Data - particularly data as it defines the behaviour of players of F2P mobile games - is now a big and complex business.
Yet as Dmitri Williams, CEO of Ninja Metrics, delved into in his GDC Europe 2015 talk entitled Advanced Game Analytics, the sector is evolving fast, particularly as more cash is spent, which is forcing a more mature scientific basis.
Of course, as a starting point, Williams says that marketeers must refer their spending and their revenue back to acquisition source; something which is reinforcing the importance of good attribution data.
Yet, around 40 percent of developers are still trying to do this sort of thing via manual processes; something which is inaccurate and unscalable.
Instead, machine learning systems i.e. automatic systems which look for patterns in data - and off-the-shelf analytics companies are now making such analysis simpler.
One dilemma for developers as they move to automatic systems is that they are black boxes, which no one understands in terms of how they generate their results.
"You don't get full understanding of the data *and* confidence in the results," Williams warns.
Machine learning
In terms of how the underlying science is changing, Williams says the sector's ability to predict what people will do - i.e churn from the game - is growing rapidly.
Machine learning is vital to set up the many rules you need to apply to your data. This is making calculations much more complex but also much more accurate.
One dilemma for developers as they move to automatic systems is that they are black boxes, which no one understands.
Another reason accuracy is improving is the impact of social networks, which provides marketeers more accuracy in terms of how an individual will behave, whether that's voting for a political party or spending money in a game or inviting a friend to join an alliance.
In addition, how players use social networks - i.e. their virality - is now being highlighted and reflected into the monetary value of those players for advertisers.
Show me the money
All this accuracy is important because for game developers, the most vital metric they should be tracking is the LifeTime Value of each player.
Williams explained LTV as being the combination of LifeTime model (i.e. when a player leaves or churns from your game) with the Value model of your players.
There are different models of how to predict churn, aka survivability models like Cox/Hazard, while work on improving Value models now includes psychological and environment elements behind spending.
Indeed, in conclusion, Williams pointed to the growing value that's associated with the social behaviour of players.
"If you're fat, you're more likely to have fat friends. We know that getting fat is a social factor. As is smoking, or losing weight or stopping smoking," Williams said.
"People are compelling and tracking their sociability in games is vital."
Indeed, when doing research for the CIA on identifying terrorists, Williams' research team improved the accuracy of its results from 50-60% to 80-90% when they added social data to its analysis.
