Five Hundred Dissertations Land in the Same Eight Weeks. Nobody Is Staffed for That.

Your taught masters cohort has one submission date. Not a spread of dates across a term, not a rolling deadline — one date, usually in late summer, for every student on every programme in the faculty. Which means the writing happens in the same eight weeks, the supervision requests arrive in the same eight weeks, the writing centre bookings arrive in the same eight weeks, and the panicked emails about reference formatting arrive in the final ten days.

This is not a busy period. A busy period is something you staff for. This is a cliff, and the establishment you have was sized for term-time demand, when the same staff are teaching. The mismatch is structural, it repeats every year, and it is almost never named as its own operational problem — it gets absorbed into general workload complaints and disappears.

Why the peak is worse than your annual figures suggest

Look at any support service’s annual report and you will see a utilisation figure that sounds manageable. Sixty per cent of capacity used. Comfortable, a finance committee concludes, and there may be room to trim.

That number is an average across a year in which the service ran at 30% for thirty weeks and at complete saturation with a two-week waiting list for eight. The annual mean is arithmetically correct and operationally meaningless, and it is the single most damaging figure in most writing support business cases. The measurement fix — reporting by teaching week and quoting a denominator of students who actually have an extended writing task in the period — is set out in our analysis of the writing centre utilisation data nobody publishes. Until you report the peak, you will keep being asked to justify capacity you have already been shown to be short of.

The peak also has a shape that makes it worse than a simple volume spike. Demand is not evenly spread across the eight weeks; it is heavily back-loaded into the final fortnight, when the work arriving is the least suitable for the help available. A student asking for structural feedback in week two can be helped. A student arriving four days before submission with an unformatted reference list and a chapter that does not connect to their method cannot be helped by a fifty-minute appointment, and everyone in the room knows it.

The three mitigations that do not work

Hiring for the peak

The obvious answer, and the one that fails first. The peak is eight weeks of a fifty-two week year, so there is no post to advertise. What you can hire is casual or hourly-paid staff, which means people who do not know your regulations, your submission requirements, your programme conventions or your students — and who need supervising by the permanent staff who are already at capacity. Casual capacity added at the peak frequently reduces net throughput in the first fortnight.

Staggering the deadline

Attempted regularly, abandoned regularly. The submission date is anchored by examination board dates, which are anchored by conferral dates, which carry visa positions, graduate job start dates and funding end dates. Moving one programme’s deadline by three weeks means moving that programme’s board, and boards are quorate committees of the same academics who are supervising. The constraint is real and it is not a failure of imagination on anyone’s part.

Rationing appointments

One appointment per student per project, thirty minutes, no repeat bookings. It preserves the service’s ability to say yes to everyone, and it converts substantive support into triage. It also produces a specific equity outcome: students who can afford to buy help privately do, and students who cannot get thirty minutes. The substitution is real and predictable, and the cohort most affected is the one that arrives with the highest need — the pattern documented in why your international cohort needs the most writing support and is offered the least.

What is actually consuming the peak

Sit in a writing centre or a supervisor’s inbox during dissertation season and sort the requests. Four categories account for the overwhelming majority of contacts, and three of them do not require academic judgement at all.

1. Reference formatting and citation consistency. Mixed styles, incomplete entries, in-text citations that do not match the list. This is mechanical, high-volume, and it consumes appointment slots and supervisor review time in roughly equal measure.
2. Structural and formatting compliance. Heading hierarchies, figure and table numbering, appendix conventions, word count boundaries, the specific submission template your programme mandates. Entirely rule-based, and entirely capable of being checked before a human sees the document.
3. “Is this the right structure?” A student who has written five thousand words and does not know whether their chapters are in a defensible order. This one genuinely needs expertise — but it needs it at week two, not week seven, and it arrives at week seven because nothing prompted it earlier.
4. Argument, method and interpretation. The reason academic staff exist. Perpetually the smallest share of contact time during the peak, squeezed out by categories one and two.

That inversion is the whole problem. The scarcest resource in your faculty — the attention of experienced researchers — is spent disproportionately on the work requiring the least expertise, at exactly the moment it is most contended. We have documented the same distortion on the supervision side in why supervisors spend review time on formatting rather than argument.

The visibility problem underneath it

There is a second cost that does not show up in any workload model. During the eight weeks when your cohort is doing the most consequential writing of their degree, you have no idea what tools they are using.

Some are using a consumer chatbot on a personal account, pasting in chapters that contain interview transcripts, participant data and unpublished findings. Some are paying an editing service with no scope limit and no disclosure. Some are doing neither and struggling alone. Your institution has a policy that covers all three situations and no mechanism whatsoever for knowing which is happening, which means the policy is unenforceable in practice and the data protection exposure is entirely unquantified.

The right response to that is not more surveillance. It is provisioning — giving the cohort a tool the institution has a contract with, so that the writing happens somewhere you can support it and account for it.

What changes when the cohort is provisioned

Tesify for Institutions is built for exactly this shape of demand: a large cohort, one deadline, and a support establishment that cannot flex. Feature by feature, what it removes from staff workload during the peak:

• Reference verification and formatting handled as the student writes, checked against the scholarly record rather than generated from memory. Category one disappears from the appointment queue rather than being triaged within it.
• Structural and template compliance checked continuously against your programme’s actual submission requirements, so the deposit-stage return that delays conferral does not happen. Category two stops arriving at a human.
• Stage prompts through the writing period, which surface the structural question at week two instead of week seven — moving category three into the window where an appointment can still change the outcome.
• Institutional visibility across the cohort: who has started, who has stalled, which programmes are behind. Not surveillance of content, but the operational picture a graduate school currently does not have until submissions land.
• A disclosed assistance record per student, reviewable at a supervision meeting, which converts an unenforceable policy into a routine conversation with something concrete in it.

Nothing in that list replaces academic judgement, and it is not intended to. It clears the queue in front of it so that the fifty minutes a student gets with an expert is spent on their argument.

Book a departmental pilot for writing software before the next season

You do not need a tender, an institutional licence, or a committee cycle to test this. You need one programme, one cohort, and one dissertation season.

A free departmental pilot runs with SSO through your existing identity provider, an institutional data processing agreement signed before any student touches it, and EU or UK data residency where you require it. It sits below the procurement thresholds that would trigger a formal tender, which means an associate dean can approve it without a business case — and at the end of one term you have your own data on contact volume, peak saturation and staff time, which is a far stronger basis for an institutional decision than any vendor’s benchmark. The sequence for scoping it properly is set out in how to run a departmental pilot of an AI writing tool.

Request an institutional demo or scope a departmental pilot now, and go into the next dissertation season with the mechanical work already handled.

Frequently asked questions