Professional Documents
Culture Documents
Heap Guide To Retention Ebook PDF
Heap Guide To Retention Ebook PDF
Retention is the True Measure of Product Value It’s for these reasons that Product Managers should
Building a product-driven business takes energy. There’s be somewhat obsessed with retention: measuring it,
marketing to generate interest, and a sales apparatus to understanding it, developing their product or features
bring deals home. On the product side, there’s the initial user with an eye towards improving it. Other growth metrics
experience to develop, and the value your product should matter too, of course. Acquisition numbers tell you how
demonstrate from the get go. well you’re able to sell your product to customers who
haven’t used it yet. Activation numbers tell you how many
These activities are all important for getting customers in people you can get into your product. Retention alone
the door. But once people start using your product, it’s your tells you if customers are finding ongoing value in
retention strategy that keeps them there. Plenty of users your product.
are willing to try a product once or twice. But if your product
or feature doesn’t provide ongoing value over time, or is too
difficult to use, or if the pains your product addresses can
For hypothesis-driven PMs, there
be more easily solved by a competitor, those users won’t
stick around. will rarely be a time when increasing
retention is not a major goal.
The diagram at left gives a quick explanation. To the question, Impact of Monthly Churn Across a Year (# Users)
“would you be happy with a monthly retention rate of 95%?”
most product companies tend to answer yes. (When we ask this
100
5% 100 95 90 86 81 77 74 70 66 63 60 57
6% 100 94 88 83 78 73 69 65 61 57 54 51
At right you can see the difference between an effective and a 120,000
If you’re at the early stages of this graph, your growth rates may 20,000
1/14
3/14
5/14
7/14
9/14
11/14
1/15
3/15
5/15
7/15
9/15
11/15
1/16
3/16
5/16
7/16
9/16
11/16
1/17
3/17
5/17
7/17
9/17
11/17
If you’re regularly measuring retention and are observing a graph
that looks like this, it should be cause for alarm.
Good Retention Example
On the bottom you can see what a user graph looks like when
9,000
a company has a good retention strategy in place. In this case 8,000
the number of users continues to increase over time, ensuring 7,000
that your company’s ability to retain users matches its ability 6,000
4,000
3,000
2,000
1,000
-
1/16 2/16 3/16 4/16 5/16 6/16 7/16 8/16 9/16 10/16 11/16 12/16
The graph at right—whose shape is typical for many Stage 1 Stage 2 Stage 3
companies—shows the standard way we measure retention.
The initial number—62%—represents the number of acquired
users who actually engage with your product. Typically this
70%
number drops significantly after a short period of time, here
60%
a month, and levels off into a more standard user rate. That 50%
number is generally what you’ll use as your baseline as you run 40%
20%
This number can also be helpful for forecasting revenue. If you
10%
know that 16% of acquired users are likely to keep using your
0%
product and you have a decent idea of how long they’ll stick
Same
month
After 2
months
After 4
months
After 6
months
After 8
months
After 10
months
After 12
months
around, you can produce a useful forecasting model.
In this graph, you also see how user numbers slowly decrease
over the course of the year, indicating slow but steady churn. NOTE: Retention doesn’t have to be measured on the level of
The final significant drop occurs at the end of the year, when the your entire product. It’s just as important to track retention of
large majority of users still in the product or using the feature a specific feature. Doing this can help you decide whether it
drop off entirely. makes more sense to prioritize further development, to sunset,
to refocus, to simply to do more investigation.
As a Product Manager, it’s your job to know the retention
rates at each of these stages—initial use, longer sustained
period of time, and a larger drop off at the end—and to develop
hypotheses around improving them.
Improving Retention that the “visits” metric doesn’t tell you what actions users
are taking in your product. For many products, this can be
a problem.
Defining Retention
For example, many gaming apps add “daily reward” games
In order to develop and test hypotheses about increasing
to the user experience, where users who visit the product
retention, it’s important to first measure your retention
on successive days win a prize. (Often the size of the
rates and assemble them in a graph like the one on the
prize increases the more days in a row a user visits the
previous page.
site.) But these kinds of visits rarely correlate with
long-term retention.
Doing this can be tricky. At bottom, your retention metrics
should measure how many users performed an event
once, and then performed that event again later. The main
difficulty is that it can take work to decide what event in
your product counts as “usage.”
The key is that when defining retention, you want to brainstorm 1 YEAR 2 YEAR 3 YEAR
with your team, and home in on the sorts of behaviors you want
your product to drive long-term. Evernote’s famous “smile graph.” The longer a customer
used Evernote, the more likely they were to return. That’s
stickiness.
