Key Takeaways

• Multi site labour forecasting estimates demand by location and time band, not one company-wide average

• Convert each site forecast into role-level hours, then review rosters against relevant award, skill, and availability inputs

• Score forecast versus actual and bias by site so network roll-ups do not hide local misses

• Use the network view for comparison and targets; use site truth for roster decisions

It is Friday night across your network. One site is short on floor staff and the queue is growing. Another site is quiet, with people standing around on the same labour percentage target you set for every location.

That gap is the multi-site problem in plain view. Multi site labour forecasting estimates labour demand by location and time band, helps translate those needs into role-level hours, and gives managers a structured basis for reviewing rosters against relevant rules and constraints. The win is not a single company average. It is a repeatable forecast-to-roster loop per site, with central comparison across the group.

In operations practice, you treat each location as its own demand system. You still share one method and shared metrics. You do not force one headcount ratio onto every venue. That is the difference between a network average and site truth.

This sits on a broader workforce planning idea. Strategic workforce planning is about getting the right people with the right skills in place at the right time so the organisation can meet its goals. At a foundation level, workforce planning aims to balance supply and demand over the short and longer term. Those ideas come from strategic and system-level planning. Your job is more tactical: hourly coverage by site, by role, and by shift.

This guide shows why single-site logic frays at network scale, what inputs and steps good multi-site labour demand forecasting needs, how area managers can run a weekly forecast-to-roster loop, and how to judge quality without vanity scores.

Why single-site logic breaks across a network

Single-site forecasting can work when one location owns its peaks, people, and rules. Across a network, the same method often still works. The same numbers usually do not.

Demand patterns differ by suburb, street, and channel mix. Local events and school holidays hit some sites hard and leave others flat. Role mixes change: one store needs more supervisors on Saturday; another needs more pickers midweek. Award coverage, skill rules, and opening hours can differ by role and site. When each manager keeps a private spreadsheet, you lose a shared view of the gap between plan and reality.

A simple example shows the trap. You set one labour percentage target for every site. Site A has a sharp Saturday lunch spike. Site B builds slowly toward a late retail evening. The network average labour percentage looks fine. Site A still runs thin at peak. Site B still pays for idle hours before the evening rush. The average hid both failures.

Workforce planning literature often frames the work as supply, demand, gap, and solution. That four-area model is a useful checklist when you scale. Demand is not one number for the brand. The gap is not one network miss. Solutions only stick when you close the gap where the work happens. Geographic imbalance is a known multi-location planning problem in health systems research; hourly retail and hospitality networks face a practical version of the same issue when local workload diverges from the group mean.

Multi-location workforce management can be more effective when your process keeps site-level demand visible while leaders compare sites on shared metrics. Control without site truth becomes a smoothed spreadsheet. Site truth without a shared method becomes roster chaos.

Demand is local even when the brand is not

Your brand standards can be consistent and your trading curves can still diverge.

In retail workforce management, a CBD store may peak at weekday lunch while a suburban store peaks after school and on weekends. Click-and-collect volume can bend the curve again for the same banner.

In hospitality workforce management, a CBD venue may lean on weekday corporate trade while a residential venue leans on Friday and Saturday nights.

Build a demand profile per site. Use the units you already trust: sales, transactions, covers, bookings, visits, or occupancy. Split profiles by daypart where peaks move. Keep the brand promise the same. Let the labour shape follow the work.

Rules and constraints differ by site and role

A forecast that only outputs “hours needed” is unfinished work. Those hours still have to become a roster people can work under your pay and scheduling rules.

In Australia, modern awards are legal documents that set minimum pay rates and conditions of employment. They commonly cover pay rates, penalty rates and allowances, hours of work, breaks, rosters, and overtime. An employer can be covered by more than one award depending on the jobs employees do. That matters in multi-site groups where retail, hospitality, and support roles sit under different instruments, or where different roles at the same site sit under different coverage.

If you convert a demand spike into roster hours without reviewing those constraints, you can create higher-cost staffing patterns, potential issues around hours and breaks, or push managers into last-minute fixes. Treat relevant award settings, skill requirements, opening hours, and availability as important inputs to the planning and review process, not as cleanup after publish.

Disclaimer: Award coverage and employment conditions depend on your circumstances. This article is general information, not legal advice. Check current Fair Work Ombudsman guidance or get advice for your sites and roles. Employers remain responsible for meeting their obligations.

What good multi-site labour forecasting includes

Good multi-site labour forecasting is a data and decision system, not a single chart.

Inputs checklist

• Historical sales, transactions, covers, bookings, or occupancy by site and time band

• Foot traffic, queue, or appointment signals where you capture them

• Labour hours actuals by site, role, and daypart

• Local events, school holidays, trading-hour changes, and weather-sensitive periods you already track operationally

• Labour budgets and productivity or coverage standards managers still accept

• Skills, qualifications, and minimum coverage by role

• Approved leave and known unavailability

Method building blocks

• History: what this site usually needs in this week type

• Drivers: what is changing demand next week

• Manager override: documented adjustments for known events or outages

• Scenarios: baseline, busy, and quiet ranges so you are not locked to one line

Cadence

• Short-term: roster horizon refresh (often weekly, with daily tweaks in high-variance sites)

• Medium-term: hiring, training, and budget checks on a longer cycle

• Planning assumptions: revisit on a steady rhythm; mature strategic workforce planning practice often refreshes at least quarterly

For industry demand factors that feed labour demand forecasting, see Deputy’s guide on how to forecast labour demand across sectors.

From demand signal to labour hours

Demand units are not roster lines. You still need a conversion step.

