Connector name: Marketo
Table name: email_template_history
API documentation link: https://experienceleague.adobe.com/en/docs/marketo-developer/marketo/rest/assets/emails

Additional details:
Request: please document, in the Marketo connector ERD or schema notes, that
EMAIL_TEMPLATE_HISTORY cannot contain A/B-test or Champion/Challenger email variants, because
Marketo's Asset API does not return them.

This is a Marketo platform limitation rather than a Fivetran defect. Adobe's documentation
states that Get Email by Id, Get Email by Name and Get Emails all exclude emails belonging to
an email program that uses A/B testing, while still counting those variants in the totalItems
value. I am not asking for a sync change. I am asking for the limitation to be stated where
somebody joining activity data back to the email asset would actually find it.

Why it is worth documenting: joining ACTIVITY_EMAIL_DELIVERED.PRIMARY_ATTRIBUTE_VALUE_ID to
EMAIL_TEMPLATE_HISTORY.ID silently fails for exactly these rows, and the failure is invisible
unless you decompose the result by asset name. Measured on one Marketo instance:

- ordinary sends: 1,627,748 of 1,627,776 rows resolve (99.998%)
- A/B-test variants: 0 of 369,595 rows resolve (0%)

The blended figure is 81.5%, which reads like a data-quality problem rather than a documented
API boundary. 27 distinct asset ids account for all of the unresolved rows, and every one of
them is named with a Marketo A/B suffix (.Subject Line Test, .Whole Emails Test,
.From Address Test). Those ids share the same numeric range as ordinary email assets, so
nothing about them looks unusual on inspection.

It is also not a historical-depth issue. The unresolved share rises over time as A/B usage
grows: 2.15% in 2023, 18.54% in 2024, 18.97% in 2025. ACTIVITY_SEND_EMAIL shows the identical
pattern.

Impact scales with how much a given customer A/B-tests. In one instance it is under 1% of
sends; in another it is around 18%, which is close to one delivered email in five that cannot
be resolved to its asset at all.

One more reason this is worth writing down: there is an existing community post
(https://support.fivetran.com/hc/en-us/community/posts/11963547119895) whose only reply says
these data points are in EMAIL_TEMPLATE_HISTORY. That is true for ordinary sends and not for
A/B variants, so the current public answer is misleading.

Note for anyone who finds this while debugging: PRIMARY_ATTRIBUTE_VALUE_ID is the correct join
key. EMAIL_TEMPLATE_ID is populated on only about 35% of activity rows in the instance I
measured and is NULL on the highest-volume sends, so it is not a workaround.