Hello, we're integrating Growthbook + Mixpanel int...
# ask-questions
m
Hello, we're integrating Growthbook + Mixpanel into our application. Our primary usage is a checkout flow. Facing some issues with how the IDs are handled alongside the
StickyBucketing
Config 1. Configured a simple experiment with StickyBucketing activated a.
id
-> hash b.
deviceId
-> fallback attribute c. LocalStorageStickyBucketService 2. Configured mixpanel integrations without any additional property 3. When mixpanel is loaded, we set the deviceId. Otherwise, the SDK starts with an empty attributes config 4. Next.js integration with React sdk 5. Using
LocalStorageStickyBucketService
User Flow 1. The user (generally) lands anonymous a. We add a
deviceId
property that points from
mx.get_distinct_id()
2. The experiment starts 3. The user fills out the form and signs up a. We
mx.identify(user_id)
b. Add set the
id
proeprty of the user to growth book sdk while keeping the deviceId c. Dispatch the event to mixpanel 4. After login, an additional One additional call to
trackingCallback
seems to be made, with the new id as an attribute Expected Result • A single metrics result is generated, pointing w/ 100% rate Current result • 2 metric results are created(one anonymous, one logged most likely?), resulting in a 50/50% rate Are we missing something? Shouldn't
StickyBucketing
handle this? Or is it a misconfiguration? Thanks!
f
That is working as intended
when you go to analyze results, you should select anon / device id as it will include 100% of your traffic (as thats what you're randomizing on)
As you have it, If someone is logged in initially, that will be used to bucket them - for everyone else will use the anon id to bucket them
g
@fresh-football-47124 our problem is that we: • Init Growthbook with the Mixpanel anon distinct ID; • We initialize an Growthbook AB test using the anon distinct ID; • After the user completes the frontend flow, we create an user account automatically, and call mixpanel.identify(UUID); At the end of the process, MP can track the previous anonymous ID and correlate the metrics with the new UUID using the simplified merging strategy, however seems like GB can't do that and is unable to tell how many users completed the AB test, since it thinks two different users are using the experiment (Random ID !== UUID), how can we correlate the events on GB too? See baseline
0/3
Also what is
Extra UserId Property (optional)
used for when configuring the MP data source?
Do you have any example of GB + MP integrated with the IDs merging?
m
Update: Some of our tests on experimentations have worked. We're unsure why the initial status was 0%, and then, after a few hours, it went to 100% as shown below Is there a process that runs in the background? Or is it something else?
f
it does update periodically
🙌 1
depending on what your update frequency is set to
m
We've had a mixpanel event, on the 12th, but it hasn't appeared as a metric for the experiment. Our other metric(for wire) was triggered and related. Can we debug this? I couldn't find a way to do so
Hello @fresh-football-47124, we're still facing issues. Is there a way to help us? We've got 2 more events from mixPanel, which are still not showing in the experiments. Is there a way to debug this? Matching the IDs of the events?
f
Let me check
m
Hey, @fresh-football-47124, unfortunately, it's still not working. Can we sort this out? Could we do a 15-minute call with someone who can help us? We would love to have it working! Maybe needs a fine-tuning? We can show our codebase and check if everything is correct.
f
ya, one sec
are you free now?
m
Yeah!