Predicted lifetime value (pLTV)

To better understand how pLTV serves your measurement and performance goals, we first need to nail down what LTV means, the immense added value of predictive analytics, and the mighty potential of their power combo — pLTV. 

Lifetime value — aka LTV — is an estimate of the average revenue a customer will generate over the time that they use your app or service. But, given the recent data privacy revolution, how does one measure LTV without the same level of access to granular and especially long-term performance data?

This is where predictive analytics or modeling comes into the picture. It leverages machine learning and artificial intelligence (AI) to examine historical campaign data, past user behavior data, and additional transactional data — in order to predict future actions. 

By creating different behavioral characteristic clusters, your audience can then be segmented not by their actual identity, but by their interaction with your user funnel in its earliest stages, which can indicate their future potential to drive meaningful value to your business.

Why is it important?

Armed with knowledge, predictive modeling enables you to make rapid campaign optimization decisions without missing a heartbeat. Nip unsuccessful campaigns in the bud, or quickly double down on investment that can drive even better results — without compromising your users’ privacy. 

In other words, pLTV allows you to leverage data science horsepower, and predict how much money your customers will spend in your app over a predefined window of time based on their past behavior. 

It also enables you to segment users by their acquisition source and forecast projections accordingly, making it an ideal tool for determining which of your marketing channels will produce your highest spending users now — and in the future. 

Especially during the acquisition and re-engagement stages, understanding user behavior patterns and the typical milestones that separate high-potential users from low potential — can be incredibly valuable. 

Creating a pLTV model

While the average marketer will only measure a maximum of around 25 metrics, an app could have well over 200 available for measurement. Take a machine, on the other hand, and it will be able to ingest all this data in a matter of milliseconds and process it into actionable marketing insights. 

Calculate all these indicators based on your definition of success and LTV logic, a machine learning algorithm can apply all that to a significant amount of data, while finding correlation between early engagement signals and eventual success.

This means that advertisers no longer need to know who the user is, but rather which pLTV profile and characteristics they fit into. 

Two key guidelines to keep top of mind in this context:

• Your pLTV profile should be as accurate as possible, and made available during the campaign’s earliest days.
• It should represent your LTV requirements for it to be considered valid and actionable, and allow the algorithm to cluster general audiences into highly granular, mutually exclusive cohorts.

Because there are obvious differences in the way various apps retain and monetize their users, a universal cookie-cutter LTV model is not a real thing. This makes building an LTV model to predict mission-critical metrics like Return on Ad Spend (ROAS) somewhat challenging — yet still doable.

To learn more about how to set up predictive models for LTV calculations — visit our guide.

5 tips for building and maintaining LTV prediction models

1 – Success is a constant feeding loop

When building data models that are designed to guide significant decisions, it’s not only important to build the best system possible — but also to constantly test and tweak it to ensure effectiveness and accuracy.

For both purposes, make sure that you continuously feed your pLTV prediction model to keep it trained on the most relevant data, and always check whether your model’s predictions come to fruition based on new or near-new observations. 

Not following these steps could mean that a model with an initial useful prediction power could go off rails because of seasonality, macro auction dynamics, your app’s monetization trends or other factors. 

By observing your leading indicators or early benchmarks and looking for significant changes in data points, you can gauge when your own predictions are likely to break down. 

2 – Choose the right KPIs

There are several options to choose from, each with a set of trade-offs in viability, accuracy, and speed to produce recommendations. Go ahead and test different KPIs (e.g. more or fewer days of ROAS / LTV). You might be surprised at how poorly correlated the standard measures prove to be.

3 – It’s all about segmentation

Segmenting users into groups is a proven method to reduce noise and improve the predictive power of your pLTV model. 

Additionally, by creating different behavioral characteristic clusters, your audience can then be categorized not by their actual identity, but by their interaction with your campaign in its earliest stages. This interaction can indicate their future potential with your app.

For example, a gaming app developer can predict their 30-day LTV based on a tutorial completion (engagement), number of returns to the app (retention), or the level of exposure to ads across each session (monetization). 

4 – Timing is everything

Acquisition cost trends during the first week after a new app launch will be very different during the fifth month, or the second year for that matter.

So, although the influence of seasonality on breaking down predictions is a given, the lifecycle of your app / campaign / audience / creative could also influence the ability of your model to make accurate predictions. 

5 – Define team responsibilities

Whether opting for an in-house, strategic-minded analyst that can lead decisions around which model should be used and how, or a more junior analyst that owns day-to-day pLTV calculations, or outsourcing to a 3rd party — allocating pLTV ownership is completely up to your business needs and budgetary constraints. 

If you’re looking to hit the ground running and don’t sport tons of subject matter expertise, outsourcing pLTV analysis could very much jumpstart the process. In the long run, however, after viability and ROI have already been established, moving this in-house could enable more in-depth analysis and scale.

Key takeaways

• Leveraging data science horsepower, pLTV allows you to predict how much money your customers will spend in your app over a predefined window of time based on their past behavior. 

• Harnessing pLTV you’ll be able to segment users by their acquisition source and forecast projections accordingly, making it an ideal tool for determining which of your marketing channels will produce your highest spending users now — and in the future. 

• To ensure accuracy, be sure to continuously feed your pLTV prediction model to keep it trained on the most relevant data, and always check whether your model’s predictions come to fruition based on new or near-new observations.

• Also, be sure to choose the right KPIs, segment your audience to reduce noise and improve predictive power, factor in timing and seasonality, and lastly – predefine team responsibility based on your business needs and budgetary constraints.