IG notifications ecosystem

Redesigning Instagram's Notification Ecosystem

Notifications ecosystem

The problem

Instagram's notifications were generic. Every push used the same format regardless of content type, sender or what a user actually cared about. The display format didn't surface the most engaging signal, the onboarding suite was too transactional to drive the behaviors most correlated with long-term retention and no one had tested notification content systematically. The team had hard engagement goals and the content layer had never been treated as a meaningful lever.

How I knew

I audited the notification suite end to end and worked with eng to trace the onboarding funnel data. The retention-correlated actions (syncing contacts, completing a profile) had notifications, but the content didn't explain why those actions mattered now. I also read the signal data: notifications that led with an action type rather than the display name of someone the user knew were opened far less often. The problem wasn't the frequency or timing. It was what the notification led with.

My goal

Redesign the ecosystem to surface the most relevant signal first, build a new onboarding suite that drove users toward the specific actions most correlated with retention and establish a repeatable framework for testing voice and tone in notifications.

My process

I pitched the display name-first approach, which required working through integrity implications with engineering before we could ship. Surfacing display names more prominently created new risks around spam and offensive language in teen-facing notifications. We shipped the design change alongside new safeguards, and the engineering team ended up building additional blocklists to detect and filter spam and offensive language.

The content change made Instagram safer for teens as a side effect.

For the onboarding suite, I mapped the actions most strongly correlated with long-term retention, then designed notifications for each that explained why rather than just what. Completing your profile and syncing contacts aren't self-evidently valuable to a new user. The content had to make the case for them. I ran the suite through a structured launch process and iterated on the copy through early performance data.

I also designed and launched experiments on personalized in-feed recommendation notifications and a new frequently-engaged notification type, each run as a controlled experiment with a clear behavioral hypothesis.

Outcomes

Display name-first: +3.0M IG DAU (+0.2%), +41K developed market teen DAU, +0.18% ads revenue, +0.077% overall time spent
Onboarding suite: +15.5% CTR, +0.22% gen-pop DAU, +0.33% more contact syncs, +0.26% more profile completions
Personalized in-feed recs: +3M total DAU; frequently-engaged type: +1.25M DAU
Integrity win: engineering built additional teen safeguards as a direct result of the display name work

What I learned

Notification design is a trust problem as much as an engagement problem. The display name change worked because it surfaced something real: a person the user actually recognized. Personalization that's genuinely specific outperforms personalization that's just optimized for click-through, and the integrity outcome reinforced that. The right signal is also the safer signal.

The onboarding result that surprised me most was profile completion. Users weren't avoiding it because they didn't want to do it. They were avoiding it because nobody had explained why it was worth doing right now. That's a content problem.