Since Heap captures all your data and enables retroactive —— Granularity: This determines the interval by which we view
analysis, it’s easy to look back and see how many people took how often a person completes the return event. It can be
a given action once, then when and how often they took that set to Day, Week, or Month. The granularity you choose
action again. Here’s how. depends on the expected usage of your product (is it the
sort of product that people should come to twice a week,
—— Navigate to Analyze → Retention. Ensure the Retention
or once a month?).
Analysis option is selected at the top, then select
the following: Run the query and you’ll see a line graph; you can also switch to
table view to see a breakdown of activity for each cohort.
—— Start Event: The start event is the event that you’d like to
use as the foundation of the retention report. This would
generally be something basic that indicates initial interest in NOTE: this is only the most basic way to run a retention query.
your product, like “signup,” “start trial,” or “purchase. As you develop your analytics practice to run more hypotheses,
you can start comparing customer segments and further break
—— Return Event: Generally the repeated action that you want to down your user base. To learn more, see Heap Docs, or watch
see over time. This should be the event that best defines a the Heap University training course.
user’s getting value from your product (as described on the
previous page).
Dashboard
Retention Retention Analysis
Reports
Start Event
Analyze Session
Session
Funnel
Paths
Date Range Past 7 Days grouped by Day
Queries
First Time
Define
Run Query
Capture
Activate
Users retained from Click Login to Click - Save Report comparing 1 segment
Account
3. You can figure out how people can continue to get value Stage 1 Stage 2 Stage 3
Get more Raise this Get this line
from your product, even after using it for a long period 70%
new users to
stick around
line, get more
people to find
to flatten
out/get more
repeatable value people who
of time. longer
continue to
60% find value
long-term retention.
20%
After 2
months
After 4
months
After 6
months
After 8
months
After 10
months
After 12
months
strategies that can help.
They tend to cluster around a few key areas, though as always Cross-Functional Efforts
you’re encouraged to come up with other creative ideas. Cross-functional efforts can go far towards increasing the number
of initial users. If launching a new feature, you can write blog posts
about it. You can run an email campaign, targeting specific users
Trial Periods
whom you think would be most interested.
Depending on your product, you may want to offer (or may already
offer) a trial period. There are many experiments you can run here. You can hold a launch party. You can do sales enablement to get your
You can extend your trial period, or shorten it. You can add more salespeople to mention the new feature more often. You can identify
features to your trial period, or remove them. You can ask users to a group of users from whom you’d most like feedback and interview
provide a credit card number before the trial period, or move that them. As with any business proposition, the possibilities truly are
request to the end. And so on! limited only by your time and creativity.
Tiers of Usage
Spend time identifying what the beginner, intermediate, and
advanced features (or types of usage) in your product are. Then
figure out how to help users at different stages uncover them. You
Deprecate
can do this in your product (with in-app guides, say or by targeting One powerful strategy is to deprecate features that are not used
pop-ups to people who have signed up months ago) or through often but which make your product more complex. PMs can hate to
marketing and outreach (you could have your Account Managers call let go of features they’ve had a hand in creating, but doing this—or
customers and make sure they know about your new feature, or you testing the possibility out—can make for a simpler, easier-to-use
could launch a social media or paid advertising campaign). product, and more quickly get users to value.
SCENARIO 3
That said, there are a number of areas to investigate and
experiment with. If you’re only dealing with a feature, the
discussions may not be as intense. But they still merit the same
Hypotheses for Creating kind of attention.
Longer-Term Retention
As mentioned above, if you’re witnessing a major drop-off after a Examine Product-Market Fit
period of time, it’s imperative that you deal with it immediately. As before, these discussions will touch some more strategic issues.
If you don’t fix this problem, you’re putting your entire business Take a deep dive into your product and business model to assess
at risk. (If the issues is retention of a feature, your whole whether you’ve really achieved product-market fit. Are you selling
business may not be at risk, but it may be worth considering the wrong product? Are you selling to the wrong people? Is your
whether the feature is needed or not.) pricing model appropriate?
End-stage drop off is a thornier issue to tackle, since the The issue could come from any part of your business. Are you giving
customers enough support? Are you following up in the right ways?
problems involved may have little to do with the product itself.
The issue could be a new competitor to the space. Or a new Figuring out how to solve these issues will likely involve other key
direction in the way your customers work or do business. (It’s stakeholders in the company. Should you expand your product?
not really the horse’s fault that it got replaced by the train and Change your pricing model? Sell to a different market? And so on.
the car.)
