Conversion Measurement API Origin Trial Page Usage Exception

[John Delaney]

Hello API Owners,

We would like to request an exception to the .5% Origin Trial global page load usage limit for the Conversion Measurement API [1] Origin Trial currently running though M91.

The Conversion Measurement Origin Trial is being run as a 3P origin trial, and has numerous partners who use both 1P and 3P tokens. Our origin trial supports the  "Exclude a subset of users" usage restriction[2]. Origin trial tokens configured with this usage restriction only enable the API on a fraction of Chrome devices.

The Conversion Measurement API involves multiple actions across different top-level sites to use the API. The "Subset exclusion" configuration prevents developers from needing to do any kind of persistent cross-site experimentation in order to try out the API. Developers who wish to do their own experimentation are free to use the Standard Limit tokens which work across all Chrome clients.

Currently, these subset exclusion tokens only enable the API for ~3% of Chrome users. We would like to increase this fraction to roughly 50% of Chrome users.

We expect that this change will push total page load usage past the .5% Origin Trial threshold based on historical data as seen here:

https://chromestatus.com/metrics/feature/timeline/popularity/3365

In December there was ~.02% global page load usage across ~1% of clients, scaling this to 50% of clients results in >1% global page load usage.

Based on these numbers, we think a  5-10% global page load usage limit would be sufficient.

We think this increase is necessary to facilitate more robust testing of the API. Namely:

• For sites which generate small numbers of conversions (e.g. <100), it is difficult to have confidence in measurements when only 3% of clients have the API enabled.
• We have received feedback that partners are seeing data loss between data gathered from the API and existing measurement solutions, which larger experiment volumes would give us more confidence in.

For sites using the subset exclusion tokens, burn-in risk is mitigated as sites still need to account for the portion of users with the API disabled.

Note that this change would not affect behavior for Origin Trial tokens configured with the Standard Limit, and sites using these tokens do not benefit from the same mechanism to mitigate burn-in.

As noted in our initial I2E [3], we believe the risk of developers relying on the API over their existing mechanisms to be low due to the many restrictions the API imposes on the data it reports.

Happy to answer any questions/concerns you may have, thanks in advance! If this belongs on blink-dev instead of this list, I would be happy to post it there as well.

[1]

https://github.com/WICG/conversion-measurement-api

[2] https://docs.google.com/document/d/1xALH9W7rWmX0FpjudhDeS2TNTEOXuPn4Tlc9VmuPdHA/edit#heading=h.gxula5b2ky88

[3] https://groups.google.com/a/chromium.org/g/blink-dev/c/Ib9-tDFitns/m/Sm9RXTNaBwAJ

[Chris Harrelson]

Hi John,

Sorry for the slow reply. I was on vacation last week and didn't process this email in time for the API owners meeting yesterday.

We've also received requests to increase from two other experiments with similar motivations - in particular, getting enough data from a collection of sites to be statistically significant for analysis.

Our plan is to discuss these situations at the API owners meeting next week on Thursday. Would it be ok for us to discuss your case then and then get back to you right after? You're also welcome to attend to help us make a good decision (it's at noon Pacific time).

Regards,

Chris

--
You received this message because you are subscribed to the Google Groups "blink-api-owners-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blink-api-owners-d...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-api-owners-discuss/96cb7356-8318-4bf7-ab2c-54552d604b96n%40chromium.org.

[John Delaney]

Thanks Chris! That plan sounds good to me, thanks again!

John

[Chris Harrelson]

Hi John,

I sent an email to blink-dev announcing the new agreement to allow increased Origin Trial limits for cases such as Conversion Measurement. Could you send an official request to increase limits to blink-dev, so we can approve it tomorrow?