Is it possible to preserve the behavior of an expe...
# announcementsb
busy-horse-73824
09/14/2023, 12:14 PMIs it possible to preserve the behavior of an experiment, but turn off its tracking? When we finish up with price testing, we typically want to keep the affected users' prices consistent forever, but no longer want to consume analytics events with it
busy-horse-73824
09/14/2023, 12:15 PMHistorically I've found that rollout rules seem to work in a similar way for the most basic setups (experiment key == feature key, variations == 2) - but they don't work with different experiment keys, or multi variant tests
b
brief-honey-45610
09/15/2023, 4:10 AMHi James, we had an unusually large number of support requests come through the Slack workspace today and we're still catching up. We'll look into this tomorrow. I appreciate your patience!
h
happy-autumn-40938
09/15/2023, 4:44 AMHi, you should be able to use a "temporary rollout" to apply the winning variations across the board until you've had time to make permanent changes in code. Using the temporary rollout will also disable the trackingCallback in the SDK.
Here's a screenshot showing where to set up the temporary rollout. It happens after stopping an experiment
✅ 1
b
busy-horse-73824
09/15/2023, 5:26 PMAh that is interesting, but we actually need to preserve the existing assignment. So if the distribution was e.g. 33/33/34, we'd need to ensure it remained with that same distribution (being evaluated with the same hashing inputs) permanently
busy-horse-73824
09/15/2023, 5:27 PMIt's effectively that we want the experiment's effect on the feature to remain, but have it no longer call the tracking callback
h
happy-autumn-40938
09/15/2023, 5:36 PMI don't think we support a streamlined way of doing this within GB at the moment. Until then, could you just leave the experiment running and semi-disable the trackingCallback? The callback does pass in the experiment object, so you could filter your 3rd party tracking calls based on experiment id.
b
busy-horse-73824
09/15/2023, 6:09 PMFair enough, thanks
27 Views