973293 - Firefox counts the number of times that it detects a translation opportunity

Status: RESOLVED FIXED

(Firefox Health Report Graveyard :: Client: Desktop, defect)

Description

[Marco Mucci [:MarcoM]]

1.  Goal
* As a product owner I want to know how many times Firefox offered to translate a page so that we can get a sense of the size of the user problem we are addressing by offering this feature

2.  Acceptance Criteria
* I can view a report that tells me how many times our users visited a page that was written in a language other than their default language
* This report will be updated weekly

Comment 1

[Benjamin Smedberg]

Do we really expect to be monitoring this data weekly indefinitely? Is it sufficient to measure the audience who have opted into telemetry? What data do we need to correlate this against (build locale, offered translation pair)? Do we need to measure how individual user behavior changes over time, or can we use simpler point-in-time metrics?

This should probably be duplicated with bug 973294: having them as separate work items doesn't make much sense.

Comment 2

[Chad Weiner [:chadw]]

Adding additional acceptance critiera:

* I can view the gross number of times that Beta users of Firefox on every build visit pages written in languages other than the user’s default for all users in the test and control groups
* I can view a breakdown of this number that lets me know how frequent each non-default language was detected (e.g., users say 10 pages written in French, 9 pages written in Spanish and 8 pages written in Sanskrit)
* I can view this report for every Firefox locale (i.e, this is not limited to the test locales, this is for every Firefox Beta user)

Comment 5

[Steven MacLeod [:smacleod]]

(In reply to Chad Weiner from comment #2)
> * I can view a breakdown of this number that lets me know how frequent each
> non-default language was detected (e.g., users say 10 pages written in
> French, 9 pages written in Spanish and 8 pages written in Sanskrit)

This would require tracking how many times a profile visits pages for each language which isn't their locale... I don't know how I feel about the privacy implications here. Is there a way we can track this in a privacy respecting way?

Comment 6

[Benjamin Smedberg]

Remembering that this is a telemetry measurement not an FHR measurement, can you clarify what you're specifically concerned about? Do you fear that this would be an identification risk or that the data itself is sensitive?

I don't see a problem with a measurement of visited-page-languages in general.

Comment 7

[Steven MacLeod [:smacleod]]

(In reply to Benjamin Smedberg  [:bsmedberg] from comment #6)
> Remembering that this is a telemetry measurement not an FHR measurement,

Ah, okay. I was under the impression this was part of FHR since it's blocked
on Bug 978158.

> you clarify what you're specifically concerned about? Do you fear that this
> would be an identification risk or that the data itself is sensitive?

I was concerned the data itself might be sensitive. It just gave me a strange
feeling which I thought I should bring up.

Comment 8

[Tim Taubert [:ttaubert] (inactive)]

Bug 978158 fixes this one by calling the new method "recordTranslationOpportunity()". This should be closed when bug 978158 lands.

Comment 9

[Tim Taubert [:ttaubert] (inactive)]

Marking as fixed by bug 978158.