Metrics is in the core of making games. The idea for metrics is simple. Count how many times something happens. You can measure
In one of my games, Bounce & Bloom, I had a simple tutorial but the completion rate was 66%.
This totally surprised as I aimed to create a really simple tutorial, but no, still something was wrong.
So I had people playtest it. I did not find any single reason, but several things were too complex.
So I cut down everything not needed, simplified and made things totally obvious.
Now the completion rate spiked up to 99%. I was happy, I'm not losing those players anymore :)
Another sample is that when Wooga started using metrics, they used have this intro for one of their games. They did an A/B test. In an A/B test half the players got version A with intro, half got version B with no intro.
The results were the same. Conclusion?
Intro was cut out as there was no proof it made any difference except it was taking time and money to produce and maintain.
Key about metrics is to understand what to measure, but even more how to interpret the results of what has been measured.
You should try to measure everything people do.
You can define the paths you expect people to take when they monetisize or as they play the game. Then you can compare the metrics to the your expectations.
Let's say your release 3rd version of the game and all of a sudden player's completing the games first 15 minutes drops by 15%.
Why? You need to track all the changes you did. Maybe the metrics can tell you which of the changes has caused, or it can be a combination of the three or it can be also a result of changes in the previous version but the results are starting to show only now.
Interpreting the metrics can be straightforward but it can also be hard and complex. Metrics can help you find the cause, but they will not always give you a clear answer what is the cause for certain behaviour. But can they can be used to verify a certain behaviour.
This is something you should know. What your players do after the in-app. Where do they spend their coins.
You should really know and learn
An A/B test means that you have two versions of the game, version A is shown to other group, version B is shown to the other group. You can do even with a soft launch where you launch the game in only limited territory or country. In this case you could determine by the player's IP address whether to give A or B version.
Purpose is to find best working method.
Many companies use these regularly to test their games and they trust these more just in house opinions.
Some companies even say these are the absolute must in making a successful game.