Most teams that plan a passkey launch spend their time on the integration: the WebAuthn server, the login page, the enrollment prompt. Once that works in staging, the question changes. How do you get from a working build to millions of customers without a spike in support calls, a broken device segment or a campaign that promises something half of your users cannot use yet?
Enterprise Passkey Whitepaper. Practical guidance, rollout patterns and KPIs for passkey programs.
This is where many passkey programs lose time and trust. A rollout treated as one big release leaves little room to react. Customer support learns about passkeys from the first confused caller. Marketing announces "passwordless" while Windows users still see no offer. And when one Android version misbehaves, the only way to stop it is an emergency release.
This passkey rollout plan describes how teams in banking, insurance, retail and the public sector can move consumer passkeys from a working build to all users in controlled steps. It answers:
Recent Articles
A passkey rollout is a sequence of stages. Each stage has its own audience, its own focus metric and an explicit exit rule. Plan for roughly two to three months from the first staging tests to all ready devices.
| Stage | Audience | Focus metric | Move on when |
|---|---|---|---|
| 1. Test and staging | Test devices per OS and browser | Tracking works end to end | All in-scope combinations tested, tracking complete |
| 2. Internal pilot | Colleagues in production | Every dialog and its error class | Fixes shipped, support FAQ ready |
| 3. First customers | One small OS segment | Readiness, first offer, login success | Two clean days, no new error class |
| 4. Device waves | More OS versions every one to two days | Fallbacks and aborts per OS | Wave check passed |
| 5. All customers | All ready devices, plus offer tests | Rates per offer, entry path and cohort | Rates stable, support calls calm |
| 6. Steady state | Everyone, platform updates watched per version | Passkey login share against the target | Target met or next program phase |
The reporting rhythm changes with the stages: per test run in staging, daily per segment during the pilot and the waves, weekly with a projection once all customers are in and monthly to leadership in steady state.
The important part is the last column. Without written exit rules, every stage ends when someone feels it is time, which is usually when the release calendar says so. The following graphic condenses the exit rules into one line per stage:
The internal pilot is the cheapest place to find what will otherwise reach your call center. It runs in production, with the same tracking customers will get, a few weeks before the first customer sees a passkey.
A group of a few dozen to around a hundred colleagues is usually enough to cover your most important device segments and most questions. The mix matters more than the size:
Run a simple daily loop: collect feedback and tracked errors, ship fixes within days, turn questions into FAQ entries and support scripts, then re-test with the same group.
Internal pilots surface the same categories of issues again and again:
Sophos' CISO playbook reports the same lesson from a workforce rollout: their early pilots underestimated edge cases and the state of personal devices, and support staff training turned out to be key.
Pilots get tricky when the login page is usernameless or uses Conditional UI. An allow-list of pilot accounts works well for passkey creation, because creation happens after login when the account is known. For login, the account is only known after the passkey ceremony. You cannot decide whether to show the passkey button or autofill based on a list you can only check afterwards.
The practical answer is to combine two controls:
Customers who already have a passkey should always be able to log in with it, regardless of the pilot list. Removing passkey login for someone who created a passkey generates exactly the support calls the pilot is meant to prevent.
See how many people actually use passkeys.
After the pilot, the passkey offer goes out by device segment. Passkey behavior differs by operating system, OS version and browser far more than by customer, so a broken segment should only ever affect that segment. Random customer sampling can still be added within a segment.
Segment the passkey offer on signals you already have at the time of the request:
PublicKeyCredential.getClientCapabilities(),
which reports e.g. passkeyPlatformAuthenticator, hybridTransport and
conditionalCreate. MDN lists it as Baseline 2025 newly available, so feature-detect it.
See WebAuthn client capabilities for details.A typical wave order starts with small segments of current iOS, macOS and Android versions, then adds older Apple versions and smaller Android versions, then the largest iOS and Android versions, then Windows, then the long tail. Large segments go live only after smaller segments with the same platform behaved.
Every wave needs a rule that the team agrees on before the rollout starts. Give each new segment two working days of data, then decide:
Plan one day of analysis between waves and start new segments Monday to Thursday, so the team watches the first days. The decision rule fits on one page:
The approach mirrors what app stores offer for app updates. Apple's phased release spreads an update over seven days from 1% to 100% of users with automatic updates and can be paused for up to 30 days in total. A passkey rollout needs the same ability, per device segment and controlled server-side.
Each control needs requirements, a build and tests in staging. Decide the set before testing starts, about two months before the first customer, so the controls exist when you need them.
The global kill switch is deliberately blunt and meant for incidents that affect all segments. If your login runs in native apps or on a hosted page with its own release cycle, an emergency release can take days. A configuration flag takes minutes. For problems in one device segment, the enrollment wave rule is the more precise tool.
