Key takeaways

• Multi site labour forecasting estimates demand and hours at each location, then compares sites so you can make informed decisions about coverage and budgets with shared data

• Start with site signal, not a network average: each location can benefit from its own demand inputs and labour model

• Use the network roll-up to spot variance, inform staffing decisions, and set exceptions, not to force one identical rota template everywhere

• Close the loop every week: forecast → draft rota → actuals → accuracy review by site category, with managers owning the final call

You already feel the split. One site is three-deep at the counter while another burns paid hours on a quiet floor. The budget looks fine in total until you open the location tabs and see the real mess.

That is the multi-site trap: you copy last week's "average" rota, or you force every venue onto one corporate template, and you risk putting coverage in the wrong place. Multi site labour forecasting can work better when you separate two jobs. Site signal is the local demand story that should drive hours at that location. Network roll-up is the shared view that helps you inform staffing decisions, budgets, and exceptions across the group.

In plain terms, multi site labour forecasting means estimating demand and the hours you need at each site, then comparing those sites so rota and budget decisions rest on shared data, not isolated spreadsheets. Labour demand forecasting is the workforce management work that helps you decide where, when, which role, and how many people you need to meet projected customer demand so staffing can follow a demand curve rather than a flat headcount guess.

These methods are general workforce-planning practices. They are not a list of automated Deputy product outcomes. Use them as an operating model, then decide which tools can support the parts of the process you choose to run.

If you treat every location like a clone of the average site, you can miss the signals that should inform Friday night coverage at the city-centre flagship and Tuesday lunch at the suburban store. The rest of this guide shows how to build site signals first, run an honest network roll-up second, and keep managers in charge of the final rota.

Build site signals before you trust the network roll-up

A roll-up is only as useful as the location models underneath it. Build each site's demand picture and document the relevant labour and scheduling constraints before you compare venues on a dashboard.

Collect demand inputs that actually move each site

Start with the drivers that change how busy a location gets, not with a national average sales line.

Useful inputs may include:

• Sales or POS history by daypart and day of week

• Foot traffic, covers, bookings, or appointments where you capture them

• Seasonality and trading calendar (school holidays, pay weeks, bank holidays)

• Local events near the site (sport, festivals, road closures)

• Weather where it clearly shifts that site's trade

• Manager notes on one-off factors the data may miss

Store-specific demand forecasting research on restaurants notes that POS-based methods are common, yet a real store still needs a store-specific model that can factor in location, weather, events, and similar drivers. Multi-site foodservice sales research models sales with site, time of day, and weather in the mix, and discusses those forecasts as useful for labour planning. In that study, weather and daypart helped predict some drink categories, but they were not significant for total sales, so treat weather as a local test, not a blanket rule.

When you use demand forecasting to inform scheduling, keep the goal practical: help managers estimate hours for review while keeping appropriate manager oversight in the process.

Translate demand into a labour model per location

Demand without a labour model is only a sales guess. A labour model turns expected trade into roles and hours.

For each location, define:

• Staffing rates (for example, team members per £1,000 of sales, per cover, or per appointment block)

• Minimum and maximum coverage by role and open hours

• Role mix (who needs to be on for service, safety, or skill, not only total heads)

• Open hours and peak windows that differ by site format

A single corporate ratio may not reflect the operating needs of every site. A high-volume quick-service restaurant, a quiet concept store, and a late-trading venue may not share the same productive hours per pound. Build demand estimates and labour models as one connected process so hours can follow work, not habit alone.

Write the model down. If Friday "always" gets eight people because it always has, consider whether the staffing level still reflects current demand rather than relying only on historical practice.

Categorise sites so roll-ups stay honest

Before you rank locations, group them so comparisons stay fair.

Common categories:

• Format (flagship, standard, express, dark kitchen, etc.)

• Size or sales band

• Trade pattern (weekday office traffic, weekend leisure, tourist strip)

• Operating hours

Compare like with like. A city-centre flagship that runs late should not "lose" to a suburban day site on the same labour % target without context. Multi-location workforce management can help when you standardise the process and still see each site clearly, rather than forcing every venue into one story.

Run the network roll-up without erasing local reality

Once site signals exist, the network view can earn its keep. Use it to inform staffing decisions and management attention, not to flatten every rota into the same shape.

Compare forecast, rota, and actuals across locations

Build a simple weekly variance view for every site in a category:

Feed attendance and timesheet actuals back into the next forecast. Deputy's timesheet and scheduling records can help teams compare planned and worked time by site so the next plan can reflect what people actually worked, not only what the rota promised.

If forecast said busy, the rota stayed lean, and actuals moved into overtime, that may indicate an opportunity to review the forecast, roster, and actuals together rather than treating the variance as unexpected trade.

Share people and hours between nearby sites where appropriate

For organisations that share employees between locations, a network view can be useful when one site peaks and appropriately trained people are available at another location.

