This was a topic asked through our ticketing system, so i thought i'd share it in brief here.
The question was on how to get the extra data attached to the analytics event keys.
So that additional data is not directly accessible through NoSQL.
You will need to build a custom chart to view it.
You can see that charts tab to the right of 'Screens' and 'Snippets' tab...
Then click on the 'filter' button and you should see your data options available...
You can customize your chart and hit 'Test' so see the data for that key...
In my case ive only got one element with a count of 1, so i'll do more tests so i can show you more data but this should get you going.
Whilst it is possible to filter by the custom data using these charts, is there a way to group by it instead ?
e.g. if I have custom data that shows a level number the player has chosen, how can I generate a chart that breaks it down by each value in that custom data field (e.g. how many players in level 1 versus 2 versus 3) ?
The "group by" drop down only contains @type
As far as I can see I can either have a separate key per level instead when I log the analytics to begin with, or generate multiple charts with multiple filters and combine it together outside. I'm hoping there's a neater way?
Thanks Sean - however adding multiple rules or groups still seems to amalgamate everything into one chart - it doesn't give any extra options for "group by" ? No matter what I click on, the only grouping option seems to be by @type
That would be really useful - do you have any update on that tutorial ?
We asked a while about using the custom data with a lot of nested structures sent in custom analytics events (we were basically locally storing a lot of events then sending them as custom data at the end of a level in our game) and were told that that use case isn't supported yet, so we split up our data into numerous keys with the level's name as a prefix and we don't send custom data anymore.
Do we have to build a chart for every single one of these keys to be able to see statistics for these events?
I had considered that approach also, but the number of combinations of levels and bits of data would be pretty large for us - and hence building charts selecting from a massive list of keys would be a bit of a nightmare.