The quotation says one number. The cost is decided by five things that are not that number: how a seat is counted, what integration takes, who does the training, what happens at renewal, and what it costs to leave.
Every institution that has been surprised by a software cost was surprised by one of those five, not by the headline rate. What follows is what an institutional licence should include, what to refuse, and the counting rules to settle before comparing any two quotations.
Nine things the licence should include
- A written seat or volume definition, with the counting rule. Not “per student” — which students, counted when, and what happens when the number changes mid-year.
- SSO and LMS integration at no separate fee. Integration is not a professional service; it is the product working in a university. Ask which LTI Advantage services are required rather than supported — the selection is a data-minimisation decision, as set out in our SSO and LMS integration guide.
- Implementation delivered inside your academic calendar, with a named date outside an assessment period.
- Training for staff, included and repeatable. One session at go-live serves the people who happened to be free that Tuesday. You need a repeatable session for each intake.
- A data processing agreement with a current sub-processor list, plus separate answers on retention and on training use — they are two different permissions.
- Export in a documented, non-proprietary format, available throughout the term and not only at exit.
- A support SLA with named response times and a defined escalation path to a person, not a queue.
- Accessibility conformance documentation, because you will be asked for it and retrofitting the question after signature is expensive.
- Price protection across the first renewal. A three-year price with an uncapped year-four is a one-year price with a delay built in.
Item nine is worth pressing on, because it is where the negotiating leverage actually sits. At initial purchase you have alternatives and the vendor knows it. At first renewal you have a deployed integration, trained staff, accumulated content and a migration cost — which is precisely when an uncapped increase lands. Negotiating year four during year zero costs nothing and is the single most valuable clause most institutions do not ask for.
Five things to refuse
- An implementation fee for standard integration. Custom development is chargeable; SSO and a standards-based LMS connection are not custom development.
- A seat count you cannot audit. If you cannot reconcile the vendor’s count against your own student record system, you cannot dispute an invoice.
- Automatic true-up without notice. Growth clauses are reasonable; discovering one applied at renewal is not. Require notification at the threshold, before the charge.
- Data export as a paid service at termination. Charging for exit is charging for the option to leave, and it should be priced at zero.
- Bundling you did not ask for. If the quotation includes products your policy would not permit — an evasion tool in a consumer catalogue, for instance — ask for them to be excluded in writing and keep the reply.
The counting rules that decide the real cost
Four questions, and the answers vary enough between vendors to reverse a comparison:
| Question | Why it moves the number |
|---|---|
| Does an interrupting student consume a seat while interrupted? | Interruption is normal in graduate education. If seats are held, your effective population is larger than your active one |
| Does a returning student consume a new seat? | Combined with the row above, this can double-count the same person |
| Do resubmissions and drafts count as separate submissions? | Formative use during supervision multiplies volume against an end-of-process assumption |
| Do appendices and datasets count towards a volume allowance? | Appendix-heavy theses consume allowance disproportionately |
Ask all four in writing, and ask for a worked example using your own numbers. A vendor that will model your population is a vendor that expects to be measured; one that will not is telling you something.
There is a second-order effect worth anticipating. A licence that charges per submission rather than per person quietly discourages exactly the behaviour you want — formative use during supervision, where feedback is cheap and intervention is early. If the commercial model penalises early drafts, the tool will be used only at the end, and you will have bought an end-of-process product at a whole-process price. Check that the pricing shape matches the pedagogical use you intend, not just the budget you have.
Total cost is not the licence
A realistic institutional model has four lines, and the first is usually the smallest:
- Licence. The quoted figure, over the full term including renewal.
- Implementation. IAM, LMS and information security effort, in days. This is real money even when nobody invoices for it.
- Governance. The data protection review, the accessibility check, the committee cycles.
- Adoption. Training, documentation, and the support load in the first two terms.
Line four is the one most often omitted and the one most likely to determine whether the deployment succeeds. A tool nobody was trained on has a licence cost and no benefit, which is the worst available outcome.
Why the free departmental pilot is the better first step
Not because it is cheaper — although it is — but because of what it produces and what it avoids.
What it produces: real usage data from your own population, an integration you have actually built rather than been shown, staff feedback against criteria you set in advance, and a defensible answer to “how do we know this works here”. Those are the four things a business case needs and none of them comes out of a tender exercise.
What it avoids: committing budget before you know the seat definition matters, and committing a governance cycle to a specification written from last cycle’s specification. A pilot that commits no spend clears approval routes that a purchase order does not.
The one condition: agree the success criteria before it starts. A pilot without pre-agreed criteria produces an argument about interpretation rather than a decision. The design, the criteria and the artefacts are in our guide to running a departmental pilot of an AI writing tool, and the parallel data protection sequence is in our DPIA guide.
One honest note on price transparency
No vendor in this category publishes an institutional price as static text, ourselves included, because institutional pricing genuinely depends on population, scope and term. That is a reasonable position and it has an obvious consequence: every comparison you make will be between quotations you obtained, on assumptions you specified. Write the assumptions down first and send the same set to every vendor, or you will be comparing answers to different questions. The question set is in our procurement question bank, and the market-level checks to run at renewal are in what changed in academic integrity platforms this year.
Request an institutional evaluation and we will scope a free departmental pilot with the seat definition and the exit terms stated up front.
Frequently asked questions
What should a university site licence include?
A written seat definition and counting rule, integration without a separate fee, implementation inside the academic calendar, repeatable training, a data processing agreement with sub-processors, documented export, a support SLA, accessibility documentation and price protection across the first renewal.
Why does the seat definition matter more than the price?
Because it determines how many seats you consume. Interruptions, returns, resubmissions and appendices can each multiply consumption at the same rate.
Should we pay an implementation fee?
Not for standard SSO and LMS integration. Custom development is a different matter.
What is a true-up clause?
A mechanism charging for growth above a threshold. Reasonable in principle; require notification at the threshold rather than a charge at renewal.
Should exit ever cost money?
No. Charging for data export at termination prices the option to leave, and it should be zero.
When is our negotiating leverage highest?
Before the first signature. At renewal you hold a deployed integration and a migration cost, which is exactly when an uncapped increase arrives — so negotiate the renewal price at the start.
Does per-submission pricing cause a problem?
It can. Charging per submission discourages formative use during supervision, which is where the tool does the most good. Match the pricing shape to the intended use.
What is usually missing from the cost model?
Adoption — training, documentation and first-term support load. It frequently exceeds the licence in year one.
Do vendors publish institutional prices?
Generally not, because pricing depends on population and scope. Send identical written assumptions to every vendor so the quotations are comparable.
Is a pilot better than a tender?
For establishing fit, yes. A tender is the right instrument once you know what you are specifying.
What does a free departmental pilot commit us to?
Coordination time and pre-agreed success criteria. No budget and no procurement cycle.
Can a pilot run alongside our existing contracts?
Yes. Support and screening are complements, so a pilot does not require displacing a detection contract.
What if the vendor will not state the seat rule in writing?
Treat that as the answer and record it. An unwritten counting rule becomes the vendor’s counting rule at invoice time.