Practical rules that can keep shared staffing sane:

• Prefer hour moves over permanent headcount moves for one-off peaks

• Train a multi-site pool on the roles that travel cleanly

• Set travel and notice expectations before the crisis text thread starts

• Maintain appropriate site coverage so shared staffing does not leave a location short

Retail workforce management and hospitality workforce management often live with uneven peaks across locations. Where appropriate, shared staffing can be one option for responding to uneven demand between locations. For rota mechanics that can make multi-site imbalance worse, see common multi-site rota challenges.

Do not claim a measured pound win from pooling unless you have your own numbers. Track whether shared shifts reduced emergency call-outs and overtime at the receiving site.

Keep manager judgment inside clear override rules

Local knowledge can still beat the model when the street is closed, a coach party lands, or a key person calls out an hour before open. The goal is structured judgment, not gut-only rota building.

Set override rules such as:

1. Managers may adjust draft hours within a set band without approval

2. Larger moves need a short reason code (event, weather, temporary closure, VIP booking)

3. Overrides stay visible in the weekly review so the model can improve

4. Keep managers responsible for reviewing and approving staffing and scheduling decisions

Deputy scheduling tools can support managers in building and reviewing rotas.

Customers remain responsible for published rotas, pay outcomes, and how they apply workplace rules. Tools support the workflow. They do not replace your judgment.

A practical weekly cadence for multi-site labour forecasting

Use one repeatable rhythm so site managers and regional leads work from the same clock.

1. Lock demand inputs (POS, bookings, events calendar, known closures) by a fixed cutoff

2. Generate site-level demand forecasts for the planning horizon you actually build rotas to

3. Apply each site's labour model to turn demand into role hours and coverage bands

4. Build draft rotas from those hours, including multi-site pool options where relevant

5. Publish with enough lead time for your operation and any notice expectations you follow

6. Run a mid-week check on early actuals, call-outs, and cost drift

7. Close the week with an accuracy review by site category: forecast vs actual demand, hours on the rota vs worked hours, and override themes

Keep the meeting short. Review exceptions and chronic bias, not every quiet Tuesday that behaved as expected.

What good looks like when you measure accuracy

"Good" is not a vendor accuracy percentage. It is a stable habit of measuring the right gaps by site.

Track at least:

• Forecast vs actual demand by location and daypart

• Forecast or target hours vs worked hours

• Bias (a site that is often over-forecast or often under-forecast)

• Special weeks tagged separately (stocktake, festival, extreme weather) so they do not poison the baseline

Improve the labour model where bias repeats. Improve the input list where events keep surprising you. Review the shared-staffing approach where the same two sites repeatedly need last-minute coverage.

How to choose multi-site labour forecasting tools

Buy for the operating model you just built: site signal first, network roll-up second, manager review always on.

Score tools against questions like these:

• Can each location keep its own demand drivers and labour model?

• Can regional leads see forecast, rota, and actuals side by side across sites?

• Do POS, bookings, or other demand systems connect cleanly?

• Is there a clear path from forecasted demand into a draft rota for review?

• Can managers work the plan on mobile when they are on the floor?

• Can you see labour cost impact while you still edit the rota?

• Does the product help support your compliance workflows with configurable rules and records, while you remain responsible for applicable legal requirements?

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 a tool only gives you a company-wide chart, you may need to recreate site-level analysis elsewhere. If it only helps one manager build one rota, a true network roll-up can be harder to run.

Put site signal and network roll-up to work

Multi site labour forecasting is not a prettier version of last year's template. It is a two-layer system: local demand and labour models that tell the truth about each site, plus a network view that can inform staffing decisions, people moves, and management attention where the work actually is.

Do this next:

• Write down demand inputs and a labour model for every location category you run

• Stand up a weekly forecast-rota-actuals review with override reason codes

• Consider a trained multi-site pool before you default to overtime or panic hires

• Measure bias by site, not only group labour %

You own the published rota and the outcomes. The right process, and the right tools beside it, can make that ownership clearer week after week.

Contact for a Deputy demo today.

FAQ

How is multi-site labour forecasting different from single-site forecasting?

Multi-site forecasting still needs a solid forecast at each location, then adds comparison across the group. Single-site work can stop at one demand curve and one rota. Multi-site work adds a shared view so you can make more informed decisions about hours and people between locations without erasing local demand signals.

How often should multi-site labour forecasts be updated?

Update on the same cadence you build rotas, with a mid-cycle check when trade or call-outs move early. If you build rotas weekly, refresh the forecast with that cycle and adjust mid-week when actuals diverge. Feed attendance and sales actuals back into the next forecast so the plan reflects what really happened, not only what you planned.

Should every location use the same staffing ratio?

Not necessarily. Shared categories can use a starting model, but individual sites may need rates, minimums, and role mixes that reflect their format and trade pattern. Set labour models by location or category, then test them against actuals instead of locking one ratio for every address.

Can AI-assisted forecasting replace site manager judgment?

No. Where assistive or AI-enabled tools are used, they should support rather than replace appropriate manager review and decision-making. Managers still review and approve what gets published, document overrides, and own how local conditions, people issues, and workplace rules apply on the day.

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.