Key takeaways: a workforce forecast plan estimates, interval by interval, the volume of upcoming work (calls, emails, cases, walk-in customers) and the number of people needed to handle it. It is built in five steps: set the horizon and the granularity, analyse historical data, forecast the volume, convert that volume into workload and then into headcount, and compare that requirement with the resources available. The core calculation fits on one line: workload = volume × average handling time, divided by the target occupancy rate, then increased to cover absences and non-productive time. Example: 180 five-minute calls in one hour represent 15 hours of workload, which means 26 people to schedule at 85% occupancy with 30% shrinkage. A spreadsheet is enough for a small team on a single channel; beyond that, forecasting software recalculates everything with every new data point.
A workforce forecast plan answers the question every team leader asks on Monday morning: how many people do we need, and when, to absorb this week's activity without queues or idle time? The term covers several uses, from project planning to financial forecasting. This article is about forecasting activity and staffing: how to build the forecast, a worked hour-by-hour example, the indicators that show whether it holds up, and the point at which a spreadsheet is no longer enough.
What is a workforce forecast plan?
A workforce forecast plan turns a workload forecast into a staffing requirement. It does not yet say who works when: it sets the target that the named schedule will then have to cover. The distinction matters, because most badly sized schedules go wrong at this stage, well before anyone is assigned to a shift.
It is often confused with two other documents that go by a similar name:
- The project plan: a Gantt chart that orders tasks and milestones over time, unrelated to recurring workloads.
- The financial forecast: the cash flow and revenue plan in a business plan.
- The workforce forecast plan: the expected volume of work per interval, converted into a number of people. This is the one used by customer contact centres, branch networks, after-sales departments and every business with variable workloads.
The three forecasting horizons
A single department actually runs three forecasts, which support different decisions and are not calculated at the same level of detail. Mixing them up means steering recruitment with yesterday's figures, or managing the day with an annual average.
| Horizon | Decisions made | Granularity | Review |
|---|---|---|---|
| Long term (6 to 18 months) | Budget, recruitment, use of temporary staff or outsourcing | Month or week | Quarterly |
| Medium term (4 to 12 weeks) | Shift patterns, leave, training, campaigns | Day, then half-hour | Weekly |
| Short term (D-7 to D) | Adjustments, reinforcements, reallocation between activities | 15 to 30 minutes | Daily, or even intraday |
The rule of thumb: the shorter the horizon, the finer the granularity. A forecast that is accurate for the day as a whole can hide severe understaffing between 10am and 11am, offset by overstaffing in the afternoon.
Building a workforce forecast in 5 steps
The method is the same whatever the tool, from a spreadsheet to specialist software. What the tool changes is the time spent redoing it every week and the precision of the calculations.
Step 1: set the horizon and the granularity
Start with the decision the forecast needs to inform. To size a team for the year, a monthly level is enough. To build next week's schedules, you need to go down to 15 or 30 minutes, because that is the scale at which queues build up. Define the scope too: which activities, which channels, which sites.
Step 2: analyse historical activity
The forecast is based on past data. Extract volumes from your telephony, ticketing or point-of-sale system, at the same time interval as the forecast. Ideally, have at least two years of data to read annual seasonality, and a few months at minimum for weekly and daily cycles.
- Trend: is activity growing or shrinking from one year to the next?
- Seasonality: busy months, back-to-school week, the end-of-year period.
- Cycles: Monday morning does not look like Thursday afternoon.
- Events: marketing campaigns, invoice runs, incidents, public holidays. Identify and isolate them, otherwise they distort the average.
Step 3: forecast the volume
Several methods coexist. Their accuracy increases with their complexity, and so does the time needed to keep them up to date by hand.
| Method | Principle | Suited to | Limitation |
|---|---|---|---|
| Moving average | Average of the same intervals over recent weeks | Stable activity with no marked seasonality | Reacts late to peaks and breaks in trend |
| Seasonal coefficients | Baseline volume multiplied by a coefficient per month, day and interval | Seasonality that repeats from year to year | Coefficients have to be recalculated by hand |
| Event-adjusted model | Seasonal forecast corrected for known events (campaigns, invoices, public holidays) | Activity driven by the sales calendar | Every past event must be documented |
| AI forecasting | Model that learns from historical data and readjusts with every new data point | Multichannel, multi-site, fine granularity | Requires a dedicated tool and clean historical data |
Step 4: convert volume into workload, then into headcount
You cannot schedule a volume, but you can schedule a workload. The conversion takes three steps:
- Workload = forecast volume × average handling time (AHT), including after-call work.
- Productive headcount = workload ÷ target occupancy rate. Nobody is busy 100% of the time: there is always some waiting between two contacts. A rate of 80 to 85% is a common target in customer service.
- Headcount to schedule = productive headcount ÷ (1 − shrinkage). Shrinkage covers leave, absences, training, meetings and breaks. Measure it on your own historical data.
For a contact centre working to a service level target (for example, 80% of calls answered within 20 seconds), the Erlang C formula refines the second of these calculations.
Step 5: compare the requirement with available resources, then make trade-offs
Overlay the requirement curve and the curve of people actually available, interval by interval. The gaps show up immediately, and each one calls for a different lever:
- Occasional understaffing: move breaks, bring shift starts forward, call on multi-skilled staff.
- Occasional overstaffing: schedule back-office work, training and meetings there instead of fitting them in at random.
