For many years, LTV has quietly yet consistently sat on the throne of the metrics used by advertisers all over the ecosystem to decide on how effective their advertising is and how to allocate budgets properly to maximize returns and keep growing. But then suddenly the unthinkable happened when Apple decided to roll out iOS 14.
Now many mobile app marketing agencies are saying that LTV modeling will see the same fate as user-level attribution. An inevitable death, which is technically possible due to the limitations of the SKAdNetwork framework. That makes it extremely challenging to get to measure LTV. While it doesn’t provide you with any timestamp of the time of the event.
Let’s recap how we can calculate LTV and why it is important
LTV is an important metric that shows how much money you made of a user during their lifetime engagement with your app or service before they churn. It’s tied to Customer Acquisition Cost (CAC) which is the metric that calculates how much you’re paying to acquire a new paying customer for a channel or a campaign and if your CAC is higher than your LTV, it’s impossible to make a profit, and you’re actually losing money already.
LTV helps you understand how much a new customer will add to your overall revenue, which makes it a key metric for every business out there because you can use it to make future predictions of your future growth then use it to set your goals and decisions on where to focus your resources and how to grow and reach those goals.
Now that we know how important it is, let’s get down to some calculations. The traditional method of calculating LTV is actually pretty simple, where you multiply the average value of a sale × number of repeat transactions and divide them by your churn rate.
Here’s an example. A customer spends $60 every year on their favorite app subscription for 5 years. The customer’s LTV would be: ($60 X 5 years ) / their churn rate. Pretty simple right? But marketing experts took usually split customers into “cohorts”; groups of customers who made their first purchases around the same time who share similar characteristics.
The three primary forms of LTV modeling
Historically, every marketer relied on three primary forms of LTV modeling to assess the performance of their channels and campaign.
- Behaviorally-driven user-level model: Using user-level data and used by all apps.
- Retention model: Using cohort and used by apps with high retention and less consistent monetization behavior.
- D7 ROAS Model: Using cohort level predictions and used by apps with consistent monetization behavior.
With the iOS14 changes, the behaviorally-driven user-level model became hard, the retention model became moderately easy, and the D7 ROAS Model became easy.
So how mobile games are doing cohort and calculating their LTV?
The SKAdNetwork comes with 64 bits that you can configure through your MMP, and those conversion values can be configured to reflect values for certain events or simply configured to have revenue buckets.
Gaming apps used conversion value schema based on revenue buckets from Day0 (or the first few days) and they look at how these buckets convert from Day0 to LTV from their overall cohort (without attribution, but for similar geographies) and apply this ratio.
For example, a US-iOS user who generated $0.50 to $1.00 on Day0, ended up generating $10 over his lifetime. They apply this ratio (x10-20) so that the revenue based on the SKAN bucket is better than nothing.
But, there are many assumptions with this method. Such as each channel variance is limited, the past ratio Day0 could be greater than actual LTV in the future, SKAN data was unreliable…
The new LTV model
In the past, marketers used to build a cohort model to forecast the future estimated revenue from a channel or a campaign because before Apple released App Tracking Transparency (ATT) installs from a campaign could all be grouped and assigned the same predicted LTV based on lower-funnel revenue events. This LTV model provided a sufficient level of accuracy for user acquisition and budget allocation purposes.
Now after iOS14, users are opting out of ad tracking. That stripped advertisers from the ability to understand which campaigns or ad networks drove installs, rendering that cohort LTV model useless. So, marketing teams started rapidly testing new ways to validate the long-term effectiveness of their marketing campaigns. Let’s summarize the most important ones.
D0 ROAS LTV Model
With the D0 ROAS LTV model, marketers use the SKAN’s conversion value report to estimate the D0 return. Then they use this to estimate the long-term return of a channel or a campaign. However, this model has several drawbacks such as that the D0 ROAS is an inaccurate measure to predict the LTV or that marketers will be stuck at the 1-day view of the performance campaign.
The user-level LTV model
This model sounds totally against what Apple is enforcing with the ATT framework in the first place. However, marketers still have first-party data that they could use to predict how much a user is likely to spend. This model is a machine learning algorithm that will use data such as the analytics identifies that help in predicting user’s behavior, the revenue data from IAP or ad monetization, and the in-app engagement to predict how much a user will spend in the future.
In other words, such a user-level LTV model allows marketers to tackle the challenges raised since the rollout of the ATT framework. This means being able to measure the performance of the SKAN campaigns in the same way they measure any MMP attributed campaigns. However, this model needs many more resources than you may think, and those resources may not be available for all businesses. A startup or a medium-sized business may not find the value in investing and building such a model compared to a business spending millions advertising its products and services online.
Is LTV still relevant? Conclusion
ATT changed the way your business advertises online. And Apple isn’t the only company to roll out new frameworks to shape a privacy-first advertising world. More companies will follow. The best strategy to approach this new normal is by adapting and being flexible. Your marketing team should follow a test and learn approach to finding what works for your business. And how to still be efficient in this new normal.
However, if you have the resources, you can try building the user-level LTV model or the D0 ROAS LTV model. Then see how this will affect your mobile app marketing efficiency. Also, keep in mind that there are many more strategies that work for some businesses but not for everyone else, such as the Marketing Mix Modeling “MMM” which is a really interesting topic for another post.