A Rota Built to Lose One Person

Most of this section describes what goes wrong. This is the positive version: what a rota looks like when somebody has built it expecting an absence, which is the only realistic expectation.

The staffing problem described in “A Rota Built to Lose One Person” should be separated from assumptions about individual effort. When an organisation evaluates this workforce software resource for how to handle multiple clients, it can use time and project records to understand capacity, provided schedules, approved leave, exceptions and the correction route remain part of the same decision.

None of the six properties requires more staff. Several of them require the same staff arranged differently, which is why the exercise is worth doing before asking for establishment.

For an independent benchmark relevant to “A Rota Built to Lose One Person”, consult the Atlassian project-management guide. Use it to test scheduling, working-time limits, attendance records, employee rights and exception handling against the real operation rather than treating a software report as self-explanatory evidence.

One: every shift meets requirement with one person removed

The headline property and the one to test mechanically. Take each shift, remove each person in turn, and check the shift still meets its stated requirement.

Shifts that fail are not necessarily wrong — some will be unavoidable — but they should be known and listed, and they should be the minority. Publishing a rota where half the shifts fail this test is publishing a week with guaranteed bad mornings.

Two: two holders of every required capability, on every shift

The second mechanical test and usually the one that fails most. It is also the one most often fixed by moving a name rather than adding one.

Where it cannot be satisfied, the gap is a training requirement with a date attached rather than a permanent condition.

Three: overlap at the handover

Thirty minutes where both shifts are present gives the morning its cheapest option: the outgoing person stays, or the incoming one starts early. Without it the gap at the boundary is uncoverable by anybody already on site.

It costs the overlap hours and it is usually the best value in the whole rota.

Four: somebody on site who is not in the count

A float, a manager who is not committed to the floor, a trainer, an administrator with a current competency. One person who can absorb a gap for an hour converts a crisis into an inconvenience.

The usual objection is that this is a luxury. The usual reality is that the operation already has such a person and has committed them to the rota to save money, and then pays agency rates to replace them.

Five: no pattern feature known to produce absence

Turnarounds above the minimum rather than at it. Forward rotation. Consecutive-day limits counted across week boundaries. Start times that people can actually reach.

Each is a constraint in the build rather than a response afterwards, and together they remove a measurable share of the absence that the rota would otherwise have created.

Six: published far enough ahead to be planned around, and then left alone

Four weeks where possible, two at minimum, with a stated rule for changes afterwards.

The stability matters as much as the notice. A rota amended three times is not a four-week rota.

Testing it as a step

Before publication: the one-absence test, the two-deep test, the pattern-feature check, and the capability-expiry check. Four queries against the rota and the competency register.

Half an hour, or minutes if it is automated, and it happens while there is still time to move somebody. Everything else in this collection is about what to do once the rota is fixed and the morning has arrived.

Where the properties conflict

They will. Two-deep on every capability may be impossible with the people available. Overlap costs hours that the establishment does not have. Longer turnarounds constrain who can work which shifts, which makes the leave allocation harder.

Rank them for your operation rather than pretending they all fit. In most settings the order is: no pattern that breaks a legal rest requirement, then hard-line requirements, then capability depth, then margin, then overlap, then everything else. Writing the ranking down is what lets a scheduler make the trade consistently instead of differently every week.

What to report

Two forward-looking numbers per published rota: the proportion of shifts passing the one-absence test, and the count of one-deep capabilities in the week.

They are the only measures in this subject that predict rather than describe. An operation that watches them, and that can show them improving while agency spend falls, has the whole argument in two lines — and it is an argument about scheduling, which is within the operation's control, rather than about absence rates, which largely are not.