When we think of user targeting in advertising, we usually think of behavioural targeting (based on a consumer’s actions online), demographic targeting (things like age and gender), or retargeting (where a marketer targets a user who has already shown interest in their product).
But in the mobile app market, many marketers go a step further, and adjust ad campaigns based on consumers’ lifetime value – that is, the full value an individual will have to a brand over the entire course of their engagement with that brand.
In marketing, measuring lifetime value is used both as a metric for success and to optimise campaigns. And forecasting lifetime value for new users can be a crucial tool for ad targeting and tailoring.
For many marketers, customer lifetime value, or CLTV, isn’t particularly relevant.
When a jewellery company runs a campaign for a new watch, each conversion is essentially equal. A customer either buys a watch or they don’t, and the value of each sale is the same (barring any discounts applied, for example as part of an affiliate marketing campaign).
But this isn’t the case in the mobile world. For a mobile app which makes money through ads and in-game purchases, the value of each customer might vary wildly. Some might download the app once, play for twenty minutes, and see a couple of ads. Others might play for months and spend hundreds on in-app purchases.
So all conversions aren’t equal, and that means measuring lifetime value is an important aspect of measurement.
And if an app maker can forecast lifetime value, this can be used to steer ad strategy, in order to target those high value users.
The Technical Detail
You’ll see slightly different equations for calculating customer lifetime value.
Most will look something like this:
CLTV = Average revenue per user (ARPU) x 1/churn
Where churn measures the average lifespan of a customer.
Some calculations will also measure the extent to which users generate income by bringing in more users, for example by using an in-app invite mechanism to invite friends. So then the calculation might be:
CLTV = ARPU x 1/Churn + referral value
To calculate this, an app needs to be able to measure all of these factors – which is usually fairly straightforward. Average revenue per user is simply total revenues divided by total user count. Churn is measured by calculating the number of users lost during a given period, divided by the number of users at the start of that period. Referral value is more complicated – apps can see when users invite friends through in-app mechanisms, but this won’t capture all referrals. And marketers also have to decide how much weight to give these referrals. But it should be possible to get a rough estimate.
With all these factors, CLTV can be used to measure and inform ad campaigns.
At its most basic, CLTV can be used for apps to judge how much they’re willing to spend per acquisition. If the average cost per acquisition exceeds CLTV, an app will lose money. So CLTV can show the maximum an app should be happy to spend per user.
And CLTV can be tracked over time to measure campaign effectiveness. If an ad campaign manages to attract higher value customers, CLTV will increase. So measuring this data can be used to optimise creative and targeting parameters.
The story gets somewhat more complicated when it comes to CLTV forecasting.
CLTV forecasting means trying to predict the lifetime value of a user, based on available data you have about that user.
There’s not one specific way of doing this, and explaining in detail how it’s done would require too many mathematical calculations for a trade press article.
But on a general level, CLTV forecasting tends to involve breaking audiences into different segments or cohorts of similar users, based on data points available, and looking at past behaviour of individuals inside those cohorts to predict LTV for new users.
These cohorts might include information on demographics, device type, or really any data which the marketer has access to which might be relevant.
And modelling might be based heavily on user behaviour with the app. For example, a model might say ‘if a user logs on to the app six times within a week of downloading and spends between $10-12 on in-app purchases, their total lifetime value is expected to be $100”.
Once created, these models can be used for campaign targeting, for example by targeting ads at lookalike audiences based on which audience segments have the highest lifetime value, and bidding more for those audiences.
And they can also be used for retargeting existing users – by knowing which users are likely to come back and spend more once they’re reminded to jump back into an app.
Pros and Cons
The benefits of CLTV measurement and forecasting are that when done effectively, they can be a very effective tool for maximising monetisation.
Measuring average CLTV against cost per acquisition is important for making sure an app is actually sustainable in the first place.
And using CLTV forecasting to attract high value customers can be vital in the highly competitive app marketplace. The majority of apps don’t turn a profit, according to Gartner. And to succeed, identifying the highest value consumers is vital. One study from app testing firm Swrve found that 0.15 percent of mobile gamers contributed 50 percent of total revenues.
But CLTV forecasting is very difficult, and requires expertise in statistical modelling. There’s no ‘one size fits all’ solution which developers can plug into. Apps need to measure their own variables, and figure out a specific model for themselves.
Doing so can be expensive and time consuming. And these models need to be constantly updated, in order to adapt to changing consumer behaviour. A new update for a gaming app for example could make it more or less attractive to players, which in turn will affect CLTV.
And Apple’s upcoming restrictions to its IDFA identifier will make it harder for apps to collect data and find audiences across the app ecosystem, making it harder to forecast CLTV and to apply those models to marketing strategy.