1160636 - metrics, user stories - allow query of "how many users of type X"

Status: NEW

(Cloud Services :: Metrics: Product Metrics, defect, P3)

Description

[Thomas Huelbert]

As Quality and Metrics manager, I want to be able to query how many active users of type X we have. Allow users to query, in a performant fashion (less than minute would be good) the following: Release Channel, Locale, Geo, default browser, default search engine, telemetry on status

Comment 1

[Mark Reid [:mreid]]

Please break out any sub-tasks as separate bugs. We discussed implementing this as a method of querying the internal state of a cuckoo filter in a running sandbox.

Comment 2

[Mike Trinkala [:trink]]

This aligns very closely with the executive report; if the initial comment contains the complete set of dimensions the differences are:
- We don't track locale
- We don track the default search engine but we have the actual search counts for [Google, Bing, Yahoo and Other]
- We don't track the telemetry status

Telemetry status would be easy to add since we have a little room left in the executive report.  Locale and default search engine would incur extra storage costs based on the number of each you would like to store (please provide those dimensions i.e. are you interested in only the ISO 639-1 locales? The eight search engines we provide by default?). There would have to be some basis to query against, the last day, week, month, beginning of time etc.  To run in real time all this needs to be defined upfront.  If the specification cannot be predefined if becomes a retroactive ad-hoc analysis turning it into a report rather than a 'quick query' against a precomputed result set.

Comment 3

[Mark Reid [:mreid]]

From triage:
- Is the list of dimensions above complete, or was it just an example?
- We will create a redshift output with this information that can be queried interactively.

Comment 4

[Katie Parlante]

Rebecca is probably the best person to answer the question above (is the list of dimensions complete). It would also be helpful to define the dimensions & time frame, as trink asked in #2:

(In reply to Mike Trinkala [:trink] from comment #2)
> (please
> provide those dimensions i.e. are you interested in only the ISO 639-1
> locales? The eight search engines we provide by default?). There would have
> to be some basis to query against, the last day, week, month, beginning of
> time etc.

Comment 5

[[e-mail]]

My understanding is that this service addresses a user's need to receive an update about our ongoing performance along key dimensions, rather than be used for 1) any in-depth analysis or b) long-term longitudinal study.  

Before verifying whether we have all the dimensions or a specific time frame, a quick set of questions:

1) Are we married to our list of dimensions once this bug is closed, or is it affordable to add more later (in the case that we get an additional request)?  

2) If we start with one time frame and determine that the more common use case is longer term, is that costly in terms of both time and budget?  I understand that switching from 1 day to 1 week would require another week of time in order to gather the data (we can't retroactively add days back).