One of the most common issues I see with my clients is they simply don't know the adoption rates of their features. I've worked with many product managers that are eager to understand repeat usage rates, stickiness and a host of other advanced metrics on their features, without knowing the most basic metric, adoption.
In this post I've going to share with you a product analytics framework I've developed called the feature adoption funnel. The main purpose of this framework is to help the company as a whole understand feature adoption.
What is the feature adoption funnel?
The feature adoption funnel is a 4 step funnel which can be used to measure every feature in a product. The feature adoption funnel is a high-level way to understand the most basic usage metrics of your features.
Below is a basic representation of the feature adoption funnel.
Step #1 - Exposed
The first step in the feature adoption funnel is "exposed". In this step we want to measure the percentage of our users which were exposed to the feature. The idea is that we can't necessarily blame the feature itself for poor adoption. We need to know how many people could have activated the feature and started using it.
The easiest way to measure exposure is to measure the percentage of users which viewed the page where the feature is presented. You want a yes / no indication.
So lets say for example you have 1,000 users that signed up in September. Out of those 1,000, 450 of them landed on feature X's page. This would mean an exposure rate of 45%.
Step #2 - Activated
The second step in the funnel is "activated". In this step we want to measure the percentage of users which actually activated the feature.
Some features don't require activation, like for example importing a list of emails into Mailchimp. If the feature you're measuring doesn't require the user to perform any action to activate the feature then you can skip this step entirely.
The best way to measure activation rate of your features is by using an event-based tracking system like Segment. Most startups would simply look in their database if the feature is enabled but this an incorrect approach since features can be disabled. You need to create a historical record when it comes to enabling and disabling features and Segment can help you do this.
If you are saving a historical record of feature activation then you'll want to look for the first instance where a user enabled the feature. Just like in step 1 you want a yes / no indication per user.
So lets say that out of the 450 users which saw the feature, 300 of them activated the feature. We can now say that your exposure to activated rate is 67% (300 / 450) and your activation rate for the September cohort is 30% (300 / 1000).
Step #3 - Used
In the third step of the funnel we want to measure the percentage of users which actually used the feature. Just so it's clear we're talking about used the feature at least once. None of the steps in this funnel have anything to do with time. We are simply measuring the percentage of users which did each of the steps ever.
How a feature is used is very different from feature to feature. Some features like adding a bio to a profile for example are very straight forward while others like setting up an integration are much more complex. Like step 1 and 2, we simply want a yes / no indication per user. Yes if the user used the feature at least once, and no if the user never used the feature.
With some features simply activating them results in usage and the user isn't required to do anything else. If this is the case with your feature then skip this step.
Out of the 300 users which activated your funnel, only 50 used it. We can now say that your activated to used rate is 16% and your overall used rate is only 5%.
Step #4 - Used again
The last step in the funnel is "used again". In this step you are going to measure repeat usage and turn that into a yes / no indicator per user. The idea behind this step is to see if once someone uses a feature if they are likely to use it again. This is a very high-level indicator and you'll want to dive much deeper into repeat usage as a separate analysis.
We see that out of the 50 users which used the feature at least once, 45 of them have used the feature more than once. This means our "used again" rate is 90% (45 / 50). Our overall "used again" rate is 4.5% (45 / 1000)
What can we learn from the feature adoption funnel?
The feature adoption funnel behaves just like any funnel when it comes to analyzing performance. We can analyze the funnel both from one step to the next, or from the start of the funnel.
If we analyze our example we can see the following:
From start of the funnel
- Exposed Rate - 45%
- Activated Rate - 30%
- Used Rate - 5%
- Used Again - 4.5%
From previous step of the funnel
- Exposed Rate - 45%
- Activated Rate - 67%
- Used Rate - 16%
- Used Again - 90%
We can clearly see that overall the adoption rate is very low (only 5%) for our hypothetical feature. What also stands out is that there is a big drop between activation and used, and a very high rate between used and used again.
These findings tell us that for some reason few users that activate the feature are using it but out of those that do use it most use it again. This indicates to me that there is either a bug or technical issue preventing users from using the feature, or that the process to use the feature is too complicated for most users.
Can you see how powerful the feature adoption funnel is for helping product managers understand where they should focus their attention to improve usage?
What else can I learn from the results of the feature adoption funnel?
Low exposure rates means that product marketing need to up their game
If you notice that you have very low exposure rates to your features it could indicate that either you have a bad product market fit (AKA low retention) or that product marketing needs to do more. Be more aggressive with your marketing automation and email tools to communicate the value your features provide users.
Low activation rates means poor messaging, bugs or unattractive features
If your exposure rates are fine but users simply aren't turning on features it could mean a few different things. The most common will be poor messaging and unattractive features. Most products suffer from feature bloat and many features just don't appeal to the majority of users.
Another culprit could be bugs which are preventing users from activating features. Make sure you periodically run through your product and make sure that everything is working as expected. QA should have automation tests set up and your analysts should also help catch significant bugs by keeping an eye on your numbers.
Low used again rates indicate that users did not get value from the feature
This one is quite obvious. If a user used a feature and didn't get value from it the chances of him using it again are low. If you have poor used again rates then it might be time to scrap some features.
The feature adoption funnel is an optimization framework
One of the most powerful aspects of the feature adoption funnel is that it can be used to optimize your feature usage over time.
The screen shot below shows this perfectly.
All diagrams in this post were created using Lucidchart.
The feature adoption funnel should always be used in relation to cohorts. Product managers can use different cohorts to see if they are optimizing key usage KPIs over time. The framework keeps the product team focused on nailing the basic usage metrics first before getting caught up in the weeds.
Final thoughts on the feature adoption funnel
There are a ton of nuances that you will have to take into account when using the framework. For example, there are certain features which aren't necessary relevant for all users. In this case you'll want your cohorts to reflect that.
You might have a feature which is aimed at a specific geography. In this case it would be useful to see adoption by different segments of users by geography. Other common use cases include company size, number of users using the service, and lifetime value.
In the end of the day you should set realistic expectations when it comes to the results of your feature adoption funnels. It's probably a good idea to filter out all the users which never successfully finish onboarding, and any other relevant filtering that is needed to reduce the noise. This will be especially helpful if you run a B2C company.
