Testing for vendor lock in before you renew an event platform
Test for vendor lock in by requesting a full export a year before renewal and rebuilding your last post show report from it alone. Count the figures you can reproduce. The gap between that count and the report's total is the switching cost, measured rather than argued about.
Every renewal conversation about an event platform contains the same unspoken assumption on both sides of the table, which is that you are not really going anywhere. The vendor knows it. Your team suspects it. Nobody has checked.
Vendor lock in is measurable, and the measurement takes about three days. Run it a year before renewal and the negotiation changes character, because you stop arguing about whether switching would be hard and start discussing a number you both agree on.
What actually holds you, once you look
The instinct is to think about features. Feature comparison is the thing procurement asks for, and it is close to irrelevant, because a competitor in an active sales cycle will build or fake anything on the list.
What holds you is identifiers and history.
A new registration platform starts with no priority points, so the rule that decides which exhibitor picks booth space first has no inputs. It starts with no booth lineage, so nobody can answer which stand a company occupied in 2019 or how their footprint changed. It starts with no cross year person keys, so every returning attendee looks new and your retention rate collapses to zero on the first edition after the move. None of those are features. They are accumulated state, and the vendor holds the only copy.
The second thing that holds you is the shape of everything downstream. Reports, warehouse tables, the sales team's saved views, the finance reconciliation that keys off the platform's invoice numbers. A migration touches every one of them, and the cost sits in your organisation rather than in the vendor's fee.
How do you run an exit test?
Request a full export as though you were leaving, and then use it as though you had left. Both halves matter and most teams only do the first.
Send the request through the same channel you would use in a real termination, which usually means a support ticket rather than a call with the account manager. Time it. The elapsed days between request and delivery is a data point, and a vendor who takes six weeks to produce a routine export will not produce one in 30 days under duress.
When the files arrive, do not open them and nod. Load them into a scratch database with no access to the live platform, disconnect from the vendor's API entirely for the duration, and then attempt one specific task: rebuild last year's post show report.
That task is the right test because the report is already agreed. Its numbers were published, somebody signed them off, and every figure in it is one the business actually uses. Reproducing it from an export is a pass or fail question with no room for optimism.
Give the rebuild to someone who did not build the original pipeline. A person who knows where the awkward joins are will unconsciously repair the export with knowledge that lives in their head, and that knowledge leaves the company with them. Somebody working only from the files and the published report will find the gaps that a migration would actually hit.
Rebuilding the report, and counting what is missing
Work it as arithmetic rather than as an impression.
Say the post show report carries 14 headline figures: total registrations, verified attendance, attendance rate, new versus returning share, international share, top ten source countries, exhibitor count, net square metres sold, rebooking rate at close, average booth size, session attendance, meetings booked, lead scans per exhibitor, and satisfaction score.
Now rebuild from the export alone. Registrations, attendance, attendance rate, international share, exhibitor count, net square metres, average booth size, session attendance and lead scans all come out, because the underlying rows were in the files. That is nine.
Five do not. New versus returning needs cross year person keys that the export did not include, only the current edition's records. Rebooking rate at close needs the booth assignment history for the following edition, which lives in a module the export did not cover. Meetings booked came from the matchmaking product, exported separately or not at all. Top ten source countries turns out to be derived from an enrichment the platform ran, and the export contains the derived value for current records with no way to recompute it for anything new. Satisfaction score lives in the survey tool.
Nine of fourteen is 64 per cent, and the five missing figures are the ones the commercial team cares most about. That is your lock in, stated as a number, and it took three days.
Do the timing arithmetic alongside it. If the export took 19 days to arrive and the load and rebuild took three, a real switch needs a minimum of 22 days before anyone starts configuring the replacement. Set that against the transitional periods in the switching rules below and you can see immediately whether the legal timetable is comfortable or tight.
What the switching rules reach, and what they miss
The regulatory position improved recently and it is worth knowing precisely, because it changes the mechanics without changing the substance.