- Structural gap: review shift patterns, use temporary or part-time staff, or escalate the issue to the long-term forecast.
Worked example: a one-day workforce forecast
Take a telephone customer service team with an AHT of 5 minutes, a target occupancy rate of 85% and shrinkage of 30%. The forecast volumes come from step 3. The figures are illustrative: replace them with your own.
| Interval | Forecast calls | Workload | Productive headcount (85%) | Headcount to schedule (30% shrinkage) |
|---|---|---|---|---|
| 8am to 9am | 60 | 5.0 h | 5.9 | 9 |
| 9am to 10am | 120 | 10.0 h | 11.8 | 17 |
| 10am to 11am | 180 | 15.0 h | 17.6 | 26 |
| 12pm to 1pm | 90 | 7.5 h | 8.8 | 13 |
| 2pm to 3pm | 140 | 11.7 h | 13.7 | 20 |
| 5pm to 6pm | 70 | 5.8 h | 6.9 | 10 |
How to read it: the requirement ranges from 9 to 26 people in the same day, almost a threefold difference. A schedule with a constant headcount of 17 would be understaffed by 9 people at 10am and overstaffed by 8 people at opening time. That is the whole point of the forecast: it shows where to place shift starts, breaks and back-office tasks. Headcounts are rounded up, because you cannot schedule half a person.
Measuring forecast accuracy
A forecast is only worth something if it is compared with what actually happened. Without that check, the same errors repeat week after week without anyone noticing.
- Forecast vs actual gap per interval: track it at the forecast's time interval, not just for the whole day, where errors cancel each other out.
- Mean absolute percentage error: the average of the absolute gaps, relative to the actual volume. It is the simplest indicator for comparing two methods or tracking an improvement.
- Bias: a forecast that is always wrong in the same direction (consistently too low on Mondays, for example) points to a missing factor, which is easier to correct than a random error.
Review the method when the gap worsens several weeks in a row, and document every exceptional event as it happens: it is the raw material for the next forecast.
Spreadsheet or forecasting software: when to switch?
A spreadsheet is still a good starting point. It becomes a bottleneck when the forecast has to be redone too often, across too many dimensions, by the one person who understands the file.
| Criterion | Spreadsheet | Forecasting and scheduling software |
|---|---|---|
| Collecting historical data | Manual exports | Direct connection to telephony, ticketing and HR systems |
| Calculation level | Hour or day; any finer and the file becomes unmanageable | 15 minutes, per channel and per activity |
| Updates | Redone with every new data point | Automatic recalculation, with manual adjustments possible |
| From forecast to schedule | Schedule built separately, by hand | Schedule generated from the calculated requirement |
| Suitable for | One team, one channel, stable activity | Several channels, sites or skills |
A spreadsheet also has hidden costs: time spent on updates, formula errors and dependence on a single person.
Software such as m-work Planner runs steps 2 to 5 in a single tool: it retrieves historical data through integrations with your telephony and ticketing tools, produces a forecast you can adjust by hand at 15-minute, 30-minute or hourly intervals, calculates the number of agents needed per channel and per interval, then builds the schedule that meets it. It takes into account both deferred and real-time workloads, the existing backlog and your opening hours. m-work Planner customers report saving 110 hours per manager per year on building schedules.
That is what the customer relationship centre at Crédit Agricole Charente-Maritime Deux-Sèvres did when it replaced its legacy planning tool: its volume forecasts are now accurate to nearly 99% over the last quarter. The full story is in the Crédit Agricole Charente-Maritime Deux-Sèvres customer story.
Building a workforce forecast means moving from a hunch ("we're short-staffed in the mornings") to a quantified requirement for each interval, and then to a schedule that covers it. The method takes five steps and you can try it today in a spreadsheet, using your own historical data. When the forecast has to be redone every week across several channels or sites, discover m-work's activity and workload forecasting software, or start with our free planning generator to turn your requirement into a first schedule.
FAQ
What is a workforce forecast plan?
A workforce forecast plan estimates the volume of upcoming work per interval and converts it into the number of people needed. It is used to size teams before building the named schedule. It is different from a project plan, which orders tasks over time, and from the financial forecast in a business plan.
How do you calculate the headcount needed from an activity forecast?
Multiply the forecast volume by the average handling time to get the workload in hours. Divide this workload by the target occupancy rate, often 80 to 85%, then by (1 minus shrinkage) to account for leave, absences, training and breaks. For example, 180 five-minute calls at 85% occupancy and 30% shrinkage require 26 people scheduled for the hour.
How much historical data do you need for a workforce forecast?
Two years of data give you a reliable read on annual seasonality. A few months are enough for weekly and daily cycles. Historical data should be extracted at the same time interval as the forecast, and exceptional events (incidents, campaigns, public holidays) should be flagged so they do not distort the averages.
How often should a workforce forecast be updated?
The long-term forecast, used for budgeting and recruitment, is reviewed every quarter. The medium-term forecast, which sets shift patterns and leave, is reviewed every week. The short-term forecast is recalculated every day, or even during the day, to reallocate teams when activity moves away from the forecast.
Can you build a workforce forecast in Excel?
Yes, for a team on a single channel with stable activity: a spreadsheet is enough to apply a moving average or seasonal coefficients and convert volume into headcount. It reaches its limits when you need to work at 15-minute intervals, manage several channels or sites, and redo the calculation every week. Forecasting software then automates data collection, the calculation and the building of the schedule.
.png)
