Contact center workforce management: forecast, schedule, adherence

Agent pay is the biggest line item in a contact center budget. A line that needs 11 agents from 9:00 AM to 9:00 PM, seven days a week, takes 45 people on staff. A scheduling error of just a few people adds up to several full salaries every month. Many teams still plan in a spreadsheet: copy last month's schedule, make a couple of edits, and hope for the best. Workforce management (WFM) turns this into a repeatable process. Small teams need it too: even with ten agents, the right schedule makes a visible difference at peak times.

The five-step cycle

1. Forecast: how many contacts will arrive in each interval and how long they will take.
2. Requirements: how many agents each interval needs to hit the target service level.
3. Schedule: shifts and breaks that cover the requirements with real people.
4. Intraday management: comparing plan with actuals and reacting to deviations.
5. Review: lessons from the gaps and adjustments to the next forecast.

Forecasting by interval

The forecast is the foundation. If it's wrong, even a perfect schedule won't save your service level.

• Use history in 15- or 30-minute intervals over several comparable weeks.
• Account for seasonality: day of week, month, holidays, and the start and end of the month.
• Add known events: ad campaigns, email blasts, price changes, and product launches.
• Forecast handle time as well as volume – it changes by hour and season.
• Forecast each channel separately: calls, chat, messaging, and email.
• Compare forecast with actuals every week and refine the method.

There are usually three forecasts. Long-term, months ahead, for hiring and vacations. A few weeks out, for schedules. A few days out, for fine-tuning shifts and breaks.

From forecast to requirements: the Erlang C formula

For voice queues, you plug each interval's forecast, average handle time, and service level target into the Erlang C formula. It tells you how many agents need to be on the line in that interval. Workforce management then turns those requirements into schedules and checks whether reality matches the plan.

Example: 120 calls an hour, an average handle time of 240 seconds, and an 80/20 target require 11 agents on the line. With 30% shrinkage, that's 16 people per shift for that hour. You can calculate any interval in the Erlang calculator, and we walk through the full calculation step by step in How many agents do you need.

In chat and messaging, an agent handles several conversations at once, so requirements are calculated differently. Your planning has to account for this concurrency instead of applying the voice formula directly.

Shrinkage

Agents aren't on the line for their whole shift. This lost time (shrinkage) is built into the schedule up front.

• Breaks, lunch, and rest.
• Training, call reviews, and meetings.
• Sick days, vacations, and other absences.
• Off-phone work: email, back office, projects.

Total shrinkage usually runs 25-35%, and every 5 points costs money. For the same line (11 agents on the line from 9:00 AM to 9:00 PM, seven days a week), 30% shrinkage means 45 people on staff, 35% means 48, and 25% means 42. Three extra people at 160 hours a month (as in the AI agent savings calculator) is about 7% of the line's payroll, month after month. So measure your own value by month and by team instead of taking it from a textbook, and look for what you can cut: meetings during peaks, unnecessary reviews, and long system outages.

How to build a schedule

The requirements curve is all peaks and valleys. If the midday peak needs 21 agents and the evening needs five, identical 8-hour shifts will leave you short one moment and paying for idle time the next. The schedule has to follow the curve while respecting labor laws, contracts, and people's needs.

• Mix shift lengths and start times: full shifts cover the baseline, short shifts cover the peaks.
• Put breaks and training in the valleys, not the peaks.
• Account for skills: not every agent can work every queue.
• Collect agents' shift preferences and allow shift swaps by clear rules.
• Publish schedules early so people can plan their lives.
• Check the schedule against requirements: overstaffing by interval is as visible as understaffing.

Weekly headcount depends on your hours of operation. If you need 11 agents from 9:00 AM to 9:00 PM, seven days a week, with 8-hour shifts and 5 shifts per person per week, you need 2 shifts of 16 people and 45 people on staff. The shift calculator works it out for your own hours.

Where the 45 people come from – and where idle time hides. The line is open 12 hours and a shift lasts 8, so you need 2 shifts a day. Each has 16 people: 11 on the line plus a buffer for 30% shrinkage (11 / 0.7 = 15.7). Over a week, that's 2 × 16 × 7 = 224 shifts; one person works 5 shifts, and 224 / 5 = 44.8 – so 45 people. But two 8-hour shifts cover 16 hours, and the line is open 12: for four hours, say 1:00-5:00 PM, the shifts overlap, and twice as many people are on the line as a flat calculation calls for. That's a quarter of all paid hours. If the call peak falls right in those hours, the money is working. If not, it's idle time. That's why shift start times are matched to the requirements curve instead of set from a template.

Schedule adherence

A schedule only helps if people follow it. Two metrics show whether they do.

• Schedule adherence – whether the agent was in the planned status at the planned time: on the line, on break, in training.
• Schedule conformance – whether the agent worked the planned amount of time in total, regardless of exactly when.

Poor adherence at peak times hurts service level the most: with 11 agents on the line, one agent late back from break cuts capacity by 9%, and service level in our example drops from 81% to 65%. Covering these dips with extra people is expensive: one more agent on the line all day means six more people on staff – about 13% more payroll. It's cheaper to deal with the causes. Talk about adherence with agents as a team tool, not surveillance. And look for systemic causes: breaks that are too short, schedules that ignore real life, slow systems.

Intraday management

• Every interval, compare actual volume, handle time, and agents on the line with the plan.
• Move breaks and training if the peak arrives earlier than forecast.
• Shift cross-trained agents between calls, chat, and back-office work.
• Absorb an unexpected spike with queue callbacks or a voice AI agent.
• Log every deviation and its cause – it feeds the next forecast.

Where to start

1. Gather interval data for the last two months: volume, handle time, and agents on the line.
2. Measure your actual shrinkage instead of guessing.
3. Calculate requirements for a typical week in the Erlang calculator and headcount in the shift calculator.
4. Compare them with your current schedule and find the intervals with the biggest gaps. Put them in money terms: every extra person on staff is one more full salary each month, and every missing one means abandoned calls.
5. Fix those intervals first, then move to a regular weekly cycle.

Don't try to build the perfect process in the first month. A simple weekly rhythm already gets you most of the benefit: on Monday, review last week's forecast accuracy and schedule adherence; on Tuesday, update the forecast for the next few weeks; by Wednesday, publish schedules; and every day at the same time, check how the current day is going. Keep a short log of what went wrong and why: an unexpected email blast, a system outage, a wave of sick days. In a few months, that log becomes your best source for forecasts.

A very small team can get by with a spreadsheet. But as agents, queues, and channels grow, a workforce management system saves the planner time and cuts errors.

Get the team involved. Agents who see the requirements curve and understand why breaks are where they are accept schedules more easily and swap shifts without leaving the line short. Supervisors who see plan and actuals side by side react before service level drops instead of explaining the dip after the fact.

Workforce management in CONTACT.WORK runs on the same call and inquiry data as your reports. The forecast, calculated requirements, schedules, and actual service level are all in one place.