Manaz Afzal

Hospitality · Growing businesses

Rostering a café from the people walking past the door

A week-ahead forecast of street footfall for a café roster, tested on two years of hourly City of Melbourne pedestrian counts on three café streets.

14.7%hourly miss a week ahead on Errol Street, against 18.0% copying last week
30%miss on public holidays, against 42%
9.0%miss on the daily total that sets staff numbers

Executive summary

A week-ahead footfall forecast misses a café street’s trading hours by 14.7% against 18.0% for copying last week, and by 14.1% once the weather is added the day before.

14.7%hourly miss a week ahead on Errol Street, against 18.0% copying last week
30%miss on public holidays, against 42%
15.2%for a plain four-week average, which gets most of the gain

Situation

Café rosters are written on Thursday by copying last week. Holidays, school breaks and the university calendar move the street more than the weather, and last week knows none of them.

Approach

Gradient boosting on two years of hourly City of Melbourne pedestrian counts for three café streets, trained on the first year and tested once on the second, every forecast using counts at least a week old.

Result

The forecast beats copying last week in every month of the test year, by 3.3 points on Errol Street. A four-week average gets most of the way; the forecast adds most on public holidays.

The problem

A café or small hospitality group writes next week’s roster on Thursday, mostly by copying last week. Rostering tools such as Deputy and Tanda can show forecast sales beside the roster, and Australian venues use them to keep wages near a quarter of takings, but they need a forecast to start from. Public holidays, school terms and the university calendar move the street more than the weather does.

I wanted to measure how much better a roster could start if it began from a forecast of the people walking past the door, set a week ahead from free data, and how much a day-before weather update adds.

Team
Owner-operator or venue manager, 6 to 20 casual staff
Scale
One to five venues on a shopping strip or in city laneways
Systems
Square, Lightspeed or Toast; Deputy or Tanda for rostering
Measure
Staff hours set to the wrong level for the trade; wage cost as a share of takings

What I built

A weekly job reads the City of Melbourne’s hourly pedestrian counts for the sensor nearest the venue, adds the public-holiday calendar, and forecasts each trading hour of the week ahead. The forecast goes into the rostering tool as a demand line beside the roster, so the manager sees which shifts are thin or heavy before publishing. On the day before, the forecast is refreshed with the weather.

Street countsHourly sensor counts,City of MelbourneCalendarPublic holidays,last four weeksWeek-ahead forecastEach trading hour,set on ThursdayRoster toolDemand line besideDeputy or TandaDay beforeWeather update;call-in or release
The manager still writes the roster; the forecast shows where it is likely to be wrong.
0100200300MonTueWedThuFriSatSunForecast, set a week aheadSame hour last weekActualTrading hours, 7am to 4pm each day
A normal week. The forecast follows the shape of each day; copying last week carries last week’s quirks forward. Data: City of Melbourne Pedestrian Counting System, sensor Errol23_T, CC BY 4.0

Results

I trained on the first year of counts, from November 2024, and tested on the following year to 17 November 2026, every trading hour from 7am to 4pm, on three streets with cafés: Errol Street in North Melbourne, Lygon Street in Carlton and Degraves Street in the city laneways.

Same hour last weekwhat most rosters copy18.0%Average of the last four weekssame weekday and hour15.2%Forecast, set a week aheadcalendar, holidays, four lags14.7%Updated the day beforeadds the weather14.1%
On Errol Street the week-ahead forecast misses by 14.7% an hour against 18.0% for copying last week. Most of that gain comes from averaging four weeks; the forecast adds a little more, and the day-before weather update brings it to 14.1%.

Against copying last week the forecast is 3.3 points better on Errol Street, with a 95% interval of 2.5 to 4.3 points from resampling the test days. Against a plain four-week average the gain is under a point, which is the result I would act on: a café without any modelling should roster from the four-week average, and the forecast earns its place on public holidays, where copying last week misses by 42%.

