We can leverage:

• loggedOutReaderRetention.js
• loggedInReaderRetention

Backfill rate is about 8 minutes for 1 day of data. The backfill is 90 days, which will take approx. 12 hours to backfill at this rate. The table should be fully up-to-date by Monday morning.

@MNeisler: wmf_product.personal_dashboard_event_v1 now backfilling from 2026-03-15

By the way, @JVanderhoop-WMF has proposed that Experiment Platform team own this as a hypothesis under DE1.1 in Q1, pending approval from @ppelberg.

Wikipedia 25 microsite instrumentation had to copy MW's implementation for generating random IDs https://gitlab.wikimedia.org/repos/sre/miscweb/wikipedia25-years-of-wikipedia/-/blob/main/src/utils/generateRandomSessionId.ts?ref_type=heads

Blocked by https://test-kitchen.wikimedia.org/experiment/we-1-8-account-creation-form-v2 since that's currently configured to send events to the custom contributors experiments stream rather than the base web stream.

I'm with @JAllemandou here. Also, why not just update the UDF that produces normalized_host so that it also includes wiki_id?

Docs homepage: https://wikitech.wikimedia.org/wiki/Test_Kitchen/GrowthBook_user_guide

We didn't get around to doing this this FY. We are considering making a standard suite of metrics for experimentation in the next FY. Will create a new ticket for that if we end up prioritizing that work.

This was an early spitball around the standard suite, but won't be the same as is actually implemented. Closing as such.

The linked dashboard no longer works (possibly because the team that made it got disbanded).

@KReid-WMF @phuedx: does this task as written still make sense or should we repurpose/redefine it or decline it altogether?

@JVanderhoop-WMF: Should the dimension be a simple "assigned to other experiments' treatment (Y/N)"? (As opposed to including the exact experiment & group name)

There are three fields that represent Goal, Secondary and Guardrail Metrics. Are they required for GrowthBook experiments to work properly?

Report up at https://analytics.wikimedia.org/published/reports/test_kitchen/event_loss_remediation

@dr0ptp4kt: Hi! I was bold and updated the docs:

@JVanderhoop-WMF because of https://wikitech.wikimedia.org/w/index.php?title=Test_Kitchen%2FConduct_an_experiment&diff=2406884&oldid=2404453

@JVanderhoop-WMF gave a thumbs up

We'll want to communicate this with teams. I see a lot of hypotheses and KRs of 1% or 2% increases. It is hard to detect a change that small

https://wikitech.wikimedia.org/wiki/Test_Kitchen/Decision_Records/Keep_Bayesian_engine_as_default_in_GrowthBook

Experiment results now available: https://superset.wikimedia.org/superset/dashboard/p/ZEXOPgzLrKb/ (permalink to this specific experiment's results in https://superset.wikimedia.org/superset/dashboard/experiment-analytics/)

@JVanderhoop-WMF: Do you have any additional questions?

Okay, so having read through the docs and even played with this feature both in staging and prod, here are my takeaways:

how to enable it

Waiting for feedback from Katherine & Sam

Us

Do GB experiments have contextual attributes, in the same way that experiments in TK UI do? Same with stream name. Or are we doing away with those conceptual models?

Sounds good to me! We should just have performer_session_id as THE session identifier and then tweak the internals of that. E.g. currently it uses mw.user.sessionId() (from MW Core) but we should manage that identifier's lifetime.