Date of slack thread: 4/26/24
Anonymous: Hi team, I have 2 experiments and the use same 1 metric (with the same source event). Because we send the conversion event on the same page and also the UserId is the same. So I’m not sure if the Statsig report is double counting if I send 2 the same events for the same UserId (the difference is on the User Attribute). See the example {UserId: test_1, VisitPageA: 1} -> Pass the gate and enter Experiment A {UserId: test_1, VisitPageB: 1} -> Pass the gate and enter Experiment B
On the conversion page, If UserId test_1 entered both experiments A and B, I plan to send the event twice For Experiment A {eventName: conversion, {UserId: test_1, VisitPageA: 1} For Experiment B {eventName: conversion, {UserId: test_1, VisitPageB: 1} -> My concern here is the report of Experiment A and B is double counting because of the same eventName, same UserId
Statbot: When a user sends multiple instances of the same event, Statsig will aggregate these events together, even if they have changes in metadata. This means that if you send the conversion event twice for the same UserId, Statsig will count both events. However, you can use metadata and values to generate custom metrics in the console, which can help differentiate the events in your reporting. For example, you could include VisitPageA or VisitPageB as part of the metadata to distinguish between the conversions for Experiment A and Experiment B (source). To avoid double counting in your reports, you should ensure that the events are logged with the appropriate metadata so that they can be correctly attributed to each experiment. If you’re still concerned about potential double counting, you may want to review how you’re using metadata and values to differentiate the events in your experiments.
Anonymous: Because I setup the gate base on the custom attribute also, so when I send the event with {UserId: test_1, VisitPageA: 1} it should only pass the gate A and count for report of exp A. That is my expectation
Vijaye (Statsig): If the same user enters Exp A & Exp B, and if you send two events, those two events will be counted towards Exp A and Exp B. Events are completely agnostic of experiments you run. We just count the number of events per day and use that for calculation. If you want to dedupe it by user, then you can use the Event DAU metric instead of Event Count metric.
Anonymous: I see, thanks <@U01RGJ9N4N8>, we’ll find another way to figure it out, e.g. using different event for each experiment
Statbot: This thread has been marked as resolved. Press below to re-open.