Igor Gjorgjioski
Senior Product Lead, VicRoads
We hit 80% mobile passkey activation across 5M+ users without replacing our IDP.
See how VicRoads scaled passkeys to 5M+ users, alongside their existing IDP.
Read the case studyTreat support readiness as a launch gate: no wave goes live until support has the scripts for it. Teams that leave it to the last week pay for it in call volume and in trust. The passkey customer support readiness guide covers agent playbooks, the preparation timeline and support metrics in depth.
The aim is that launch-day calls are mostly hand-holding questions and daily call volume stays within its normal range. If it leaves that range, the next wave waits.
Most passkey support calls fall into a handful of recurring situations. The first answers below draw on Google's passkey help. Apple's passkey security documentation explains how passkeys sync through iCloud Keychain, which matters for the device change cases.
| Customer says | Likely cause | First answer |
|---|---|---|
| "No passkey found" | Saved in another browser or credential manager, or deleted | Sign in another way, then create a passkey here |
| "Works on my phone, not on my laptop" | Passkey stored on the phone or in another ecosystem | Use the QR code with the phone or add a passkey on the laptop |
| "The QR code does nothing" | Bluetooth off or one device offline | Turn on Bluetooth and retry |
| "I got a new phone" | Synced passkeys move within one ecosystem | Sign in to the same Apple or Google account on the new phone. After a switch of ecosystem, sign in another way and create a new passkey |
| "I lost my phone" | Passkey and SMS channel may be lost together | Recovery path, then remove the old passkey |
| "I deleted it, it still shows" | Removed on the website, still in the credential manager | Delete both copies, then create a new one if wanted |
| "Do I still need my password?" | Unclear what happens to the password | One clear answer based on your password policy |
Beyond these situations, many support topics come from flows around the passkey:
signalCurrentUserDetails() from the
WebAuthn Signal API can update it, for the others an FAQ
entry helps.For the wider list of operational risks after launch, see passkey day 2 problems and our passkey troubleshooting guide.

Passkey consulting - rollout materials, ROI tools & expert guidance. Used by PwC and EY.
Communication and rollout waves must be planned together. The most common mistake is a mass campaign that announces passkeys while half of the customer devices still see no offer.
Three rules keep the messaging consistent:
For enrollment prompt wording and placement see our passkey creation best practices.
Measure on three levels and agree on the definitions before the pilot:
Two signals deserve their own alert: more login calls in a wave hold the next wave and a new error class in a segment triggers the rollback rule. Both only work if tracking runs from the pilot onward, which is why the pilot also tests the measurement. The authentication analytics playbook and our overview of passkey analytics cover the metrics in more depth.

Authentication Analytics Whitepaper. Practical guidance, rollout patterns and KPIs for passkey programs.
Product, change and support teams should agree on these answers before the pilot starts:
Corbado Connect adds passkeys on top of your existing identity provider and ships rollout controls like the ones described above: gradual rollout by OS and browser, allow-lists, invitation tokens, sampling and a kill switch without a release. Corbado Observe tracks login success, error classes and fallbacks per device segment from the pilot onward, so every go, hold or rollback decision rests on segment data.
See how Corbado Connect ships this →A passkey rollout is decided in the weeks around launch. Teams that plan stages with exit rules, roll out by device segment, keep a global kill switch for incidents and treat support readiness as a launch gate can reach all ready devices within weeks while support stays calm. Start with the questions in section 9, about two months before the first customer sees a passkey.
Subscribe to our Passkeys Substack for the latest news.
Plan for roughly two to three months from the first staging tests to all ready devices. Staging and rollout controls start about two months before the first customer sees a passkey and an internal pilot runs a few weeks before. Device waves then reach all ready devices within a few weeks if no segment needs a hold.
Support should see every customer-visible change before it goes live, update scripts and the knowledge base before each release, tag passkey and recovery calls from day one and join a short daily stand-up during hypercare. The internal pilot's questions are the best source for the first support FAQ.
A kill switch stops passkey login, passkey creation and passkey management for all users at once through configuration, without a new release. Existing passkeys stay registered and users fall back to their previous login method. It is the fastest way to contain a critical incident, especially when frontends or native apps have long release cycles. Problems in a single device segment are better handled by stopping enrollment for that segment.
Publish the help page and the support FAQ before the first customer sees a passkey offer, start quietly with the offer at login and launch the mass campaign only once the waves cover most customer devices. Claim that passkeys are available on all devices only when every wave is live.
Use two working days of data per device segment. Login success at or above baseline, no new error class and creation and login rates in the expected range mean go. Rates below range or more login calls mean hold. A new unexpected error class or login success below baseline means roll back the segment by stopping new enrollment there.
Related Articles
Table of Contents