1. Pick the demand unit for the site (for example transactions per hour or covers per service).

2. Apply a staffing standard or productivity ratio by role that managers have reviewed.

3. Set minimum and maximum coverage based on your operational requirements.

4. Produce a baseline hour plan, then busy and quiet scenarios for known events.

5. Hand the hour plan to rostering after skill and rule review.

Review standards when service models change. A ratio that fitted last summer can miss a new channel mix this year.

The forecast-to-roster loop for area managers

Multi-site labour forecasting is less useful when the forecast stays in a slide pack instead of informing roster decisions. Area managers need a weekly operating loop that ends in published rosters and measured actuals.

1. Refresh site forecasts for the roster horizon using each site’s profile, drivers, and known events.

2. Convert demand into hours by role with current staffing standards and min/max coverage.

3. Review award, skill, and availability constraints before you lock names to shifts.

4. Draft rosters for manager review so local leads can challenge odd peaks, gaps, or people mismatches.

5. Publish once reviewers accept the plan.

6. Track actuals mid-week so you can adjust open shifts or coverage while the week is still live.

7. Score forecast versus actual and bias by site after the week closes, then feed misses back into next week’s profile.

Cross-site cover can help after site truth is clear. Moving people between locations without a solid local forecast can create additional staffing mismatches.

Tools should support the loop, not replace judgement. Deputy Labor Modeling can use sales data, custom business metrics, or manager forecasts to help determine required staffing levels. Deputy Rostering can help managers build and review schedules based on their staffing plans. Deputy Time and Attendance can help you capture clocked hours and attendance patterns by site so forecast-versus-actual reviews use real worked time. Where available, keep AI-assisted drafts and suggestions subject to manager review. Users remain responsible for reviewing proposed changes before confirming them.

How to judge forecast quality without vanity metrics

If you only celebrate a neat network chart, you will miss the site that keeps failing on Saturday.

Start with a simple definition. Forecast error is actual minus forecast for the same site, role group, and time band. Direction matters. Consistent over-forecasting can increase labour costs. Consistent under-forecasting can create staffing pressure and service challenges.

A common scale-free summary is mean absolute percentage error (MAPE). The US Census Bureau describes MAPE as a measure of the central tendency of errors calculated by averaging the sum of the percentage differences between projections and actuals, ignoring the plus or minus sign. In plain terms, you look at how far off you were in percentage terms, treat misses as positive distances, and average them.

Also track bias by site: do you systematically staff high or low? A low MAPE with hidden bias can still leave every Friday underdone.

Watch the limits. Research on forecast accuracy notes MAPE limitations for near-zero actuals: MAPE can produce infinite or undefined values when actual values are zero or close to zero. That shows up in quiet overnight slots, closed dayparts, or roles with intermittent demand. For those bins, use counts of absolute hour error, coverage hit rates, or other measures that do not blow up at zero.

Prefer site-level scorecards over one network MAPE. A single group score can look healthy while two sites carry the entire miss. There is no universal “good MAPE” target for labour forecasting in this guide. Choose thresholds from your own history, service goals, and cost tolerance, then review them as the business changes.

What poor multi-site forecasts cost your people and your service

Poor forecasts can affect labour costs and workload.

Overstaffing pays for idle time when demand never arrives. Long idle periods can leave people under-used and disengaged, especially when they cannot take on other tasks. Understaffing can contribute to queues, missed service standards, workload pressure, and managers filling gaps themselves.

Safe Work Australia treats job demands as a psychosocial hazard when demands are too high or too low. High demands include having too much to do in too little time, or working long hours without enough breaks. Low demands include long idle periods. Controls include planning the workforce so you have the right number of skilled workers for the job, and scheduling enough time for difficult work to be done safely. Multi-site labour forecasting can form part of a broader workforce-planning approach.

Skip vendor-style labour cost percentage claims unless you have your own finance baseline. Your internal forecast-versus-actual and overtime patterns will tell a clearer story than a generic industry saving figure.

Deputy is designed to support compliance workflows but does not provide legal advice or guarantee compliance. Customers remain responsible for configuring the platform appropriately and complying with applicable laws and regulations.

If you want one place to connect multi-site forecasting work to rostering and time data, try Deputy for free or talk to sales about your network.

FAQ

How often should multi-site labour forecasts be updated?

Refresh on the roster horizon first. High-variance retail and hospitality sites often need weekly rebuilds with daily tweaks when events or weather shift trade. Use a medium planning cycle for hiring and budget checks. Revisit planning assumptions on a steady rhythm; many mature strategic workforce planning programs refresh at least quarterly. There is no single rule for every network.

Can one labour model work for every location?

You can share one method, one metric set, and one review cadence across the group. Identical headcount ratios or labour percentages may not reflect the demand patterns of every site. Keep shared process, local demand profiles, and local rules.

Should area managers override site forecasts?

Area managers may need to adjust site forecasts when they have relevant information the model does not capture, such as a local event, outage, renovation, or one-off booking. Document the reason, time-box the adjustment, and score whether it improved forecast-versus-actual. Silent overrides train the wrong habits and hide model gaps.

How does Deputy support multi-site labour forecasting work?

Deputy can help multi-site teams connect demand-informed planning to draft rosters and site-level time data for manager review. Use shared visibility across locations, then keep final staffing decisions with your managers. Start with a free trial if you want to test the workflow on your own sites.

Disclaimer: This article provides general information only and is not legal advice. Workplace obligations vary depending on your circumstances. Deputy is designed to support leave, payroll, time & attendance and scheduling workflows but does not provide legal advice or guarantee compliance. Employers remain responsible for complying with applicable workplace laws and configuring the platform to meet their requirements.