A Short Guide to LTV
Mehdi Abbassi – Jun 29, 2022
Lifetime Value (LTV) is easily one of the most important key performance indicator (KPI) that a developer needs to consider. Data from LTV forecasting can be used to ascertain an estimate of how much revenue a game will generate. LTV forecasting also provides developers with crucial information for marketing strategies.
Despite how fundamental this metric is, many developers don’t have a good grasp of LTV forecasting. TapNation is dedicated to helping our clients achieve their business goals. So, we’ve put together this brief overview of how LTV forecasting works and why it is so important.
What is LTV forecasting?
LTV forecasting predicts how long a user will interact with a game and more specifically how much revenue they may generate in this time. This can be done by looking at the history of a typical user and extrapolating from this data.
A simple method of determining LTV is to ascertain how long user X spent playing game Y and how much money they spent during this time. This figure is then compared to the cost of acquiring a user, known as the cost per install, or CPI. If the LTV of a game is greater than the sum of its CPI, then it should be a profitable investment.
For example, the concept of a user’s in-game ‘lifetime’ can vary wildly. A user may spend as little as a few days or just a week playing a hyper-casual game, whereas a mid-core game could hold their engagement for over six months.
What’s more, the rate of user engagement during these periods may not be consistent. Many games see high levels of engagement and spending within the initial ‘honeymoon’ period. For a hyper-casual game, this peak could be anywhere from three days to three months. Other games may have a steady rate of monetary spend by players over a long period of time, even up to a year or more.
Player and game relationships can be complicated. The dedication a player shows to a game is determined by demographics, branding, genre and even the country the player resides in. All these types of inconsistencies must be considered when ascertaining the typical player lifetime.
These types of complexities underpin the fact that LTV forecasting is not a perfect science. It is a prediction of future behaviour based on past behaviour. LTV forecasting must factor in inconsistencies and account for future trends. The model that is used to determine LTV needs to be sophisticated and flexible.
A look at how to calculate LTV
It is not easy for a budding game development company to research LTV calculation models. Because LTV forecasting is such an important part of game development, many companies keep their exact LTV processes secret.
While we’re not in the business of giving away our trade secrets, below is a short guide to the process we use here at TapNation to ascertain LTV forecasts.
The methodology used to forecast LTV must incorporate a wide range of data. TapNation provides our clients with precise LTV forecasts by using a machine learning model that extrapolates from historical data. The old saying ‘garbage in, garbage out’ is a foundation of good LTV forecasting. Every piece of data that could impact LTV must be collected and fed into the algorithm.
LTV calculation can be impacted by:
- Previous historical LTV data
- The country of data origin
- The platforms used
- The media channels used
- Current and past trends
- The calculation period
Before any of this data can be integrated into the machine learning model, it must go through a variety of preprocessing steps. At TapNation, we break these steps into two components:
Data Cleaning – This is where our team carefully go through the accumulated data to manage any missing values and control outliers that may impact the final outcome. Some figures are replaced by averages, data is aggregated and events that may cause income spikes are considered.
Data Structuring – Next, we need to build the appropriate structure of our data bases before any input to the models. This step is very important to assure a good connection between the tables in our data base and to avoid any problem during the training of the model.
TapNation uses two classic machine learning models (random forest and gradient boosting) and an other model built from scratch with a grid search of several hyper-parameters before selecting the most accurate results. The three results are then aggregated under a meta-model. This model is tested against the historical LTV curve over time, which is then expanded to provide LTV predictions.
How LTV forecasting can benefit your company
Every business should have a good grasp of what its LTV is now and what it might be in the future. LTV predictions allow companies to invest in areas that will drive growth without too much risk.
Having a reliable LTV prediction allows a company to allocate a reasonable amount of funds to user acquisition (UA) campaigns. Without a good idea of what your LTV will be, you won’t know if you are potentially spending too much money on UA. Likewise, LTV is a crucial metric when deciding how much to spend on user retention.
LTV forecasts can also be used to identify any weak areas. LTV predictions can show that a game has a long user lifetime rate but low rates of spending. Developers can then focus on strategies to increase in-game purchases.
Taking a good look at a competitor’s LTV can identify what concepts are viable in the marketplace. Ongoing LTV predictions can provide useful metrics to measure growth and profitability. For startups, a solid LTV forecast can demonstrate a game’s feasibility to potential investors.
Reliable LTV forecasting is incredibly important for both fledgling developers and well-established game companies. Whether you already have a game on the market or you are currently in development, TapNation can deliver effective LTV forecasting solutions.
All in all, If you have any further questions about publishing with us and our data tools, check out the publishing page on our site here. You can also email us or get in touch via our social channels.