Retention isn’t the holy grail of metrics, but its importance is too often underestimated. Most product developers overlook retention as a secondary metric, one that should naturally come after the growth stage of a product. Those who go about building a product in such a way indubitably either question their original decision of pursuing growth before anything else or inevitably fail altogether.
For many reasons, retention is the only metric that any young product maker needs to take seriously. Above all else, the best sign that you’re providing value is that people use what you’re building. And more than simply using what you’ve built is that they use it again and again; it becomes a recurring solution to a problem or need that they face.
Just as the activation metric relies on you being able to define a meaningful interaction that showcases the value of your product and drives the user to re-engage, retention must be defined as one or several meaningful interactions that a user regularly engages with. The idea behind retention is that you’re able to provide regular, recurring value to your users because of a single or a set of single actions that help the user respond to a need or solve a problem.
As you’ll see retention can be measured in 2 ways. The first is by analyzing the frequency of use of your product. The second is by analyzing the efficacy of your retention funnels.
Frequency of use
When establishing criteria for the frequency of use, as a product developer, you’ll likely have an idea of how often a user should use your product. If your analysis suggests that a user is less frequently engaged with your product than you expected, then there are 3 ways of interpreting this information.
- Your estimate of how often users should use your product is off. Some products are meant to be used daily like email. Some products like car hire might be considered to be used weekly. Products around purchasing a home might be used every decade. Although you might make assumptions about the frequency of use of your product, you may have to adjust your perceptions to accommodate the needs demonstrated by your users.
- You need to modify your product to increase usage. Sometimes products start out with low usage, but by making progressive, incremental changes, you can increase the reliance your users have on it. Perhaps an app that proposes discount vacations is interesting to its users because of its unbeatable prices. Its users express their happiness, but few come back for to reserve again. It could be that by proposing a calendar integration, it makes proposals when a user doesn’t have any weekend plans in order to bring users back into its ecosystem and generate more frequent reservations.
- Your product’s value proposition is less significant than what you thought. The most common case of low recurring usage is due to a lack of perceived value. Some products simply miss the mark. It’s important to be honest with yourself about the expectations you have for anything you build. The faster you come to the conclusion that you’ve missed your mark, the more quickly you can recover by pivoting or simply abandoning your product.
In some cases, you’ll find that your analysis beats your expectations. If you notice this, then you’re one lucky guy or gal. This means that your users are finding value in your product at a level that’s higher than what you expected. You can continue to optimize the frequency of use of your product by experimenting, a subject we’ll get into a bit later.
Retention funnel efficacy is about how well you do (or your product does) at bringing users back into your product ecosystem and engaging with your core value proposition. Some products naturally draw their users back in because they appeal to a deep or psychological need. A dating app like Tinder for example can re-engage its users because of its core value proposition: an opportunity to meet someone that likes you. In the context of retention, it doesn’t get better than this.
Retention can also be artificial. Those notifications that have been blowing up your smartphone for the past few years – that’s retention at work. Tinder also uses this type of retention. When someone likes you and you’ve already liked them, you receive a notification. The intention behind this notification is to nudge you to start a conversation.
Products that have a natural retention mechanism are those with which we find it hard to live without. If we dig deep enough into this concept, of course we can live without most of the modern technology we’ve adopted, but in today’s world, most of us can’t go anywhere without a phone, a tablet or a computer.
Think of the apps you use regularly without any particular nudge from the product creators. Could you get anywhere without Google Maps or Citymapper? Would it be possible to collaborate efficiently on a document without Google Docs? How did you listen to music before Spotify?
Most of these products have done so well at designing their experiences, it’s difficult to imagine a time when they didn’t exist. Let’s take the example of Spotify. Could you imagine carrying around a discman and a book of CDs on the subway? Why would you ever pay per song on iTunes?
It’s hard to express how difficult it can be for a product to arrive at this level of integration in its users’ lives. Most of us must accept that we’ll never build something so “sticky,” but it’s worth mentioning that when it comes to retention, this is the ideal.
For those of us unfortunate enough to have not developed a product so integrated into our users’ lives, it becomes necessary to rely on strategies that draw users back into our product ecosystems.
