About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
Direct answer: Before you connect a customer’s Microsoft Entra ID, Active Directory or other IdP to your hosted application tenant, complete an identity discovery: authoritative source of users, group-to-role mapping, MFA and conditional access expectations, lifecycle (joiners/leavers), break-glass admins, and what happens when federation is offline—then implement SAML or OpenID Connect for sign-in and, where your product supports it, SCIM for automated provisioning; the two layers are complementary, not interchangeable substitutes.
Why it matters
Customers do not want another password store. They want leavers disabled in Entra to lose application access the same afternoon. Botched federation shows up as duplicate accounts, blank roles after SSO login, or worse—every SSO user becomes admin because a claim mapping default was wrong.
This is onboarding identity design for ISVs, not a guide to repairing Entra Connect sync errors or disaster recovery for hybrid identity. The hosting provider may manage servers; the ISV still owns application session creation, authorisation and audit.
How it works
Discovery worksheet (minimum)
- IdP vendor and edition (Entra ID P1/P2, on-premises AD, Okta, etc.).
- Required identifiers: email,
employeeId, UPN format. - Groups or roles that must map to in-app permissions (finance approver, site manager).
- MFA: enforced at IdP only, or also step-up inside app for sensitive actions?
- Guest users / contractors: in scope for SSO?
- Maximum acceptable session length and logout behaviour on shared PCs.
- Break-glass local accounts: allowed? how many? monitored?
Patterns
| Pattern | When it fits | Hosted-app notes |
|---|---|---|
| SAML/OIDC login | Browser SSO standard | Redirect URIs must match prod hostname and any custom domain |
| SCIM provisioning | Automated user lifecycle | The customer identity provider calls your application’s SCIM endpoint over HTTPS—document inbound firewall and TLS requirements |
| LDAP bind (legacy) | Old LOB constraints | Prefer migration to modern federation; TLS and IP allow lists |
| Local accounts only | Tiny customers | Still plan MFA and export for exit |
Claim mapping
Translate IdP groups to application roles in a table reviewed by customer IT. Test negative cases: user in no groups, user in conflicting groups, renamed group. Log effective roles at first login during UAT.
Federation reliability
If the IdP is unreachable during new sign-in, the secure default is to fail closed—do not mint fresh sessions without successful authentication. Cached session grace means users who already hold a valid, unexpired application session may keep working until that session ends; it is not permission to extend expired authentication, ignore logout policy or treat an IdP outage as “open access.” Document both behaviours for helpdesk scripts.
Hosted infrastructure touchpoints
Custom domains affect redirect URIs. Outbound calls from your app to customer webhooks are separate from inbound federation—do not conflate firewall tickets.
What to choose and why
Enterprises: SSO + SCIM with staged UAT tenants. SMBs: OIDC with just-in-time provisioning if volume is low. Regulated: no JIT admin—provision privileged roles through workflow.
Always run a pilot cohort with real customer staff before company-wide SSO cutover.
Read next
Related articles you may want next.
- HostingWhat partners need to see that end customers should notGive partners the oversight they need while protecting each customer’s data through tenant boundaries, limited permissions and clear support access rules.Read article
- HostingWhite-label and branding on a hosted ISV product: what usually can changePlan hosted product branding across logos, domains, email and sign-in screens, with clear limits and checks for authentication and certificate changes.Read article
- HostingFranchise and multi-company groups on one hosted ISV platformDesign access and data boundaries for franchise and multi-company customers, including group reporting, site permissions and changes in ownership.Read article
Risks and limits
- Infinite redirect loops when HTTP/HTTPS hostnames mismatch custom branding.
- Offboarding: disabling a user in Entra must trigger session revocation in your app (or short-lived tokens plus explicit logout), not rely on token validation alone—disabled accounts may stay usable until refresh or expiry.
- Stale sessions after IdP disable if the app does not revalidate tokens or maintain a revocation list.
- SCIM delete handling: soft-disable vs hard-delete affects audit and billing seat counts.
- Not a step-by-step Entra wizard; commands not validated in lab.
Sources
- Develop and plan provisioning for a SCIM endpoint — Microsoft Learn; accessed 2026-11-16 — identity provider calls the application SCIM endpoint.
- Microsoft identity platform and OAuth 2.0 OpenID Connect — Microsoft Learn; accessed 2026-11-16 — browser SSO and claims.
- NIST SP 800-63B-4, Digital Identity Guidelines — NIST; current edition superseding SP 800-63-3 (August 2025); accessed 2026-11-16 — authentication assurance orientation for sensitive actions.
Related reading
- What partners need to see that end customers should not
- White-label and branding on a hosted ISV product
- Franchise and multi-company groups on one hosted ISV platform
- Handling change requests after an ISV customer is live
- Software vendor hosting (related service context)
Learning notes
- Identity discovery prevents more go-live failures than server sizing.
- Role mapping belongs in a signed table, not improvised in prod.
- Custom domains and SSO redirects must be tested together.
- Break-glass accounts need monitoring, not shame.
- SCIM and SSO are related but not the same project.
This article was originally published on .