Temporary Worker Feedback: What Your Site Is Actually Telling You

One direction is yours: what your supervisors think of the workers on site, which the agency needs in order to send better people next time. The other is theirs: what workers make of your site, which determines whether the good ones come back.

These get conflated, and they should not be. A site with a worker quality problem and a site with a worker experience problem look identical on a headcount report and need opposite responses. Treating the second as the first is the most expensive mistake available in temporary labour, because it means replacing people who were never the problem — repeatedly, at full recruitment and induction cost each time.

Authorised users at your sites can submit worker ratings across four named dimensions — punctuality, work quality, attitude and communication — on a five-star scale, with an optional comment. The overall score is the mean of the four.

Rating four specific things rather than giving one overall impression is what makes the result usable. “Good worker” means different things to different supervisors and cannot be compared across a site, let alone between sites. Punctuality can. It also separates the fixable from the unfixable: a worker rated low on punctuality but high on work quality has a transport or shift-pattern problem, not a capability problem, and replacing them would be a waste of someone who is good at the job.

Four things make feedback worth the thirty seconds it takes:

• Rate the shift, not the person in general. A bad Tuesday is information; a general impression formed over a month is not.
• Rate promptly. Feedback given at the end of the week has already been smoothed by memory.
• Use the comment when something specific happened, and leave it blank when nothing did. A comment on every record trains everyone to ignore comments.
• Rate the good ones too. A site that only submits negative ratings gives an agency no way to identify who to send back, which is the more valuable half of the information.

And it has to be low-effort for supervisors, because feedback that costs a supervisor ten minutes will not be given during a busy shift and should not be expected. Four taps and an optional comment is the design point, not a compromise.

This is the most common frustration with agency labour and it usually has a mundane cause: the feedback never reached the booking decision. A supervisor mentions it to a coordinator, the coordinator is not the person who books the shift next Tuesday, and nothing in the system records it.

Where ratings are recorded against the worker, site history is visible at the point of booking rather than existing as something someone was told once. That is the mechanism — not a policy, not an assurance, a record attached to the person who is being allocated.

Where someone genuinely should not return to your site, say so explicitly and in the system rather than in passing. A recorded exclusion is actionable; a remark in a corridor is not, and the person who heard it will eventually be on holiday.

The same record works the other way, and that is the half worth more. Where a worker has been rated well at your site, that history sits against them when shifts are being filled — so “send us more people like the ones who worked out” becomes something the booking can act on rather than a hope expressed on a call. It improves with use: the more your supervisors rate, the better the matching gets, which is the practical reason to rate the good shifts as well as the bad ones.

Accept does not run a post-shift survey of temporary workers, and it is worth being straightforward about that rather than implying otherwise.

What exists instead is behavioural: whether a worker accepts the next shift offered, whether they attend it, whether they return the following week, how long they stay, and whether they leave in the first fortnight. Those are not opinions about a site — they are decisions about it, taken by people who have worked there, and they are considerably harder to give a polite answer to than a questionnaire.

Where a worker does raise something directly, it reaches us through the people who deal with them — the branch teams, and the development side of the service, where coaching and recognition conversations happen. That is human-led rather than automated, which makes it slower and less structured, and also more honest.

The practical consequence for a site manager: if you want to know what your workers think of your site, look at what proportion of them come back. It is a more reliable measure than anything they would write down.

So if you are shopping for an agency that supplies worker feedback data, be precise about which kind you mean. Ratings your own team submits, and anonymised patterns in attendance and retention, are things we provide. Survey responses from workers about their experience are not, because we do not collect them. An agency offering you those is worth two questions: what proportion of workers actually reply, and what happens to the experience of the ones who do not — because the people most unhappy with a site are the least likely to fill anything in.

Nothing stops you asking directly, and a site that does it well learns things no agency report will tell it. Four things separate useful answers from polite ones:

• Ask early, and in person. The first week is the most informative, because someone has not yet decided how they feel about the place and still notices things a long-standing worker has stopped seeing.
• Ask about specifics, not satisfaction. “Was the induction clear about what you would actually be doing?” gets a usable answer. “How are you finding it?” does not.
• Ask someone who is not their supervisor. A temporary worker asked by the person who allocates their shifts will say it is fine.
• Ask the ones who leave, and ask quickly. Somebody who left last week has nothing to gain by being tactful, which makes them the most honest source available to you and the hardest to reach a month later.

Then change one thing visibly. A site that asks and changes nothing gets asked-and-changed-nothing answers next time, and workers tell each other which sites those are.

Some of what workers experience shows up only in aggregate — across a site, a shift pattern or a group who started together, rather than in any one person's record. Accept reports these to clients as anonymised patterns in attendance, reliability and retention. They come from the operational record and from the conversations described above, not from a survey, and no individual is identifiable in them.

Absence that is protected in law is separated out and is not counted against a worker or reported as a reliability signal. That is a safeguard rather than a feature. It is also why an absence figure we give you may be lower than the one you would reach by counting days yourself, and it is worth knowing that before the two numbers are compared in a meeting.

Governed also means withheld. Where the data is not fresh enough, where the sample is too small to mean anything, or where a particular cut would make an individual identifiable, you are told that rather than given a number. A pattern drawn from six workers over two weeks is not a finding, and presenting it as one would be worse than saying nothing at all.

That matters most at the moment you ask for a breakdown and do not get one. “The sample is too small to tell you anything” is a more useful answer than a confident figure built on very little, and being given it is the difference between a report and a decoration.

Four checks will usually tell you which one you have, and they take an afternoon rather than a project:

1. 1Do the same workers perform well elsewhere? If people who struggle on your site attend reliably at other sites, the variable is the site.
2. 2Does the problem concentrate on one shift, one line or one supervisor while the rest of the operation holds? If so the cause is local, and no amount of supply will fix it. Review attendance by shift pattern to see where anomalies emerge.
3. 3When do people leave? Departures in the first fortnight point at induction, the first day and how people are treated on arrival. Departures at three months point at the work itself, the pay, or a better offer nearby. Tracking first-fortnight leavers isolates induction quality from wider retention issues.
4. 4What happens to the ones who stay? Are they asked back, do they get consistent shifts, is there any reason for them to be loyal?

Answering those four honestly is uncomfortable, and cheaper than the alternative — which is replacing the same roles indefinitely and concluding that agency workers are unreliable.