Scheduling software promised nurses their weekends back. The rotas tell a more complicated story.

Rostering tools are the least glamorous software in a hospital and among the most consequential. They decide who works nights, how many hours sit between a late finish and an early start, and how often a unit runs one body short.

Nurses interviewed across three systems described the same split. Where the constraint set was written with ward staff in the room, the tools genuinely improved fairness and reduced the informal favouritism that paper rotas invite. Where the constraints were written by finance, the tools simply found cheaper ways to be short-staffed.

The constraint is the policy

This is the point that gets lost. An optimiser has no opinion. It will happily produce a legal, fully covered rota that leaves a nurse with the minimum permitted rest gap eleven times in a month, because nothing in the objective function penalised that pattern.

The technical fix is trivial. Add a soft penalty on short turnarounds, cap consecutive nights, weight requests by how often the request has previously been refused. The organisational fix is harder, because each of those constraints costs money and someone has to sign for it.

What good looks like

The better implementations publish the constraint set to the staff it governs. Nurses can see what the system is trying to do, which turns an opaque decision into an argument about priorities that can be had openly.

The measure that matters

Not shifts filled. Turnover at twelve months, and the proportion of rotas where a request was declined twice in a row for the same person. Those two numbers tell you whether the tool is helping the ward or just the ledger.