Almost every applicant tracking system on the market was designed around a permanent placement: one vacancy, a shortlist, an offer, a start date, done. A temporary desk does not work like that. “Ward 4 needs a nurse on early, Monday to Friday” is not one thing that can go wrong. It is five.
This page is about the specific places that difference bites — the ones agencies usually discover in week three, after the data is already in.
The permanent model treats a role as a container that is either empty or full. Every workflow follows from that: a pipeline that moves in one direction, a status that advances, a placement that closes the record. It is a good model, and it is the wrong shape for temporary work.
A temporary requirement is a pattern of shifts, each covered or not covered on its own day, by a person who may be perfect for Monday and unreachable on Thursday. The unit is the shift, not the vacancy. Everything below is a consequence of getting that one decision wrong.
A perm placement ends the record. A temp booking ends on Friday, and the same requirement is open again on Monday with a different person in it.
Who is suitable barely changes month to month. Who is free changes hourly, and is the only question that matters when a client rings at 6am.
A perm shortlist can take a week. A shift starting in four hours is filled by whoever the consultant can establish is available, compliant and not already working — from a phone, between calls.
None of which is exotic. Health, education, early years, care, construction and driving are all overwhelmingly temporary desks, and between them they are most of the UK recruitment market by placement volume.
The commonest failure by a distance. “Availability” exists on the candidate record as free text: “Mondays and Wednesdays, some weekends”. It is how a candidate describes themselves, and it is useless to a desk filling the 14th.
The trap: it looks like the feature is there. Nobody discovers it is not until they try to answer “who can work Tuesday night” and find the only way is to read every record.
Healthcare runs early, late, night and long day. Education runs AM and PM. Construction and driving run days and nights. Offshore runs a rotation that is not a shift at all.
The trap: any system that ships a fixed list of shift types has picked one sector's vocabulary and is wrong for everybody else — the same mistake as a hardcoded list of compliance document types. If your desk has to call a night shift “Shift C” because that is what the dropdown offers, the software is now training your consultants in its language rather than yours.
To a database filter they look identical: not returned. To a recruiter they are completely different. One is somebody to try again next week. The other is somebody already earning, on your desk, whose booking you might be about to double.
The trap: collapsing the two loses the single most useful piece of management information a temp desk has, which is how much of the available pool is actually working.
A candidate who has said nothing about next week is not available. They are unknown. Systems that return everyone who has not explicitly blocked a date produce a list that is mostly people who never offered.
The trap: this one has a cost outside the software. A desk that rings people who never said they were free is how an agency wears out its own candidate base, and the people who stop answering are the reliable ones who get rung most.
Two consultants, one candidate, two clients, the same Thursday. On perm software nothing prevents it, because a permanent candidate genuinely can be in two pipelines at once — that is normal and desirable.
The trap: the clash has to be refused at the point of booking, not flagged in a report somebody reads on Friday. And a whole-day booking has to block every named shift that day, because somebody working all day is working all day.
Most systems hold one mandatory document list for the whole agency. Real desks supply clients who each want something different, and a candidate who is compliant for one client is not automatically compliant for the next.
The trap: a per-client list that adds to the agency default sounds richer and is much harder to reason about. “Why is this person compliant for Client A and not Client B” stops having a short answer, and a system only an expert can operate is as bad as one only an expert wants.
Not the candidate record. The list. If the answer involves a search filter over free text, or a report, the feature is not there.
If the list is fixed, ask what happens when a client uses a word that is not on it. The answer is usually a note field, which is another way of saying nobody can query it.
Watch whether it is refused, warned about, or allowed. Then ask what a whole-day booking does to the named shifts around it.
Does the candidate now show as free, while they are still working the other one? This is the bug that is invisible until it puts somebody in two places.
If cancelled bookings are deleted rather than kept, that question has no answer — and reliability is one of the few things a temp desk can actively manage.
A rota grid demonstrates beautifully on a laptop and is unusable in a car park between calls, which is where a great deal of temp booking actually happens.
Availability is stored as a row per date, not as a recurring pattern. A candidate can still say “Mondays and Wednesdays” and the screen will write the dates — which means “free Mondays, but not that one” is an ordinary correction rather than an exception the system has to remember.
The shift label is whatever your desk calls it. There is no fixed list to fit into; suggestions come from what your own organisation has already typed, and a day with no label means the whole day. A candidate can be available, unavailable, or already booked — three states, because the last two are not the same fact.
Somebody who has said nothing is never returned as available, and the empty list says so in words rather than looking like a fault.
Booking somebody onto a shift sets their availability for that date to booked, and nothing else in the system does. That is what stops a second consultant ringing someone who is already working.
The refusal names the booking they are already on. A whole-day booking blocks every named shift that day; two different named shifts on one date are fine.
Cancelling frees the day only if they have no other active booking overlapping it — and a candidate who said unavailable keeps saying it, because a cancellation is not new information about what they can do.
Shifts can be booked against a client before anybody has written a job posting, which is the common case on a temp desk rather than the exception. How many people a client asked for is stored; how many are booked is counted, so the two can never drift. Cancelled bookings are kept, so “who drops out, and how often” stays answerable.
Each client can carry its own list of required documents. That list replaces your agency default rather than adding to it, deliberately, so the answer to “which rules is this client on” is always one list and never an inheritance puzzle. A client with no list of its own follows your standard set.
There is no rota grid, on purpose. A grid is the thing a desk cannot use on a phone between calls.
The eight credential regimes UK agencies actually have to track, the renewal cycle for each, and the trap in every one.
How long an agency can keep a CV, where the answer really comes from, and the five questions that decide it.
What exports cleanly, what never does, and the checklist to work through before you give notice.