Designing “Memory Rescue” With OneSignal (Without Turning Slovo Into a Notification Machine)

The article discusses a retention experiment called "Memory Rescue" for a Russian learning app called Slovo. The app utilizes push notifications, but only strategically, to avoid overwhelming users. To determine if a notification is useful, Slovo tracks specific state tags such as the learning level, streak days, review due count, and daily completion status.

Based on these state tags, the app decides whether to send a notification inviting the user back to recall recently learned words. Eligible learners are those with reminders enabled, today's goal incomplete, and their last completed word sufficiently old. The notification content varies depending on whether review due count is zero or not, aiming to provide a clear call-to-action.

When a user clicks the notification, the app uses deep linking to direct them to the appropriate task, such as reviewing a set of words or viewing their progress. The app also has exit rules to prevent notifications from being sent once certain conditions are met, like completing the daily goal or finishing a streak. To avoid potential issues, the campaign includes a 10% holdout group and measures various metrics like retention, notification opens, and opt-out rates.

The author acknowledges that the implementation should not be considered evidence that the campaign works, as there could be alternative explanations for improved retention. Potential improvements to the idea include reducing frequency, removing redundant messages, or relying more on widgets for reminders. The author plans to publish further evidence to determine the success of this campaign.