Start with the service people must still be able to use
Picture a county replacing the sign-in system for its permit portal. An applicant who signed in on Friday should be able to return on Monday, open the same application and submit it. An inspector should be able to see the cases assigned to them, and a former employee’s access should stay revoked. Those are three separate checks. A demo of the new login page proves only that someone can sign in.
Before opening a search, the program owner should name the services and deadlines the change must protect. Ask application owners to identify a real task for each group affected, including people with limited access to a second device and users with delegated permissions. Pick a pilot that covers meaningful differences in access without putting the highest-stakes service in the first wave. Record what would cause the agency to pause that wave.
Give the account mapping its own owner
List each application, the identity provider it trusts, the field it uses to identify a returning person, and the place where its permissions live. Include staff and public accounts where both are in scope. Give each application owner responsibility for confirming the mapping. If an application has two records with a shared or outdated email address, route the case for authorized review before the cutover. A convenient match can attach the wrong person to a case file.
NIST’s federation guidance separates the account maintained by the identity provider from the account maintained by the application. These accounts can have separate life cycles, and the application can hold its own access rights. Ask the engineer how they will test the sign-in, the account match and the permissions separately. For a record that cannot be matched safely, the program owner needs a support route and a way to track resolution. NIST SP 800-63C-4, Common Federation Requirements.
Staff the failure paths as well as the happy path
Put an identity engineer in charge of federation and configuration. Assign an application engineer to mapping, permissions and app-level testing. Give a tester responsibility for showing what happens to a returning applicant, a suspended inspector and a user who loses their sign-in device. The support lead should rehearse the actual recovery handoff with the security owner. NIST treats account recovery as a distinct process for regaining access after a person loses control of their authenticators. NIST SP 800-63B-4, Authentication and Authenticator Management.
A smaller agency can combine roles when the workload allows. Keep someone outside the configuration work responsible for checking access results. In the interview, give each candidate the same fictional case. Ask what they would test if sign-in succeeds but a returning applicant sees an empty dashboard. Score whether they distinguish an identity mismatch from an authorization problem, protect the records involved and say what evidence would stop the release. The work sample can use synthetic accounts and requires no production access.
Give the agency a cutover it can run
The program owner should make the go or no-go decision after the application owner, security owner and service desk lead review the pilot evidence. Set the decision before the first wave. For example, a confirmed cross-account exposure stops the rollout immediately. For login failures, agree on a threshold and a monitoring window based on normal traffic and the support team’s capacity. Rehearse returning traffic to the former sign-in route. If the pilot creates account links or changes permissions, assign someone to reconcile those changes during a rollback.
This applies the Vitaver Playbook’s Mission First principle. The agency owns the result and should be able to verify it with its own people. Before the contractor leaves, have an agency operator demonstrate how to diagnose a failed account match, hand a recovery case to the right team and find the tested rollback procedure. Decide who reviews unresolved exceptions and when. Then the contract can end while the agency retains the ability to operate the service.