Talk to Customers
If you’re not doing this already, it’s critical that you talk to
customers who have churned. Find out why they’ve left and 2013 2014 2015 2016 2017 2018
take copious notes. Base your decisions on this information.
Dropbox’s retention graph, from the company’s S-1.
Nice and healthy!
Start Event
Return Event
The first, “Saving Reports,” told us that when a Heap user saves Execute Query
a report in their first session, they are 2x more likely to be long- Group By Has done sequence Accept Invite → Save Report
is likely to indicate that a user is actually using Heap to run Compare Segments Add Segment
queries and produce an analysis that they found important Date Range Dec 31, 2016 - Apr 27, 2018 by Week
enough to save.
First Time
The second metric, “Group by” told us that when a Heap user
uses the “Group By” feature when running a query, their Retention Start Event
engagement rates tend to be 2x those of users who don’t use Start Event
Return Event
As noted above, we chose “Group By” because it was a feature View - Any App Page
that other tools, including Google Analytics, do not offer. So Group By Add Group By
repeatedly using “Group By,” we hypothesized, indicates that a Filters Add Filter
As with “Saving Reports,” this query didn’t tell us whether it was Has Run a Query
the act of using Group By that caused users to be retained (that Date Range Past Year by Month
by using Group By they realized how powerful Heap was, for First Time
example), or whether the fact that they used Group By indicated After 10 months
something else about these customers that made them more Has Run a Query
w/ GroupBy
51.86% (-10.15%)
likely to keep using the product (that they had already found Has Run a Query 21.85% (-22.84%)
A Hypothesis Story:
Suggested Reports
Earlier this year, we hypothesized that our customers would
benefit from having more guidance in the product. From
customer interviews, we learned that our UI didn’t give
customers clear enough information about what questions
they could answer and why those questions mattered. We also
learned that customers found our query builder difficult to use,
since it had so many options.
SUGGESTED REPORTS These numbers suggested a few options. Maybe the issue was
that Suggested Reports needed to be more discoverable. Maybe
we needed to do more work to show users why Suggested
What We Learned Reports would be a good option for them. Or perhaps users were
aware of what Suggested Reports could do, but simply found it
not useful.
After rolling out Suggested Reports, we gave it some time and
started measuring. Or, rather, we looked at the data. (Since Clearly we needed to dive deeper.
Heap autocaptures all customer data, we never needed to “start
measuring”; Heap was already measuring anything, without
needing“track(‘event’)” tags.)
Impact on Retention Did Suggested Reports improve query retention? Retention Retention Analysis
Start Event
Executive Query
Return Event
One way to start answering these questions about Suggested Executive Query
Reports—was the issue that Suggested Reports wasn’t Filters Add Filter
Date Range
Add Comparison
to measure the feature’s impact on retention. Our theory was First Time
were in fact finding the feature valuable. If so, our energy Has done Click Analyze (All) 56.54% 3.81%
Has not done Click Analyze (All) 41.47% 2.53%
First, we found that that users who had successfully run a Heap While this information doesn’t prove causation (only correlation),
query with Suggested Reports had a 50% higher weekly query as a first step it was quite encouraging.
retention rate at 10 weeks vs users who hadn’t.
Start Event
Return Event
First Time
Run Query
Final Learnings
Whenever we launch a new feature, we like to examine the
hypotheses we had at the beginning and see how they turned —— The bad: Of customers who use Heap and were exposed
out. We also like to look at the data and see what next steps to the Suggested Reports feature, only a very small
it suggests. percentage—10%—actually used the feature.
Hypothesis: Next Steps:
Our hypothesis was the Suggested Reports would make it easier Given that Suggested Reports has a positive correlation with
for users to answer questions in Heap, and as a result would retention, it seems worth continuing to build it out. Before doing
increase retention. that, however, we clearly need to make Suggested Reports
What we found: more discoverable. There are many ways we can do that: we can
present it more prominently in the product; we can add guides
—— The good: It does seem that users who actually used to the product that direct users to the feature; we can market the
Suggested Reports found value in it. Compared to users feature more aggressively; we can direct our Account Managers
who did not use Suggested Reports, users who did use to help let customers know what Suggested Reports is and what
Suggested Reports were more likely to run queries, use it can do for them.
Group By, and engage in the activation features that tend
to indicate retention in Heap. On to develop more experiments!