StreetPassers-by an hourLast weekFour-week averageForecastDay-before updateDaily total: last weekDaily total: forecast
Errol Street, North Melbourne16818.0%15.2%14.7%14.1%11.1%9.0%
Lygon Street, Carlton22924.5%19.2%18.6%16.7%18.9%13.8%
Degraves Street, city laneways83214.4%12.2%12.0%11.2%10.2%8.2%
StreetPublic-holiday hours: last weekPublic-holiday hours: forecast
Errol Street41.9%29.6%
Lygon Street37.9%34.3%
Degraves Street29.1%23.8%
Same hour last weekForecastNov 202510%Dec 202517%Jan 202618%Feb 202612%Mar 202613%Apr 202616%May 202612%Jun 202617%Jul 202612%Aug 202613%Sep 202613%Oct 202618%Nov 202614%
Every month of the test year, the forecast misses less than copying last week. The gap is widest in January and October, months with public and school holidays, and narrowest in June.

The daily total, which sets how many staff are on, is easier to forecast than each hour: on Errol Street the forecast misses the day by 9.0% against 11.1% for last week.

Data and method

Data
City of Melbourne Pedestrian Counting System, hourly counts, 19 November 2024 to 17 November 2026, CC BY 4.0; Open-Meteo historical weather for Melbourne, CC BY 4.0
Streets
Sensors Errol23_T, Lyg161_T and FLDegS_T, trading hours 7am to 4pm
Split
Train on the first year, test once on the second; every forecast uses counts at least seven days old
Methods
Gradient boosting with an absolute-error objective on hour, weekday, month, public holiday, the same hour one to four weeks back and their average; the day-before version adds temperature and rain
Error
Weighted absolute percentage error: total hourly miss divided by total passers-by

Limitations

How I would take it live

Financial impact

About A$12,645 a year in staff hours set closer to the trade

Value is the share of rostered hours that no longer sit above or below the trade, at a casual wage rate.

InputValueSource
Rostered staff hours a week180 one café, seven days
Hourly miss: copying last week18.0%Measured Errol Street
Hourly miss: forecast with day-before update14.1%Measured
Casual wage with on-costsA$35

Calculation

  1. Miss removed3.9 points
  2. Hours a week set closer6.9
  3. Value a yearA$12,645

Annual value of a closer rosterA$12,645

Footfall is not trade; the forecast is checked against the venue’s own till data by hour before shifts follow it.

Implementation and rollout

Six weeks from a sensor feed to a demand line in the roster tool

  1. Week 1

    Connect

    Nearest sensor feed, public holidays and twelve months of till transactions by hour.

    Owner, Manaz

  2. Weeks 2 to 3

    Check

    Forecast against the till for a month; buy rate set by hour.

    Manaz

  3. Weeks 4 to 6

    Run alongside

    Demand line beside the roster in Deputy or Tanda for one venue.

    Venue manager

  4. Then

    Extend

    Wage share of takings compared with last year; other venues added if it holds.

    Owner, Manaz

Tools

City of Melbourne pedestrian sensor API, Square, Lightspeed or Toast till exports, Python gradient boosting, weekly and day-before, Deputy or Tanda demand line

Team

  • Owner or venue manager writes and publishes the roster
  • Shift lead calls in or releases staff the day before
  • Manaz feeds, forecast, till check and monthly review

The timeline is a plan for a typical engagement; it depends on data access and how many systems are involved.

Risks and mitigations

What could go wrong in production, and the control for each

RiskMitigation
Passers-by do not convert at a fixed rateBuy rate set by hour from the till and refreshed monthly.
Sensor outage or roadworksOutages flagged and the four-week average used instead.
Events not in the calendarKnown-event list kept by the manager and added to the forecast.
Staff rostered to a number they distrustForecast shown as a guide beside the roster; the manager decides.

Key takeaways

Key takeaways

  • Copying last week carries last week’s quirks forward.
  • A four-week average gets most of the gain with no modelling.
  • Public holidays are where the forecast earns its keep.
  • The daily total that sets headcount is easier to forecast than each hour.