1270225 - FX_MIGRATION_ENTRY_POINT buckets for first-run and refresh-firefox seem much lower than expected

Status: RESOLVED WORKSFORME

(Firefox :: Migration, defect)

Description

[Matthew N. [:MattN]]

http://mzl.la/1UzwZsx

Where the migration wizard was entered from (Release 45):
Other/catch-all		0 (0%)
first-run		67 (0.27%)
refresh-firefox		6 (0.02%)
Places window		5.75k (23.46%)
Password manager		18.69k (76.24%)

I would have expected first-run to be highest and refresh-firefox to be second but those are near zero. I have a hard time believing that we only had 67 sessions where the user was running Firefox with no existing profiles (like a new computer).

The probe is marked opt-out so it can count first-run/refresh users who didn't have a chance to opt-in for the profile yet but maybe there's an issue with recording this data so early in startup?

Comment 1

[Matthew N. [:MattN]]

My other guess is that the dashboard doesn't include payloads from users who didn't opt-in to telemetry.

Comment 2

[Benjamin Smedberg]

The t.m.o dashboards only show the opt-in population, at the moment. I don't know if there are short-term plans to change that, but you can of course run your own queries; looking at the 1% longitudinal sample might be the fastest/most efficient way to ask your question.

Comment 3

[Matthew N. [:MattN]]

OK, that's unfortunate and probably should be more clear on the dashboard as I suspect people don't realize that

The longitudinal SQL table has sane data (not version-specific):
0	41
1	976,776
2	152,956
3	16,569
4	25,426

https://sql.telemetry.mozilla.org/queries/304/source#503

Comment 4

[:Gijs (he/him)]

(Thanks for looking at this stuff!)

(In reply to Matthew N. [:MattN] (behind on reviews) from comment #3)
> OK, that's unfortunate and probably should be more clear on the dashboard as
> I suspect people don't realize that
> 
> The longitudinal SQL table has sane data (not version-specific):
> 0	41
> 1	976,776
> 2	152,956
> 3	16,569
> 4	25,426
> 
> https://sql.telemetry.mozilla.org/queries/304/source#503

Should we mark this as WFM and file a followup bug to make a note on the limitations of the dashboard?

Comment 5

[Matthew N. [:MattN]]

I think this is no longer an issue with how opt-in/opt-out works.