How Housing Drives Worker Turnover and Retention

Every HR lead at a contractor or an industrial employer carries a turnover number and a ready list of explanations for it: pay, a better offer elsewhere, the direct supervisor. Accommodation is easy to leave off that list, even though it holds every hour a worker is not on site.

Accommodation is rarely the leading cause. Pay, contract terms and supervision often outrank it. It still deserves attention, because it goes missing from the analysis more consistently than any of them, and because it is one of the few things an employer can change inside a season.

The replacement cost nobody attributes to housing

Replacing a worker here costs more than an advert and an interview: recruitment, visas and travel, screening, arrival and induction, supervisor time pulled off the site, then a stretch during which the new arrival is not yet producing what the leaver produced. In a trade on the critical path, the cost moves into the schedule.

The accounting problem is that these items are scattered: recruitment and mobilisation in one budget, supervisor time in overhead, schedule slip in project risk. Accommodation sits in a line of its own, and no report puts the two sides on one page, so trimming it looks like a saving while the offsetting cost grows elsewhere.

Extra spend does not automatically come back as retention. But a discussion that opens and closes inside the housing line has read only one side of the ledger.

What actually drives the cost of worker housing→

What actually pushes people out

Employers assume accommodation quality is a question of the building. What a resident lives inside day to day is the operating layer before it is the building: what works at night, how long they wait at peak, what they eat, and whether anything they say reaches anyone.

• Sleep and overnight cooling. A room that does not cool at night means shifts worked short of sleep, night after night, and the effect is cumulative.
• Washrooms, wash points and kitchens at the peak hour, not on a daily average. A ratio that reads well on paper collapses in the window before a shift.
• Food. Repetition, menus that do not suit the nationalities actually present, serving times that do not match shifts. It comes up every day because it happens every day.
• Privacy and personal storage. A lockable place for a passport, a phone and money. Where there is none, you get arguments over missing property, and a room that never quite settles.
• Distance and the commute. A journey at both ends of a long shift adds hours that are neither counted nor paid, and fatigue shows in performance before it shows in a resignation.
• Contact with family. A connection that works at the hour they promised to call. A worker who cannot check on their family carries a problem they cannot solve from where they stand.
• Whether a complaint produces a result. A report closed without a repair teaches the resident that the next one is pointless. Reporting stops, the log goes quiet, and a quiet log is the worst reading on the report.

None of this appears in a tender document or on a daytime inspection arranged in advance. It shows at night or at peak, when nobody visits.

Summer readiness: what has to work before the heat arrives→

Turnover is a lagging indicator

The conditions that push people out produce their effects long before anyone leaves: absence and lateness, output below what a worker is capable of, injuries that read as fatigue rather than carelessness, room disputes, requests to move site. All of it is cost paid before a resignation.

Departure is more collective than it looks. It gets made in the room, among people who have known each other since recruitment, and when someone they respect leaves, the others ask themselves the same question. That is why an attrition line looks flat and then steps up all at once.

Exit interviews and the complaint log: reading both honestly

If you run exit interviews and keep a complaint log, you already hold two sources that speak to the accommodation. Both are useful, both are biased in opposite directions, and the common mistake is treating either as a measurement.

Why exit interviews under-report housing

The leaver wants a clean exit: a reference, the option of coming back. Criticising the accommodation criticises something you provided, not the market. And the conversation is usually held in a language that is not their first, often by people close to the housing operation.

• Pay is a socially acceptable answer that ends the conversation in a sentence. Accommodation needs explaining, and explaining needs trust and time an exit conversation does not provide.
• Housing degradation is cumulative and slow, and hard to put into words on the spot. Nobody leaves over a hot night; they leave over a season of them.
• Someone who has already left has nothing to gain from candour, while the people who stayed hold the more useful information and are never asked.

Why the complaint log over-represents the loudest issue

A log records who complains, not what is broken. A resident who trusts their supervisor opens entries; a block where the residents and the supervisor have no relationship opens none, and the second one looks better on the report. The log also skews toward the fixable, a broken light, over the structural, a washroom queue before a shift. Closure rates measure administration, not resolution.

Read exit interviews as a floor rather than a proportion, and the log as a map of trust rather than of faults. Then add a source neither controls: an unannounced visit at the hour the problem bites, by someone with no operating stake in the answer.

Building a complaints process that produces repairs, not closures→

What to measure before the number moves

This needs no new system. Much of this data is already collected, aggregated at project level rather than accommodation level. Segmenting by block is what makes it useful: if one block moves and the rest stay flat, that block is telling you something.

1. Time to close cooling and water faults, measured from the report rather than the work order, and at night specifically. Track the worst cases, not the average.
2. Repeat rate for the same fault at the same location. On its own it separates what was fixed from what was closed.
3. Peak-hour load on washrooms and wash points, not their ratio to total residents. The problem lives inside a narrow window and does not exist outside it.
4. Meals served and not eaten. Uneaten food is a cost and a signal at once, and it tends to move quickly when something changes.
5. Short-notice absence, segmented by accommodation rather than by project or trade.
6. Requests to move between sites or rooms, including informal ones that reach a supervisor verbally. These often come before a resignation.
7. Turnover itself, segmented by accommodation and by length of service. Early departures point at arrival conditions; later ones point at conditions that accumulated.

Take a baseline, then measure again after a season. You will not get a clean causal relationship, but you will get a direction of travel in indicators that move ahead of turnover and sit inside your control.

What accommodation cannot fix

If pay is below the market, if the direct supervisor is the problem, or if the work itself is unsafe, good accommodation will not hold anyone. It pushes harder than it holds. Nobody stays for the accommodation alone, but bad accommodation can push out people who would otherwise have stayed.

Do not expect a clean causal read. The season changes, the project phase changes, the recruitment channel changes, the labour market in the source country changes. All of them move the number for reasons that have nothing to do with the accommodation. Anyone who promises that better accommodation will cut turnover by a stated amount is selling a certainty they do not have.

The reasonable position: for most employers the accommodation is easier to change inside a quarter than the pay structure. Judge it on what you can evidence, on whether conditions improved and whether the leading indicators moved, rather than on what you cannot isolate. If nobody has the time to look, that is an answer in itself.