Case study · Hospital network

A bed forecast planners act on every morning

A daily bed-demand forecast now drives staffing and elective scheduling across 31 wards.

Context

A network of four hospitals with 31 inpatient wards and a central bed-management team.

Challenge

Bed managers planned from yesterday's census and phone calls between wards, so electives were cancelled on the day when beds ran out.

Approach

How the work ran

WK 0–2
Frame
Sat with bed managers to learn which decisions happen at 07:30, and agreed the 72-hour horizon.
WK 2–6
Prove
Trained a forecast on three years of admissions and ran it in shadow alongside the manual plan.
WK 6–10
Build
Published the forecast at 06:00 daily with its error range, and set up weekly accuracy reviews.

Results

92%

Forecast accuracy at 72 hours

Wards covered

−18%

Elective operations cancelled on the day

“For the first time our planners argue about the plan, not the numbers.”
Director of Operations, hospital network

Related work

We will review your admissions data and tell you what accuracy is realistic.