Unified Account Recovery
Unified Account Recovery
Scope
Scope
I acted as Growth design lead on a cross-org effort with the Integrity org to unify Meta's fragmented account recovery flows — the first step toward a consolidated authentication platform. We piloted the unified experience across three cohorts (Supertrusted, Trusted, Untrusted), rolling out in phases from H1 2025 through 2026.
I acted as Growth design lead on a cross-org effort with the Integrity org to unify Meta's fragmented account recovery flows — the first step toward a consolidated authentication platform. We piloted the unified experience across three cohorts (Supertrusted, Trusted, Untrusted), rolling out in phases from H1 2025 through 2026.
Problem
Problem
Complex, outdated systems Facebook has multiple recovery products owned by different teams, each with its own logic, auth methods, and security bars. Mapping out the how these products interact made it clear how the current state negatively impacts users. Users trying to recover their account can face redundant challenges, get routed through a maze of multiple flows, or fail to recover at all — only 41% of hacked users who attempted recovery succeed.
Complex, outdated systems Facebook has multiple recovery products owned by different teams, each with its own logic, auth methods, and security bars. Mapping out the how these products interact made it clear how the current state negatively impacts users. Users trying to recover their account can face redundant challenges, get routed through a maze of multiple flows, or fail to recover at all — only 41% of hacked users who attempted recovery succeed.
Reputational damage Users having account access issues became a growing, visible problem for Meta — as hacked-account complaints rose, so did the reputational and press risk of leaving users locked out of their accounts. This made improving account recovery a company-level priority.
Reputational damage Users having account access issues became a growing, visible problem for Meta — as hacked-account complaints rose, so did the reputational and press risk of leaving users locked out of their accounts. This made improving account recovery a company-level priority.
North Star Vision
North Star Vision
All users can regain access via a single, entrypoint-agnostic experience powered by a central signals and decisioning framework and a challenge set that is secure and efficient to optimize, scale, and maintain.
All users can regain access via a single, entrypoint-agnostic experience powered by a central signals and decisioning framework and a challenge set that is secure and efficient to optimize, scale, and maintain.
Decisions
Decisions
Unify by trust level, not by entry point. Rather than one-size-fits-all, we segmented users into three cohorts by their device/account trust signals — Supertrusted (15%, no challenge needed), Trusted (50%, one challenge), Untrusted (35%, two challenges), and built one decisioning system that treats users consistently regardless of where they enter the flow.
De-risk by starting with the Supertrusted cohort first. We first piloted with the safest cohort of Supertrusted users, validating the unified flow's UX and guardrails before expanding to riskier cohorts — a deliberate sequencing choice to reduce rollout risk.
Unify by trust level, not by entry point. Rather than one-size-fits-all, we segmented users into three cohorts by their device/account trust signals — Supertrusted (15%, no challenge needed), Trusted (50%, one challenge), Untrusted (35%, two challenges), and built one decisioning system that treats users consistently regardless of where they enter the flow.
De-risk by starting with the Supertrusted cohort first. We first piloted with the safest cohort of Supertrusted users, validating the unified flow's UX and guardrails before expanding to riskier cohorts — a deliberate sequencing choice to reduce rollout risk.
Trusted
Trusted
For Trusted users, we consolidated multiple challenge picker screens from the different recovery products into one version. We also baked in known growth wins like auto-sending codes to contact points we’re confident users have access to.
For Trusted users, we consolidated multiple challenge picker screens from the different recovery products into one version. We also baked in known growth wins like auto-sending codes to contact points we’re confident users have access to.
Untrusted
Untrusted
For Untrusted users, chain two challenges (a first-factor and second-factor) using a single dynamic method selector rather than two rigid, separately-categorized pickers — since challenge categories weren't clean binary buckets, a unified selector that adapts based on the first choice was simpler for users and easier to build.
For Untrusted users, chain two challenges (a first-factor and second-factor) using a single dynamic method selector rather than two rigid, separately-categorized pickers — since challenge categories weren't clean binary buckets, a unified selector that adapts based on the first choice was simpler for users and easier to build.
Outcome
Outcome
The Supertrusted pilot shipped with strong results that validated the success of UAR: attempt-to-clear rate improved from 79% to 92% (+16%), a projected lift of +9 pts in hacked-account clears/day in US/Canada, and +55K benign MAU — all while holding integrity guardrails (rehack rate, topline compromise) neutral.
UAR cohorts will continue to rollout in phases: Trusted is testing in H1 2026, and Untrusted is sequenced after.
The Supertrusted pilot shipped with strong results that validated the success of UAR: attempt-to-clear rate improved from 79% to 92% (+16%), a projected lift of +9 pts in hacked-account clears/day in US/Canada, and +55K benign MAU — all while holding integrity guardrails (rehack rate, topline compromise) neutral.
UAR cohorts will continue to rollout in phases: Trusted is testing in H1 2026, and Untrusted is sequenced after.
Recovery clear rate
Recovery clear rate
79% → 92%
79% → 92%
Hacked clears
Hacked clears
+9 pts/day
+9 pts/day
Monthly active users
Monthly active users
+55K
+55K
Integrity guardrails
Integrity guardrails
Neutral
Neutral
Next Steps
Next Steps
How might we expand the features that UAR offers? I vibe-coded a proof of concept for AI support in UAR via Xcode and Cursor to kickstart discussion amongst the team.
How might we expand the features that UAR offers? I vibe-coded a proof of concept for AI support in UAR via Xcode and Cursor to kickstart discussion amongst the team.
Learnings
Learnings
Org politics usually slows work down as teams get protective of their surfaces and resistant to anything that risks their metrics (Growth optimizes for topline MAU, Integrity for keeping bad actors out). But despite different goals, this was genuinely one of my favorite collaborations and we shipped something both teams were proud of.
Org politics usually slows work down as teams get protective of their surfaces and resistant to anything that risks their metrics (Growth optimizes for topline MAU, Integrity for keeping bad actors out). But despite different goals, this was genuinely one of my favorite collaborations and we shipped something both teams were proud of.
"Thanks for all the productive collaboration… and spending so much effort understanding how the compromise space works to integrate that effectively with growth flows. I've done a bunch of x-org projects and few have gone so smoothly as this one"
"Thanks for all the productive collaboration… and spending so much effort understanding how the compromise space works to integrate that effectively with growth flows. I've done a bunch of x-org projects and few have gone so smoothly as this one"
—Integrity eng lead
—Integrity eng lead