Mark Robinson, the Co-Founder and COO of GamesAnalytics , was generous enough to write the first guest post on my blog, getting into the mechanics of determining Lifetime Value (LTV). This post does a great job of putting many of the ideas I have discussed in my LTV series into practice. Here are Mark’s thoughts on calculating LTV.
The games industry is quickly learning how to design engaging player experiences and make money from free to play (F2P) games. The transformation from console to online has placed analytics at the heart of game design and management. There are two types of analytics. Game Performance metrics let us interpret the health of our games. Player Behavioural metrics tell us what to do about it to make things better.
So in a simple example, if our Game Performance metrics show us our average sessions per player is 1.2 and our Day 1 retention rate is 25 percent, then we know we have a problem. Player Behavioural metrics will show us why players are leaving the game and not coming back. Typically there will be several reasons: the tutorial is too obscure, the game tries to monetize players too early, players are running out of resources etc etc.
The objective for analytics is to embed this iterative approach in monitoring game performance and use behavioural analysis to suggest solutions to the designers and developers to take forward and A/B test.
And so it is with Lifetime Value (LTV).
LTV is one of the most important Game Performance metrics as it fundamentally determines how successful the game is in terms of revenues. By comparing LTV to Cost per Install (CPI), it allows developers to understand if there acquisition budget is delivering the required revenues for the game to be profitable. And importantly it shows which acquisition channel is being effective so that future spend can be re-orientated.
Calculating CPI is trivial but there are a number of factors in calculating LTV which need to be considered to get the correct metric from which to make important decisions with confidence and accuracy.
There are three possible contributions to LTV
- Actual In-Game Revenue generated by the player from real money purchases or consumption of in-game advertising.The latter is an important component of game revenues so this data needs to be collected against individual players as well as the revenue events. This is a consideration when tagging the game for data collection so that all revenues regardless of source can be allocated to an individual player.
- Viral Revenues i.e., revenues from players that have been introduced into the game by the originator.It is common that a player that introduces paying players into the game is recognised for this with a contribution of those viral revenues applied to them. For example, if Player A introduces Players B, C, and D and they generate $0, $10, $50 respectively, then it is good practice to split this viral revenue giving 20 percent to the originator Player A, or $12 in this case. Then the introduced players are allocated 80% of their actual revenue, or $8 for Player B and $40 for Player C.Calculating viral revenues in this way can get complicated, especially if you start to consider second- and third-step introductions. For simplicity, I normally suggest first-step introductions are sufficient.(By the way note that the 80/20 split is undertaken so that the total revenue still adds up to the correct amount and there is no double counting.)
- Future Revenues – this is the most difficult aspect of LTV to calculate but in some ways can have the most influence on the overall figure.Estimation of future revenues is an area that is rarely explored and many organisations content themselves to have a handle on the first two aspects. As with all player analytics there are a number of important factors which will determine the extent of future revenues:
- Playing time – are they new players or mature?
- Payers of non-payers – have they paid already?
By setting up a grid of player segments using these two factors and calculating actual future revenues from a carefully selected cohort of players on this basis, it is then possible to overlay future revenues on all players. However, be aware that the calculation moves from individual player level to an average by future revenue segment. An example of the resulting matrix is shown below.
Levels 1-3 Levels 4-6 Levels 7-9 Levels…. Paying $1.21 $1.02 $0.88 $0.75 Non-paying $0.22 $0.17 $0.12 $0.08
And so the estimated future revenue is a average based on the position of individual players in the matrix.
Pulling all of this information together gives us the following way of calculating LTV which incorporates a rounded view of the contributions individual players have made to the success of the game and will make to the success of the game.
The math looks like this for player i:
LTV (i) = Total Revenue (i) + Total Viral Revenue from first degree introductions by i + Est Future Revenue (level by paying/non-paying)
And so we have a way, with a bit of work, of accommodating the three main components of LTV. It does need the ability to query at individual level and it also requires some careful calculation of cohorts to calculate future revenues, but who ever said analytics was easy?