Engineers at Instagram have combined causal inference and machine learning (ML) to improve the user experience for daily digest push notifications about stories. This new method provides a more personalized experience on Instagram and brings people the notifications that are most important to them.

How it works

We’re sharing an example of how we used causal inference and ML to control sending for a type of daily digest push notifications – digests of stories for people to view. Under older models, ML was used to predict how likely someone is to click on a notification and thus judge whether someone believes a notification is of high or low quality. The lower quality the ML model infers a notification to be, the less likely the notification will be pushed to the user.

This older system worked well, but it also opened an opportunity to serve fewer notifications to people who are highly active on Instagram. After all, if you’re likely to find and view a story organically, there’s no need to receive notifications about it.

Using a combination of statistical methods and ML, Instagram’s engineers were able to reduce the number of notifications sent to highly active people on Instagram without negatively impacting their overall engagement with the platform.

Why it matters

Personalized notifications play an important role in everyone’s experience on Instagram. People want to see the notifications that are most important to them and avoid ones that they see as excessive or unimportant.

By applying machine learning and statistics, engineers at Instagram have been able to significantly improve overall user experience. We’ve taken a quality over quantity approach that helps people see fewer notifications and ones that are the most important and interesting to them.

Take a deeper dive:

Learn more about, “Improving Instagram notification management with machine learning and causal inference.”