Sponsored By

Retaining players is perhaps the most important goal for a game developer. The problem is that it’s also the most challenging… So what can you do to get retention right? One of the best strategies to adopt is running a cohort analysis.

Ran Avrahamy, Blogger

June 15, 2015

10 Min Read


The million dollar question game developers keep on asking themselves day in and day out is how to get players to keep coming back for more. Not only do you have to compete with the 42 apps on average on a user’s phone (with most opening no more than a handful on a regular basis), there are also 372,169 active games in the app store alone (and counting...) just waiting for their chance to take your spot.  


Retaining players is perhaps the most important goal for a game developer, especially in a space that’s dominated by F2P games. The problem is that it’s also the most challenging… So what can you do to get retention right? One of the best strategies to adopt is running a cohort analysis.


Let’s take a step back to explore this strategy. A cohort report enables you to group users with common characteristics and measure specific KPIs over different timeframes. For example, one cohort can be users who first launched an app any time during the month of January while another cohort are those who launched it during the first week of February and live in the US.


This form of grouping enables an “apples to apples” comparison and therefore a really good indication of change over time. It tells us about the quality of the average customer and whether it’s increasing or decreasing over time. Unlike standard retention, the metric is calculated per different timeframes, which represent the first X activity days per user, and then accumulated among all users (that’s why the graph never drops).


Case Study - cohort analysis across the funnel

A F2P gaming developer with high install volume but lower than average lifetime value was looking to improve his retention and ultimately, the volume of in-app purchases. By diving deep into his user base and analyzing his entire funnel, he pinpointed the best performing ad networks to work with and ended up improving retention and as a result increasing the average number of in-app purchases per user by nearly 15%. How did he do it?


The gaming marketer divided his funnel into four stages - each of which was mandatory: tutorial completion, Facebook connect, level 10 completion, and in-app purchase. He then configured AppsFlyer’s dashboard to track these events and defined each cohort based on the media network that acquired users who launched the app during the first week of April.

Tutorial completion

image: AppsFlyer

Key takeaways:

  • When it comes to tutorial completions, network A reigns supreme starting off with a bang at about 24% of users, climbing to 27% and then 31% after which the line plateaus (remember, the graph never drops as it’s accumulated among all users)

  • Network E ran with incentivized traffic. Since it offers rewards for tutorial completion, we see that it started off high but then showed marginal growth and began to flatten on day 5

  • Networks B and C performed fairly well with modest growth until day 7

  • Network D performed poorly across the board starting off at only 7% and showing very little growth going forward

Optimization performed:

  • Networks B and C were broken down to campaign and ad level and those that underperformed were removed

  • Network D was dropped on day 5 as there was no desire to continue bringing in new users who weren’t moving down the funnel

  • The budget of network A was increased to maximize the pool of users entering the engagement funnel so that more reach the finish line, and hopefully become high spenders (players who contribute most of the developer's revenue)

Facebook Connect

image: AppsFlyer

Key takeaways:

  • Network A continued to perform well in this stage of the funnel, picking up speed on day 3 and only plateauing on day 14

  • Network B optimization worked well as a growing number of users on average moved down the funnel

  • Network C optimization worked fairly well but still underperformed compared to B

  • Network E’s invenvitized traffic reached a plateau very early on, clearly demonstrating that’s as deep in the funnel as its users will go

Optimization performed:

  • Network E was dropped after day 5  

  • Network C targeting was changed to better reflect network B’s targeting

Level 10 completion

image: AppsFlyer

Key takeaways:

  • Network A, B, and C performed well delivering users that moved down the funnel to complete level 10 within about 14 days.

In-app purchase

image: AppsFlyer

Key takeaways:

  • Networks A and B delivered great value with a growing number of in-app purchases from day 5 and climbing after day 7

  • Network C generated some value but plateaued after day 14

Optimization performed:

  • Budget of networks A and B increased

  • Retargeting campaign launched with network C starting day 14 in attempt to tackle the slowed growth

The Bottom Line

The ability to use a cohort report to understand how each network performs in each stage of the funnel proved extremely valuable to the game developer. By optimizing during every step of the way, he managed to retain more users and bring them to cross the finish line. Ultimately, these users increased the number of in-app purchases by 15%.  

Read more about:


About the Author(s)

Daily news, dev blogs, and stories from Game Developer straight to your inbox

You May Also Like