Is Your Ad LTV Model Accurate?

Ad LTV (Lifetime Value) refers to the advertising revenue made by new users who install the app over the lifetime of their activity. It takes into account user retention, ad impressions, and estimated revenue from all networks for a specific cohort of users. These are the users who first used an app on a given day, from specific countries or networks and on the basis of a select period in recent history. Many publishers do their best to create models for future LTV prediction after the app’s launch so that they could optimize their marketing strategies.

Accurate LTV Is Difficult to Come By

However, this is easier said than done and there are a lot of ways to calculate LTV with more or less accuracy. Simple solutions can result in big errors, with inaccuracies that can reach up to 400% on a user level and 50% on a cohort level.

We know that any LTV calculation has two main factors – retention and revenue, and publishers can get these two wrong when they build in-house solutions for ad LTV tracing. Those models usually assume that each impression pays the same level of CPM, which is a huge misconception, the one that can lead to dramatic errors and hugely inaccurate ROI calculations.

Here are six holistic LTV models we use at SOOMLA explained:


1. Naive Method

This most basic LTV calculation method takes the ARPDAU (average revenue per daily active user) and multiplies it by the expected lifetime. Although this is the simplest method, it assumes all users stay the same period of time. It is, in fact, impossible to know in advance how long the users will stay, which is why this method is also the weakest.

2. Modeling Lifetime Function

This method models out the retention function on the basis of limited data points – the 2nd day, 7th day, 14th day, and 30th-day retention. Then it multiplies the expected lifetime by ARPDAU in order to get the LTV. This model entails a bit more accuracy than the naive model, although it still assumes a constant ARPDAU and overweights the day-30 retention

3. Estimating Segment LTV by Other Segment

Assuming you have data from organic traffic or other segments, you can calculate the 180-day LTV in a specific segment on the basis of the proportion of revenue generated in the first seven days from the revenue in 180 days. Although this method is relatively simple and accurate, it requires 180 days of data from existing segments, which is sometimes not possible to get.

4. Segment LTV by Partial Data From Other Segment

This more advanced method uses data from existing organic traffic or other segments to create the model for the first 90 days and then uses the modeled function to predict lifetime from day-90 to day-180. Although this is a more accurate model and allows LTV calculation for newer apps, it is a bit difficult to follow.

5. Modeling Retention Function With a Spreadsheet

Utilizing tools like logarithmic functions, integral calculations, and statistical regression, this model is more flexible and can be applied over and over with a varying number of retention data points from different apps, cohorts, and traffic sources, it can be overly complex. It entails calculation of every segment separately without using data from previous segments and assumes a constant ARPDAU.

6. Modeling Revenue Function With a Spreadsheet

This method models out the revenue function by making use of similar statistical methods like regression and log functions. The positive side of this model is that it’s flexible and allows you to create a full model that takes any number of data points and accounts for both revenue and retention. However, it is very complicated as it calculates every segment on its own and doesn’t use data from previous segments.

Although the main purpose of these generic models is to provide some assistance in basic modeling if you are on a tight budget, some of them do sound very complicated. If you feel like this is too much for you, then you should let a dedicated solution like SOOMLA do all the legwork for you.

The Essentials of A Good LTV Solution

Our models calculate user ad revenue in proportion to the measured impressions, clicks, or installs, in what is essentially a curve-fitting or regression problem. Although our core product conducts attribution of ad revenue to the user level, the attribution itself depends upon an accurate recording of the activity in the app. Our methods of calculating ad LTV include six essential points:


1. Super Granular Data Collection

Our platform covers all the little details like tracking advertised apps, accidental clicks, x/y location of clicks, installs on both Android and iOS, as well as bid levels for CPI campaigns. Additionally, we are the only platform that can help you identify those 5-10% of users responsible for the majority of your ad revenue we call ad whales.

Visualization of ad whales’ role in ad revenue | Soomla

2. 100% Focus on Ad LTV

We devote an incredible amount of attention to measuring in-app advertising and activities, calculating LTV on the level of an individual user, cohort, and traffic source and constantly updating our technology to keep up with the industry trends.

3. Engaging ad networks in the solution

Covering all the main networks, our solution works regardless of mediation selection across all big mediation vendors, seamlessly listening to their SDKs. Our platform allows mobile app publishers who work with multiple ad-networks to learn the advertising revenue per user, cohort and traffic source. It integrates with nearly any media source, ad network, CRM, and marketing and attribution platform.

4. Superior modeling

Our own estimation system is more advanced than simply taking the average eCPM of the line item. We use mature ad LTV algorithms based on multiple models.

5. Unbiased third party

Attribution companies usually represent the advertisers’ interest first and fail to follow the trends of the monetization part of the combination. We are the only company that has your best interests in mind and can provide you with the expertise and benchmarking data you need.

6. Ongoing truth testing

While most platforms tend to use averages as the go-to method for calculating ad LTV from various segments, this is not the most precise strategy as it can lead to faulty decisions and negatively impact your overall success. Soomla evaluates multiple ad LTV models for each ad network and campaign type and assigns the most accurate model by constantly comparing truth sets. We have developed accurate testing methods that allow us to quickly iterate, test, and compare various models, including our own.

LTV is influenced by numerous factors and it may be tricky to keep up with all of them. So if you want to learn the secrets of how our ad LTV models work and how they may benefit you, reach out to us and we’ll show you why SOOMLA is the most accurate ad LTV platform in the industry.

Feel free to share:


Please enter your comment!
Please enter your name here