Open jklukas opened 3 years ago
I'm not aware of whether there's a defined methodology for determining tablet vs. phone from baseline pings, so we would need direction on that.
It would require some coordinated backend and frontend work to add this new dimension, so we may not be able to address this request in the near term.
If we can determine the difference, I would humbly maybe suggest that we break these out as different usage criteria instead.
@athomasmoz Are you aware of any existing methodology for extracting device type from our mobile telemetry?
cc @magorlick on the question of methodology for device type as well.
This query was made for iOS a little while ago that separates iPhone, iPad, and iPod https://sql.telemetry.mozilla.org/queries/69519/source#175287
That indeed looks pretty straight-forward:
CASE
WHEN device_model LIKE '%iPad%' THEN 'iPad'
WHEN device_model LIKE '%iPhone%' THEN 'iPhone'
WHEN device_model LIKE '%iPod%' THEN 'iPod'
ELSE 'other' END AS device_model
But this also brings up the question of longevity. The linked redash query references legacy core
pings, but it's my understanding that Firefox for iOS has recently integrated Glean telemetry and that we will want to use baseline
pings going forward.
If we're going to put some effort into defining usage criteria for iPhone vs. iPad usage, we should not be applying that logic on top of legacy telemetry.
We have not yet tackled the problem of apps transitioning from core pings to Glean. I just filed https://github.com/mozilla/gud/issues/117 to track that, though.
For core pings the relevant field is device
and for Glean pings the relevant field is client_info.device_model
. We already capture those fields in the clients_daily-type aggregates, so we have the information necessary to define a new usage criterion that filters on device.
Just a note that this is still on my mind. There's going to be a lot of effort in H1 to deploy Looker and move use cases there. Within the next month, I'd like to prepare some updates to the mobile clients_daily ETL stream to support a GUD-like dashboard inside of Looker. I will definitely include device_type
as a dimension there, and we can evaluate next steps from there.
AshleyT asked for the following on Slack: