We run a games platform with millions of users (+- 150,000,000 gameplays / month).
We want to find tools or set up a data stack to:
- collect basic metrics for a specific game such as average gameplay time, 1 day return rate, 7 day return rate,...
- be able to segment these data by any dimension that we pass along (e.g. by country, by network speed, by ...)
- generate more advanced insights for a specific game, e.g. this is the distribution of gameplay time and what you can see is that many people quit between minute 2 and 3 (corresponding to level 2).
In the long term, we want to use this data to:
- generate some kind of 'potential' metric for games
- help developers to improve their games (both in terms of technical issues, but also in terms of gameplay issues)
We are already using AWS for some things but are open to other stacks as well.
For AWS, what we guess is that we would need:
- Amazon Firehose for collection the raw events
- Amazon Glue for data processing (e.g. partitioning)
- Amazon Lake Formation (not sure what it's for)
- Amazon Athena for a query layer
It feels a bit excessive.
What tools or stack would you suggest for collecting (millions of) raw events and being able to generate metrics and insights based on them?