Where People Actually Leave
Digital reporting is most credible when the workforce knows what is recorded and how it will be used. That principle applies to review the service: the tool can support an operational check, but it cannot repair an inaccurate advert, a chaotic first day or poor line management.
When this issue touches live workplace policy, compare the local evidence with guidance published by SHRM before changing a process or communicating a rule.
Most warehouse attrition happens in the first two weeks, and most of it was decided before the person started.
Warehouse operations recruit continuously because they lose people continuously. The losses are not spread evenly across the year of employment.
The shape of the curve
A large share of leavers go in the first fortnight.
Another group goes between one and three months, once the novelty and the initial effort have worn off.
After six months the rate falls sharply, and people who reach a year mostly stay for several.
Which means the problem is concentrated at the start, and a retention programme aimed at long-service staff is aimed at the people least likely to leave.
What the first-fortnight group is telling you
They are not leaving because of pay progression, career development or recognition.
They are leaving because the job was not what they were told, because the first day was chaotic, because nobody spoke to them, or because the shift is physically harder than they expected.
All four are decided before they started or in the first eight hours, which is where the intervention has to be.
The measurement problem
Recruitment is usually measured on cost per hire and time to fill.
Both reward filling a seat quickly with whoever is available.
Neither notices that the person left on day nine and the seat is open again.
Which means the recruitment function can hit every target while the operation is short-staffed all year, and nobody is doing anything wrong by their own measures.
The measure that changes behaviour
Cost per person still present at ninety days.
Or, more bluntly: how many of last month's starters are still here?
It is a single number, available from records you already hold, and it reframes everything.
A cheap hire who lasts a week costs more than an expensive one who stays a year, and until the measure says so, the incentives point the wrong way.
What it actually costs
Advertising and agency fees, which is the visible part.
Induction and training time, including the trainer's hours.
The ramp period, during which the person is paid full rate for partial output.
Errors and their downstream cost.
Overtime and agency cover for the gap.
And the supervisor's time, repeatedly, which is the largest hidden line.
Add them and the figure per lost starter is large enough to fund a great deal of prevention.
Why it persists
Because the costs are spread across budgets that do not talk to each other.
Recruitment carries the fees. Operations carries the overtime. Training carries the hours. Quality carries the errors.
Nobody sees the total, which is the structural reason a problem everybody complains about goes unaddressed for years.
Producing the total once, on one page, is the most useful thing anybody in this subject can do.
What this collection covers
The advertisement, the interview and what people are told before they start.
The first day, the first week and the ramp to standard.
Agency labour, which is most of the workforce in many operations.
Peak, which is a hiring problem more than a scheduling one.
Shift patterns, pay structures and what people actually choose between.
Supervision, which determines more than any programme.
And the exit data, which is the only place people tell you the truth and which almost nobody collects usefully.
What it does not cover
Measuring how fast people work, which is a separate subject with its own obligations.
This is about getting people through the door and keeping them long enough to be worth the trouble, which most operations treat as a recruitment problem and which is mostly not one.
Building the curve for your own site
Take twelve months of leavers and plot length of service at exit, in weeks.
Not a table of reasons — a distribution of durations.
Most operations have never drawn it, and drawing it takes an hour from the payroll records.
The shape tells you where to spend: a spike in week one is an induction problem, a spike at week six is a supervision or a ramp problem, and a flat curve across the year is a pay or a market problem.
Separating the groups
Agency and direct staff have different curves and different causes, and pooling them hides both.
Peak hires and permanent hires likewise.
Split them before drawing any conclusion, because a headline attrition figure on a site that doubles its headcount in November describes the peak arrangement rather than the operation.
More in this section