Staffing rotas that save £18K a month across six sites.
Forecast predicts footfall by location and daypart for the next four weeks using historical covers, reservations, and a local-events calendar. Managers set rotas with 25% less overtime while maintaining service, saving roughly £18K a month across the group.
Run Forecast- 1
Describe the task
Forecast can predict footfall by location and by daypart four weeks ahead, using historical covers, live reservations, and a local-events calendar. Instead of managers rostering on gut feel and padding every shift with overtime insurance, each site gets a demand curve it can staff against, saving roughly £18K a month across the group without thinning service.
The richer the covariates, the sharper the curve. Tell Forecast which sites to model, which dayparts to split, and which events to fold in, and it produces a per-site, per-daypart forecast the rota can be built on directly.
Forecast footfall 4 weeks ahead for all six sites in covers-history.csv. - Split each day into dayparts: breakfast, lunch, dinner, late - Attach reservations.csv and local-events.csv as covariates - Return expected covers per site per daypart, with a busy/quiet flag - Translate the curve into a suggested headcount per shift at our target covers-per-server ratio End with a summary: sites forecasted, projected overtime reduction, and any daypart at risk of being understaffed.
- 2
Give Forecast context
In a Forecast workspace, load each site’s covers history into a source node and wire reservations and the events calendar in as covariates. Forecast learns each location’s own rhythm, a business-district lunch site and a weekend-dinner site get different curves, then breaks the day into dayparts so the rota matches when demand actually lands.
Make sure the events calendar is geocoded to the right sites. A stadium match three streets from one location shouldn’t inflate the forecast for a site across town, so tie each event to the venues it actually affects.
Required contextA covers history per site with at minimum site, date, daypart, and covers servedA target covers-per-server ratio, so the forecast can be turned into headcountOptional contextA live reservations feed, so confirmed bookings sharpen the near-term curveA local-events calendar geocoded to each site, so matches and festivals lift the right locationsA weather feed, so a heatwave or washout adjusts patio-heavy sitescovers-history.csvreservations.csvlocal-events.csv (optional) - 3
What Forecast creates
Forecast learns each site’s daypart rhythm and folds in reservations and events. You get a 4-week footfall curve per site and daypart, a suggested headcount per shift, and a flag on any daypart at risk of being under- or over-staffed.
From Forecast: Forecasted footfall for 6 sites across 4 dayparts, 4 weeks out, from covers-history.csv.
£18KProjected monthly saving−25%Overtime hours6Sites forecastedNext Saturday, dinner6 sitesSite Forecast covers Suggested servers Flag Camden 184 covers 9 Local gig lets out at 22:30, hold a late server Shoreditch 212 covers 10 Busy, fully booked by 20:00 Canary Wharf 96 covers 5 Quiet, weekend district lull Understaffing risk this week2 daypartsSite Daypart Gap Note Camden Sat late +1 server Event-driven spike after 22:00 Islington Fri dinner +1 server Reservations already 30% over baseline "Camden’s Saturday late daypart is forecast well above baseline because of a nearby gig. Want me to auto-add a late server to the suggested rota, or just flag it for the manager to decide?"
- 4
Follow-up prompts
Draft the rota from the forecast
Forecast can turn the per-daypart headcount into a draft shift schedule per site, so managers start from a filled grid and adjust, rather than building from scratch.
Turn next week’s forecast into a draft rota per site: shifts by daypart at our covers-per-server target, and highlight where I’d need to call in an extra hand.
Test a bank-holiday scenario
Forecast can re-run the four-week curve with a public holiday or a big local event added, so managers see the demand shift before they lock the rota.
Add the August bank holiday and the Camden street festival to the events covariate and re-forecast. Show me which sites and dayparts need extra cover.
Refresh it every week
Once the dayparts and ratio are right, save the pipeline as a Forecast saved runbook and schedule it to rebuild each week against the latest covers and reservations.
Save this as a runbook called "weekly-footfall-rota", schedule it for every Wednesday at 7am, and send each site manager their draft rota for the coming week.
- 5
Tips and troubleshooting
Split by daypart, not just by day
A daily footfall number staffs the wrong hours, a quiet lunch and a slammed dinner average into a misleading middle. Give Forecast daypart-level history so the rota matches when demand actually arrives.
Geocode events to the sites they hit
A concert only lifts the venues near it. Tie each event to the specific sites it affects, or a citywide festival inflates the forecast for a suburban location that never sees the crowd.
Staff to the busy edge on event nights
On event-driven spikes the average understates the peak. For flagged dayparts, roster against the upper end of the forecast so a predictable rush doesn’t sink service.
Ready to try it yourself?
Point Forecast at your covers history and events calendar, and give every manager a four-week footfall curve they can staff against, less overtime, same service.
Run Forecast