Induced retention is somewhat artificial on its face, but it can generate a tremendous amount of value not only for the creator of a product, but its users as well. In fact, this is the main goal of induced retention.
If users aren’t completely absorbed by your product, then they’re probably missing out on a lot of what you have to offer. To keep users engaged with your product, we most often use a notification strategy. Notifications can take many forms. They’re not just the beveled rectangular bars that show up on your phone’s lock screen. They can take the form of emails and messages as well.
It’s important to note that when we talk about induced retention, there are varying degrees to which this is executed. Have you ever signed up for a service and received 5 emails in the day or two following your signup? Sometimes product developers can take this a bit too far and alienate their users by what adds up to a spam strategy.
Good induced retention strategies do their best to deliver value and avoid unnecessary communication with users. As we’ve discussed, the boundary between good and bad design isn’t subjective. It’s very much objective, which is why induced retention must be measured.
Just like acquisition and activation, induced retention is measured through a funnel. The peculiarity about induced retention is that it can very well be made up of tens, hundreds and even thousands of funnels. Each retention campaign you run can result in a new funnel.
When you shoot your users a notification, a message or an email in the context of retention or re-engagement, you’re providing them an opportunity to re-experience your value proposition. For each of these campaigns, you’ll need to redo your analytics strategy in order to track the conversions of your users through each of these funnels.
You’ll likely want to compare your different campaigns to each other in order to decide which ones are more effective. You can also use the data you’ve collected to understand the impacts of your campaigns on subsets of users, grouped by common demographics or behavioral characteristics. We call this segmentation, and we’ll get into the details of segments in another article.
While funnels are an excellent way to visualize drop-offs and the efficacy of the sequence of interactions you propose to your visitors, they don’t offer much insight regarding the long-term performance of your campaigns. Funnels are great at analyzing the immediate outcome of a marketing campaign or a retention strategy, but it’s just as necessary to measure the long-term effects of your strategies to inform your future decisions.
For this reason, we use what is called a cohort analysis to measure the performance of your campaigns on your retention rate, or the ratio between the number of users who are acquired, activated or retained in a given period versus the number that are retained in a subsequent period.
A cohort analysis groups visitors or users together by the time at which they interacted with your product, such as when they created an account or completed the activation interaction. You can then subdivide these groups by source (the sites or domains they arrived from), by marketing medium (backlinks, newsletters, ads, etc.) or any other parameter.
For example, you may have had 200 visitors that created an account during the first week of September, 187 that did so during the second week of September and 273 during the third week of September.
These groups can then be segmented or broken down by medium, channel, source or any other parameter of a campaign. Perhaps among the 187 new users in the first week of September, 21 interacted with Facebook ads, 43 came from Google Search and 56 are associated with traffic from your Medium blog. You can do this again across the same parameters for the subsequent cohorts in order to maintain consistency in your analysis.
After you’ve defined and broken down you cohorts, you’ll want to see how many of these users interacted with your product over time. The goal of this analysis is to identify which acquisition, activation and retention strategies result in the highest amount of engagement over the longest period of time.
Each cohort of users is analyzed relative to one or many potential goals that you want them to complete in their 1st period of usage, 2nd period of usage, 3rd period of usage, etc. A period of usage can be an hour, a day, a week, a month or any other period that categorizes the expectation of the frequency of use you have as the creator of the product.
In effect, if you expect weekly usage by your users, then you would analyze how many of your new users acquired in the first week of September were retained in the 2nd week of September (period 1 for this group), then the 3rd (period 2), 4th (period 3), then the first week of October (period 4) and so on. For those users acquired in the 2nd week of September period 1 would correspond to the 3rd week of September. In this analysis the periods of use are relative to when the users were acquired. The analysis is therefore staggered. You’ll always have an additional data point for each earlier week in the analysis.
What a cohort analysis shows is the increases and decreases in long-term retention based on the modifications you make to your product and feature set and which sources of traffic or funnels are most correlated with long-term engagement.
Want to learn more about how to implement product analytics for your company? Download our paper, “The 3 components of behavioral analytics for products.”