Skip to content

Setting up single sign on across event systems without breaking show week

IntegrationsUpdated 2026-08-238 min read

In short

Single sign on centralises authentication so people use one identity across the registration platform, exhibitor portal and reporting tools. It only removes access when local password login is disabled at the same time, and it does not create or delete accounts unless the platform also supports provisioning.

Run this query on Monday morning. How many accounts can sign in to your registration platform right now, and how many of those people still work here.

Most operating teams cannot answer the first half without asking the vendor, and almost nobody can answer the second half at all. Single sign on across event systems is the usual proposal for fixing that, and it is a good proposal, but it fixes less than people expect and it fixes nothing if it is switched on without switching something else off.

The account list nobody has looked at

An event stack accumulates logins in a way a normal corporate application does not, because the people who need access are only partly employees.

The registration platform has accounts for the marketing team, the operations team, the customer service agency, two contractors who built last year's forms, and the show director. The exhibitor portal has accounts for the sales team and for the exhibitor services provider. The lead retrieval system has accounts issued during show week to floor staff. The reporting tool has accounts for everyone who ever asked for a login and never gave it back.

Nobody deliberately built that. It accumulated one favour at a time, usually at speed, usually during the fortnight before a show, and the removal step has no natural trigger. A person leaving your company triggers an offboarding process that covers email and the finance system. It does not cover a portal your show team bought with a card.

What does single sign on actually solve here?

It solves authentication. It puts one identity behind every login screen, so a person proves who they are once against your directory and the platform trusts the result.

That is genuinely valuable. It removes shared passwords, it gives you multi factor authentication everywhere at once without configuring it everywhere, and it means the directory account is the thing that grants access.

What it does not do is create or remove accounts. That is provisioning, a separate protocol with a separate specification, and the two get conflated constantly in vendor sales conversations. RFC 7644, the SCIM protocol document published by Hunt and colleagues at the IETF in 2015, defines the provisioning half: endpoints at /Users and /Groups supporting GET, POST, PUT, PATCH and DELETE, a /ServiceProviderConfig endpoint that advertises what the provider supports, pagination through a 1-based startIndex and a count, and PATCH operations of add, remove and replace for changing an existing record without replacing it whole.

The companion document, RFC 7643, defines what a user record holds: an id assigned by the provider, an externalId that the client controls so you can key records to your own directory, and an active flag described as a boolean where "a value of false implies that the user's account has been suspended" (Hunt and colleagues, 2015). That flag is the whole point. It gives you a machine readable way to assert that someone should no longer get in, and a place to check whether the assertion landed.

The same document defines an enterprise extension carrying employeeNumber, department, division, organisation and a manager reference (Hunt and colleagues, 2015). Those fields are why provisioning is worth more to an event business than it first appears. If department travels with the account, the access review stops being a name matching exercise and becomes a question you can answer in one query: which accounts on the exhibitor portal belong to people outside the sales organisation.

The setting that decides whether any of this works

Here is the part I would put in bold if the house style allowed it, so instead I will just say it plainly and early.

Turning on single sign on while leaving local password login enabled removes almost none of the risk. The accounts created before the rollout still exist, still have passwords, and still work. Your directory now controls the front door while the side door stays open, and because the login page shows a company branded button, everybody believes the problem is solved.

Every platform worth using has a setting that enforces the identity provider as the only authentication route, sometimes called SSO enforcement or disabling local login. Find it, and find out whether it applies to every account or only to accounts created after a date. Ask specifically about service accounts and API users, because those often need an exception, and an exception you did not know about is a password nobody rotates. Machine credentials are a different problem with a different rhythm, handled in rotating integration credentials.

Counting the problem on your own stack

Numbers make this an operations conversation rather than a security one, and operations conversations get budget in event businesses.

Take six systems: registration, exhibitor portal, lead retrieval, the association management system, the email platform and the reporting tool. Export the user list from each. Say they hold 84, 31, 46, 22, 19 and 63 accounts, which totals 265 logins across a team of maybe 40 people. That ratio is normal, because the same person appears in four systems and the leavers never got removed.

Now cross the combined list against your current staff directory. If 19 of the distinct people on it have left in the past 18 months, and each of them averaged 3.2 accounts, that is 61 live logins belonging to people who are gone, about 23 per cent of the total.