Regulation (EU) 2023/2854, the Data Act, sets contractual terms for switching between data processing services. Article 25 caps the notice period for initiating a switch at two months and sets a mandatory maximum transitional period of 30 calendar days, extendable where the switch is technically unfeasible in that time, with the provider obliged to propose an alternative period (EU, 2023). Article 29 handles cost: from 11 January 2024 providers may charge reduced switching charges that do not exceed the costs directly linked to the switching process, and from 12 January 2027 switching charges shall be zero.
Regulators elsewhere have reached similar conclusions about the same behaviour in adjacent markets. The UK Competition and Markets Authority published the final decision of its cloud services market investigation on 31 July 2025, finding an adverse effect on competition in public cloud infrastructure services and recommending that strategic market status investigations into Amazon Web Services and Microsoft be prioritised (CMA, 2025).
Both of those reduce the friction and the fees. Neither recreates a priority point. The rules make it cheaper and faster to move whatever the platform can hand over, and the whole difficulty in an event business is the part it cannot. What the contract itself needs to say to close that gap is a separate exercise, worked through in writing data export rights into the contract.
Scoring the result, and what to do with it
Three numbers come out of the test: the days to receive the export, the share of the report you could rebuild, and the count of objects that came out with usable keys.
If you rebuilt 90 per cent or more, you have a normal commercial relationship and can negotiate on price. If you rebuilt somewhere in the middle, as most portfolios will, you have a specific list of missing objects and a year to fix it, which is enough time to start extracting those objects yourself on a schedule. If you rebuilt less than half, switching is not currently an option and the renewal should be treated as such internally, whatever position you take in the room.
The list of missing objects is the most useful output. It converts a vague anxiety into six or seven line items, each of which is either an export request, a contract clause, or a monthly extract you start running yourself. Record who owns each one in the integration runbook so the work survives the person who ran the test.
When is lock in the right trade?
An unfashionable position, and I hold it. Some lock in is worth accepting.
A platform that holds twelve years of booth lineage and computes priority points correctly is doing something genuinely hard, and the accumulated state is exactly why it is valuable. Demanding perfect portability from it is demanding that it be less useful. The mistake is not depending on a platform. The mistake is depending on one without knowing the size of the dependency and without a copy of the state in your own storage.
That is the practical resolution. Accept the dependency, take a monthly extract of the objects that constitute the state, and store them somewhere the vendor does not control. You have then bought most of the exit optionality without a migration, and the annual exit test becomes a check on the extract rather than a fire drill. How far back the extract can reach is its own constraint, covered in the post on API history limits.
Where this stops
The exit test measures whether you could leave. It says nothing about whether the destination is any better, and a portfolio that migrates to a platform with the same export weaknesses has spent a year to end up in the same position facing a different logo.
The test also assumes the export you receive during a friendly year resembles the one you would receive during an unfriendly one. Usually it does, because the same tooling generates both. Occasionally it does not, and the only protection there is a contract term rather than a rehearsal.
The last limit is a human one. An exit test finds problems that are expensive to fix, and the natural organisational response is to file the result. If nobody owns the follow up list by name, the test becomes an annual ritual that changes nothing about the platform relationship.
Start this week by sending one support ticket asking for a full export of every object, in a documented format, with keys included. The clock starts when you send it, and how long the reply takes is the first number in your test.
Questions people ask about vendor lock in
- What actually creates lock in on an event platform?
- Identifiers and history, far more than features. A replacement platform starts with no priority points, no booth lineage across editions, and no person keys that link a 2021 registrant to a 2026 one. Features can be matched by a competitor in a sales cycle. Twelve years of accumulated identifiers cannot be recreated by anybody.
- When should you run an exit test?
- About a year before renewal, while you still have time to act on the result and while the vendor still has a commercial reason to help. Running it during a termination gives you the same information at the moment it is least useful. A year of notice also lets you start collecting the missing objects yourself if the export turns out to be thin.
- Do the EU switching rules solve vendor lock in?
- They reduce the cost of the mechanics. Regulation (EU) 2023/2854 sets a maximum notice period of two months for initiating a switch and a mandatory maximum transitional period of 30 calendar days, and Article 29 requires switching charges to be zero from 12 January 2027. None of that recreates booth lineage or priority points that the platform never exported.
Related reading
- Finding the API history limits in your event systems before a five year backfill
- Writing data export rights into an event platform contract
- What belongs in an integration runbook for an event tech stack