Retention is every app publisher’s game to lose

An emerging breed of data-savvy marketers are bridging new tech and proven statistics to level up their app retention strategy.

No matter the app size, category, or business model, retaining app users is a big problem -- and an opportunity.

The Apple App and Google Play Stores each offer over 1.5 million apps vying for the same audiences. These stores are buyer’s markets filled with products that often have high substitutability. Just search for “photo editor” or “subway maps” and brace yourself for literally hundreds of options. The market for app downloads is additionally characterized by cut-rate switching costs -- with the table-stakes-ization of free-to-download and ever-decreasing download times among the most prominent drivers of shrinking barriers to install.

If you are an app publisher today, it’s likely that a user can find and download a substitute app that seemingly provides the same value as your app - in less than a minute and or free. No wonder retention rates are abysmal.

Yet, quickly leading users to their ‘a-ha’ moment and encouraging them to ‘snackify’ their experience across shorter, more frequent intervals will become critical for the largest publishers earning revenue through ads, in-app purchases, or both. And even for the long-tail of smaller app publishers, high engagement likely has a positive effect on discoverability in the app stores.

 

Enter Predictive Churn Management

Now here’s the good news: retention, generally, is not a new problem. For decades, the minds and budgets of data-savvy marketers from big traditional industries from telecom to retail have invested in sophisticated predictive modeling techniques to predict subscriber and shopper churn risk and preemptively engage at-risk customers.

Three key trends are now surfacing opportunities for app publishers to unlock this same control and predictability over user behavior:

  1. First, the mass proliferation of connected devices is generating waves of new, rich user behavioral data that contains the information needed to build precise predictive models;
  2. Second, recent advancements in computing have made quickly running the complex analyses necessary for behavioral prediction on billions of data points cheaper and more accessible;
  3. Finally, the tried and tested legacy methods for predicting behavior like customer churn and lifetime value are being repurposed and adapted to specifically fit app user data.

The result will be a supercharged approach to how retention is treated across the user lifecycle, based on a forecast of how likely each user is to churn. Here’s how it will work:

  1. Instead of reactively nudging already churned users to re-engage, data-savvy marketers will now use predictive behavioral analytics to intervene with users just before they churn in a meaningful and targeted way, and put them back on a path to loyalty.
  2. The same predictions will also help marketers identify their most loyal users early on in the lifecycle and proactively engage them to take actions that drive user growth, such as inviting friends to the app, leaving ratings in the app stores, or completing an NPS survey. 
  3. For users that do churn, a predictive approach will give marketers the information they need to optimize spend by retargeting churned users with the greatest winback potential, rather than indiscriminately spending the same amount to re-acquire users regardless of their expected residual value.

 

Ongoing, Evergreen Journeys

This predictive approach to user retention will eventually scale with the support of marketing automation towards ongoing, evergreen journeys humming quietly in the background of more pointed, one-off blast campaigns. To help illustrate what this future approach will look like, let’s consider the example of a fictional sports media app: SportsMob.

As soon as SportsMob’s active users reach a predefined threshold of churn risk, they will be funneled into a journey of chained touch-points designed to optimize retention per dollar spent:

  • Newly identified at-risk users will first be exposed to a push campaign promoting a feature, the use of which is correlated to retention. Let’s say it’s choosing their favorite team: “Stay on top of your team’s news - set your favorite team now.”
  • Some of these newly at-risk users will be won back through this first touchpoint.  A few days later those that remain at-risk will next be exposed to an incrementally more costly email campaign, offering a 10% discount on a premium subscription to live streaming of football games.
  • Users that remain at risk following these first two touchpoints will trickle into a third Facebook remarketing campaign that is incrementally more costly than the previous, advertising a one-month free trial of the premium service.
  • The messaging and creative content of each link in these chained multi-touch journeys will be A/B tested and, leveraging techniques in machine learning, automatically optimized towards the goal of retention.

By staying one step ahead of users’ next moves, this predictive approach will ultimately give marketers the ability to measure and control the long-term value of their user base.

 

Taking Action

So what can you do today to get started with a predictive approach to churn management?

The first step is making critical inquiries into the nature of your product, and value of your users:

  • How should we define churn for our app? Is it the absence of any behavior whatsoever? Or, is it it tied to the absence of specific behavior, such as viewing content or completing a purchase?
  • Is there an amount of time that must elapse in order for churn to occur? If so, how long? 7 days? 30 days? Longer?
  • What is the value of a retained user? An additional app session? Or an additional piece of content viewed or shared?

At Localytics, our data science and predictive analytics team is dedicated to helping our customers answer these questions and putting into place the appropriate methods for forecasting and controlling churn at scale. It’s just our first step in leveraging predictive modeling for deeper app marketing automation.

 

I want to learn more about Predictions

 

 

X