Hello. Does the conversion window apply to users even if an experiment has stopped? E.g. a metric has a conversion window of 336 hours, a user is exposed to the experiment today but we end the experiment after, will his behavior still influence the metrics afterwards?

Hi Frederick, Good morning and thank you for writing in. Yes, the conversion window can apply to users even after an experiment has stopped. This is because the conversion window is designed to capture the period during which a user can convert after being exposed to an experiment. If a user's first exposure to an experiment was recent or near the end of a stopped experiment, they may not have had the full window to convert before the analysis window closes. However, it's important to note that the setting for "Include In-Progress Conversions" or "Exclude In-Progress Conversions" can affect this. If you choose to "Exclude In-Progress Conversions", users who have not had the full window to convert by the time the experiment ends will not be included in the analysis. In some cases, users who were exposed to the experiment late may still have their conversion window extend beyond the experiment end date. This is particularly relevant for subscription-based businesses where the goal metric (like revenue) might come after the experiment end. Further information can be found on the resources below: https://docs.growthbook.io/app/experiment-configuration#analysis-settingshttps://docs.growthbook.io/using/experimenting#metric-windows Hope this helps :)

Thanks for the quick response! I am wondering if we need conversion windows at all then and how they help? Could you give me a use case when it makes sense?

My pleasure, it depends on your business case but happy to outline a few usecases where conversions windows are beneficial: 1. ​*Reducing Noise from Unrelated User Behavior*​: If you are tracking specific actions like purchases, and you only want to measure the effect of an experiment in a checkout flow on purchases made soon after seeing that checkout flow, a conversion window can help reduce the noise from user behavior not related to an experiment. 2. ​*Capturing Long Run Impacts*​: Lookback windows, a type of conversion window, are good for capturing long run impacts of an experiment on regular behavior like user logins or page views. They can help mitigate the novelty effect of an experiment and focus on the long run effects of an experiment. 3. ​*Handling Delayed Effects*​: In subscription-based businesses, the action (like signing up for a trial) might come before the experiment end, but the goal metric (like revenue) might come after the experiment end. In such cases, a conversion window can be used to include data that comes after the experiment end. Hope this helps :)

super helpful, thanks!

My pleasure 🌼