Then price the manual alternative. Six systems means six removal tickets per leaver. Forty leavers and joiners a year across a show team is 240 tickets, and at 15 minutes each that is 60 hours of somebody's year spent on account admin that a provisioning protocol would do for nothing. The 60 hours is the argument that lands with a finance lead. The 61 stale logins is the argument that lands with whoever signs off on data protection.

What do you do when the vendor supports neither?

Most event platforms will not have a /ServiceProviderConfig endpoint to query. Assume the answer is no until you see documentation.

Three things still work. The first is a quarterly access review that runs off an export rather than a memory: pull the user list from every system on the same day, diff it against the staff directory, and delete what does not match. Put the date in the calendar rather than in a plan, because a review with no date never happens twice.

The second is to make the request at renewal. Vendors respond to a written requirement in a procurement document far better than to a support ticket, and single sign on support is now common enough that asking for it is unremarkable. The wider set of things worth negotiating at the same moment sits in writing export and access rights into the contract.

The third is to reduce the number of accounts that need managing. Shared read-only dashboards, exports mailed on a schedule, and a single service account for the agency all cut the login count without any protocol at all. Fewer accounts is a smaller version of the same problem and it is available immediately.

Collapsing many people onto one service account has a cost worth pricing before you do it. Microsoft notes that portal applications typically send requests through a service principal account, and because its service protection limits are applied per user, a portal's whole traffic load lands on that single account's allowance (Microsoft, 2026). The same concentration happens to your audit trail. One shared account means every action in the platform is attributed to the account rather than to a person, which is fine for a scheduled export and unacceptable for anything that can delete a registration.

Why the rollout date matters more than the design

Identity plumbing fails in a specific way: everything works until someone needs to log in urgently, and then it does not.

Show week is the worst possible moment. Floor staff arrive on the Sunday, need lead retrieval access on the Monday, and half of them are agency people who are not in your directory. If single sign on went live the previous Thursday, the answer to "I cannot log in" is now a directory question routed to an IT team who have never heard of the show. Roll out in the quiet quarter, keep a documented break glass account with a rotation date attached, and record who holds it in the integration runbook.

Test the leaver path before you need it, too. Disable a test directory account, then try to sign in to each platform. If any of them still lets you in, you have found the local login exception, and you have found it on a Tuesday in February rather than during an audit.

There is a sequencing argument here that I would defend against the usual order of work. Most projects roll out authentication first because it demonstrates progress, and leave deprovisioning for phase two, which arrives never. The reverse order is better on an event stack. A quarterly export and delete costs an afternoon and removes the 61 stale logins immediately. Single sign on costs a project and removes none of them until the local login setting is also found and changed. Do the cheap thing that reduces the count first, then do the protocol work that stops the count growing.

Where this stops

Single sign on controls who gets in. It says very little about what they can do once inside, and event platforms have famously coarse permission models.

A registration platform that offers administrator or read-only, with nothing between, will hand the customer service agency the ability to delete registrations because they need to edit them. Provisioning the account correctly does nothing about that. Ask about role granularity separately, and if the platform only has two roles, the mitigation is procedural rather than technical.

The other honest limit is that the directory has to be right. Provisioning propagates whatever your source of truth says, and if contractors are not in the directory, or leave dates are entered a month late, you have automated the delivery of stale information across six systems instead of one. Fix the directory first, then connect it to the event stack.

Start this week with the count. Export the user list from your registration platform, put it next to the current staff list, and mark the names you do not recognise. That number, on one show, is the whole business case.

Questions people ask about single sign on event systems

What is the difference between single sign on and provisioning?
Single sign on handles authentication, deciding whether the person at the login screen is who they claim to be. Provisioning handles the account itself, creating it, updating it and switching it off. RFC 7644 defines the provisioning half as a protocol with /Users and /Groups endpoints. A platform can support one without the other, and many event platforms support neither.
Does turning on single sign on remove access for people who have left?
Only if local password login is disabled at the same time. If the platform still accepts a username and password alongside the identity provider, an account created before the rollout keeps working after the person's directory account is switched off. Check for a setting that enforces the identity provider as the sole authentication route.
What does SCIM give you that a spreadsheet does not?
A defined contract. RFC 7643 specifies an active flag on the user record where false implies the account has been suspended, a userName that must be unique across the service provider, and an externalId the client controls. That gives you a machine readable way to assert who should have access, rather than a list somebody maintains by hand.

Related reading

All integrations articles