What a rota that survives a call-out costs: $5.82, one shift changing hands

Pricing call-out risk inside the roster search: on one team it removed every open hour a week of sick calls left, for $5.82; on three of five, nothing paid.

A weekly staff roster is usually scored on one week: the pay, the overtime, the fairness, the shifts it leaves open. Then on Tuesday someone calls out sick and the manager finds out whether the roster they posted was a good one.

The shift roster builder could already pick between rosters that tie at the best score, posting the one a single call-out hurts least. That tie-break costs nothing, which is also its ceiling: on four of five test teams the first roster found was already the steadiest, so it changed nothing. The open question was the other half — what could a manager buy by giving something up?

The quantity has to be cheap to ask. Measuring a roster's real exposure means playing every person-day as a call-out and rebuilding the week each time: about 20 seconds on a seven-person café. Nothing like that can sit inside a search. So the search gets the quickest version of the question: for each person-day, if this person called out, could someone else take each of their shifts that day with every hard rule holding and nobody else moved? If not, the day is "at risk" and its hours are counted. That check is milliseconds, and it is a strict over-count — it ignores the hand-over chains and rebuilds the real measurement tries.

Then put a price on it. --cover-price 100 means "I will pay up to $100 of pay and fairness for each hour of risk you clear." The build hill-climbs on score plus price times risk hours, one hand-over or swap at a time. Open hours cost $100,000 an hour inside the same score, so no sane price trades coverage away; what it trades is money and evenness.

On the retail template one shift changed hands. It cost $5.82 of score — $3 less payroll, the rest in the fairness penalty — and took the roster's measured call-out exposure from 24 hours to 0. Played through 20 seeded weeks of call-outs at the rates one real firm's timesheet export showed, the priced roster left no open hours where the default left 6 (sick-only rate, 1.0% of person-days) and 18 instead of 30 at the all-leave rate (8.4%), with fewer changes to other people and lower short-notice premiums at the lower rate.

On three of five teams, nothing paid for itself. On the café every single person-day is at risk and no one hand-over or swap changes that: nobody is free behind anybody. The salon and clinic the same. And on the fifth team the move the price bought cleared 3 risk hours, cost $15 of payroll, and changed the measured exposure not at all — the honest reading of a surrogate that over-counts. A move can clear risk hours that no call-out would ever have opened.

That is why the price is the manager's to set and off by default. The tool's job is not to decide that steadiness is worth $100 an hour; it is to say what the roster it is about to post would cost at that price, and to show, in the notice under the grid, exactly what it bought: "1 shift changed hands to cut the hours at risk from 81 to 54, at $5.82 of score."

The measurements, the sweep from $0 to $200 an hour at risk, and what is not claimed are in the README, with the harness in eval/cover-price.mjs. The tool is free, MIT, runs in the browser with nothing uploaded: Shift Roster Builder.

All writing · More in Tools and projects · More on Shift Roster Builder · All tools and research