in a single session really increasing my retention?). The good news is, that for a computer, making the distinction between tomatoes and the tomatos is a task they were designed for. Having seen our machine learning algorithm at work, I am confident that our autocollection provides an answer to the shortcomings of manual event tracking.
Enter your email below to hear from us (maximum once a week) and to secure your place among early users! 🚀
P.S. we expect to be beta test ready during Spring 2021 and are looking for test users. Drop us a line at email@example.com if you are interested in reducing churn in your SaaS application and getting exclusive access to Kirnu.