Open source platform for stress free deployments, measured impact, and smarter decisions.

GrowthBook Users

Hi all, I noticed this in the <http://If%20you%20return%20multiple%20rows%20for%20the%20same%20user/experiment,%20we%20unique%20the%20list%20(ie:%20only%20counts%20as%20one%20user%20in%20the%20experiment)%20and%20count%20the%20earliest%20timestamp.%20From%20that%20point,%20the%20user%20has%2072%20hours%20(or%20whatever%20you%20change%20the%20window%20to)%20to%20convert.|docs>.
&gt; If you return multiple rows for the same user/experiment, we unique the list (ie: only counts as one user in the experiment) and count the earliest timestamp. From that point, the user has 72 hours (or whatever you change the window to) to convert.
What query is this referring to? The metric or experiment assignment? I'm not really sure what it's saying.

This is referring to the experiment assignment. This documentation is currently being refreshed since we're replacing the "Multiple Exposure" model with the "Experiment Duration" model.

How does this work with All Exposures? The <http://If%20a%20user%20sees%20an%20experiment%20multiple%20times,%20you%20should%20return%20multiple%20rows%20in%20your%20assignment%20query,%20one%20for%20each%20time%20the%20user%20was%20exposed%20to%20the%20experiment.|Data Sources> section says
&gt; If a user sees an experiment multiple times, you should return multiple rows in your assignment query, one for each time the user was exposed to the experiment.
If you just take the minimum of all these rows, then only the first exposure would matter.

Yeah, so I think the docs are just out of sync on this point.

The Metrics doc there is assuming the First Exposures model.

This exact section will be rewritten when we replace Multiple Exposures with Experiment Duration.

Ok got it. It seemed to be describing First Exposure. Just wanted to make sure our All Exposures Attribution Model was still working as expected

Yep, it should be. Sorry for the confusion.

(but I think as I've mentioned to you or your colleague before, we're sunsetting that for the Experiment Duration model which will extend from the start of their conversion window through the end of the experiment, which is a more common use case &amp; the query is more performant)

No worries, thanks for confirming! And yes, we're excited to be able to use that new model!