search

ReflectionIO / storedata

Reflection IO Store data application
0 stars 0 forks source link

The revenue sometimes display as zero on the graph page even though the database has a value for that date #469

Closed diireh closed 9 years ago

diireh commented 9 years ago

The filters used are: 1st Jan 2015 / UK / iPhone / Games

The 1st Jan has two codes: 217 (morning - used by the graph page) 218 (evening - used by the leaderboard page)

The 217 revenue figure according to our database is: 39425592 The 218 revenue figure according to our database is: 16277495

As you can see from this screenshot the downloads on the 1st should be 39425592 but is instead 0.

screen shot 2015-02-06 at 11 16 31

Note: The queries below are slow if I add itemid="529479190" so instead I paste that into the filter rows box after doing the query to just see the info for Clash of Clans

SELECT * FROM rio.rank where deleted = "n" and code2=217 and categoryid=15 and country="gb" and type in ("topfreeapplications", "toppaidapplications", "topgrossingapplications")

code = 217 topfreeapplications revenue 0 - downloads 11124 topgrossingapplications revenue 39425592 - downloads 0

SELECT * FROM rio.rank where deleted = "n" and code2=218 and categoryid=15 and country="gb" and type in ("topfreeapplications", "toppaidapplications", "topgrossingapplications")

code = 218 topfreeapplications revenue 0 - downloads 15352 topgrossingapplications revenue 16277495 - downloads 0

diireh commented 9 years ago

On this date the download figure was correct though. i.e. 11124

screen shot 2015-02-06 at 11 21 26

diireh commented 9 years ago

The problem here is that the windowed data is incomplete, however there are numerous reasons as to why this would be the case. Database failures etc... The best solution for this is to write a tool that allows us to enter the filter criteria, which it would use to generate the windowed data. That way we could "fix" any issues.

Also, going forward we'll probably have a new windowing strategy using DataFlow's build in methods or creating our own bespoke with the Pipeline API.