Guides · Data and access
Single sign-on: one login, and one place to switch it off
With single sign-on, people use the work account they already have. When they leave, you close that one account and their access to everything goes with it. Every tool that keeps its own passwords is another list someone has to remember to update.
The short version
- With single sign-on the tool holds no password. It asks your company directory who the person is and trusts the answer.
- The main gain isn’t convenience. It’s that closing one account ends access everywhere.
- Let the groups you already keep in the directory decide who gets what, so there’s no second list of people to keep in sync.
- Before you buy, ask which standard it uses, whether password login can be turned off, how long a session lasts after the account is closed, and whether single sign-on costs extra.
This isn’t only IT’s concern. If you build something for your colleagues, it gets used more when people can sign in with the account they already have, and nobody has to remember to take them off it when they leave.
Count the internal tools in your company that have their own login screen. The booking system, the supplier register, the form a department built, the reporting tool someone paid for with a company card. Each has its own list of users and passwords, and each list has to be updated by hand whenever someone joins, changes role or leaves.
Most of those lists are a little wrong right now. A contractor from last spring is still on one. A colleague who moved department kept access to another. With single sign-on you stop keeping those lists at all.
What single sign-on is
With single sign-on the tool never sees a password. When someone opens it, they’re sent to the company’s identity provider (Microsoft Entra ID, Google, Okta or similar). It checks who they are, in whatever way the company has set up, and sends them back with a signed message: this is Kari, and she’s in these groups. The tool trusts that message and lets her in.
Two standards cover almost everything. OpenID Connect is the newer one and what most modern tools use. SAML is older and still common in large enterprise software. Either is fine. What you want to avoid is a login scheme the vendor made up.
- It isn’t a password manager. A password manager remembers lots of passwords for you. With single sign-on there’s only one, and only your directory knows it.
- It isn’t “the same password everywhere”. Syncing one password into ten systems leaves ten copies of it in ten places. With single sign-on there are none.
- It isn’t a personal “Sign in with Google”. A private Gmail account signing in to a work tool is still a private account. It has to be the company’s directory, or the company can’t close it.
What separate passwords cost
Single sign-on is usually sold as convenience: one login instead of twelve. That’s the least important part of it. The costs of separate passwords show up elsewhere.
- People who leave keep access. When someone leaves, their company account is closed. A tool with its own user list doesn’t notice, unless someone remembers it exists and has the rights to remove them.
- Your security rules stop at the door. Two-factor sign-in, blocked countries, company devices only: none of what your directory enforces applies to a tool with its own password.
- Passwords get reused. Nobody makes up a strong new password for the canteen booking system. They reuse one they already have, which is then only as safe as the least careful site they’ve used it on.
- Nobody knows who has access. When an auditor or a customer asks who can get into a system, the answer is spread over a dozen admin screens, and some of them belong to people who’ve left.
The day someone leaves
This is what single sign-on is for. Disable the account in the directory, and the next time that person tries to open a connected tool, the identity provider says no. There’s no list to work through, so nothing on it can be forgotten.
There are a few gaps, and that’s where to look.
- Sessions that are already open. Someone who signed in this morning may stay signed in until the session runs out, even after the account is disabled. Find out how long that is in each tool. Minutes or hours is fine. Weeks is not.
- Logins that go round the side. Many tools keep a password login next to single sign-on, for the first administrator or out of habit. An account made there sits outside your directory and survives all of the above.
- Keys and tokens. Integration keys and personal access tokens are often created by one person and keep working after they’ve gone. Give them to a service with a named owner instead.
Let the directory groups decide
Single sign-on answers “who is this?”. The next question is “what may they do?”, and the easy answer is to give each tool its own list again: Kari is an editor here and an administrator there. That brings back the problem single sign-on just solved.
It works better when the groups you already keep decide. The finance group can approve invoices. The warehouse group can register deliveries. When Kari moves from finance to the warehouse, someone changes her groups in the directory, which they’d do anyway, and every tool follows.
Groups decide which doors a person can open. Which records are theirs once they’re inside is a separate question, answered in the data. The guide on row-level security covers that.
What to ask before a tool gets users
Ask these of any tool, bought or built, before anyone signs in to it. If the vendor has thought about it, each one has a short answer.
- Which standard? OpenID Connect or SAML, and does it work with the identity provider you use? “We support Microsoft” doesn’t tell you much until you know which part of Microsoft.
- Is it included? Single sign-on is often only on the most expensive plan. Find out before the pilot, not at renewal. Otherwise the cheap plan with separate passwords is the one people keep using.
- Can password login be turned off? If it can’t, single sign-on is optional and the side door stays open.
- Are groups read at every sign-in? Does access follow the groups in your directory as they are today, or were they copied once and left to drift?
- How long does a session last after the account is closed? Ask for the number, and whether an administrator can end a session early.
- Where are sign-ins recorded? Your identity provider records that someone signed in. The tool should record what they did after that, and you should be able to read it.
How it works on Softeria
Apps built with Softeria sign people in with the accounts they already have: Microsoft Entra ID, Google Cloud Identity, or any OpenID Connect provider that follows the standard, such as Okta, Auth0 or Keycloak. We don’t hand out new passwords. Your directory groups decide who can open each published app, and when someone is removed from your directory, their access ends.
Whatever you end up using, the lesson is the same. A tool that keeps its own list of people is one someone will forget on the day it matters. If the directory is the only list, closing an account when someone leaves is one step.