A brief breakdown on how to find your most popular product feature in your data
Background
As a product manager or a data analyst, it is most likely your interest to know what is the most popular product feature. Knowing this helps you learn more about your users and gives you an indication on what to build next.
In this short post, I will provide a walkthrough on a few ways to dig this out from your data. This post assumes you have your data instrumentation set up. If not, here’s a good guideline.
Definition
Let’s define our goal clearly. By ‘most popular feature’, I mean ‘the feature which most of the users use the most’.
I will be using ‘event’ and ‘feature’ interchangeably so you can assume they are referring to the same thing. ‘Event’ here represents the tracked data point for a feature. For instance, if your product triggers a share event every time a user makes an online purchase, it will be a proxy to refer to your sharing feature.
Methodologies
There are generally 3 ways you can do this:
1 - Via event count
Get a list of events, sort by their event count — the event with the highest count will be your most popular feature. To get top N most popular features, you will need to apply judgement to the list, as some product feature may send an almost equal amount of events. Eg, event #1 and #2 might be triggered by the same feature, hence you will need to take event #3 as the 2nd most popular feature.
The downside for this is it can be swayed by a power users, as a feature is not necessarily the most popular one if a small number of power users use it a lot.
2 - Via unique user count
Get a list of events, sort by unique user count — the event with the highest unique user count is your most popular feature. At first glance, this looks promising as you have the feature which most users use.
However, this can also be misleading, as it factors in users who try it out (perhaps just out of curiosity) and does not consider the frequency of usage. For instance, you can influence this metric by placing it at a prominent place in your product.
3 - Via event which is ranked the highest for most users.
This is by far the most complete methodology I have come across. The idea is this: For each user, rank their most used feature (usually the highest event which is related to a product), extract out the highest ranked features and count their occurrence. This means that the feature which is ranked first for the most users is your most popular feature. This plays nicely to satisfy our definition — most of the users use it the most.
While I recommend this method, it may not be easily obtainable as most of the product analytics platform (eg Google Analytics) do not have this function readily. You will likely need to run an SQL query on top of the tracked events. For a detailed walkthrough which includes a sample SQL query, please refer to this article.
Conclusion
In short, our goal here is to find the feature that is used the most by most users as the most popular feature.
If you find your most popular via the highest event count, you will get the feature that is used the most, but not necessarily by the most users.
If you find your most popular feature via the highest unique user, you will get the feature that is used by the most users, but not necessarily used the most.
Finally, if you find your most popular feature via the event which is ranked highest by the most users, you will get the feature that is used the most by the most users. While this is recommended, it may not be readily available.
Thanks for reading this far!
Feel free to comment below if you have any feedback or questions, let’s discuss!
