Forecasting Call Volume for Small Teams: A Practical Method
A step-by-step way to forecast calls by day and half-hour from your own history, adjust for events and trends, and turn the forecast into a staffing plan.
Forecasting sounds like a job for a specialist team with expensive software. In reality, a small team can build a perfectly usable call forecast in a spreadsheet using the calls it has already handled. The aim is not to predict the future perfectly. It is to avoid the two common failures: too few people at the busiest times, and too many at the quietest.

This guide shows a simple method. After you have a forecast, Erlang C Explained Without the Math Headache shows how to convert it into the number of agents, and Shift Planning Basics shows how to place them on the schedule.
Why forecasting matters
Calls do not arrive evenly. They bunch on certain days and hours, and they spike after events like a billing run, a promotion or a service problem. If you staff for the average, you will be short at peaks and idle in quiet spells. Customers wait, agents get frazzled, and money is spent on empty time. A forecast tells you what to expect and when, so you can plan.
What a forecast helps you decide
- Staffing levelsHow many people you need in each part of the day.
- Shift patternsWhen to start and end shifts, and where breaks go.
- Leave and trainingWhen it is safe to schedule time away from the phones.
- HiringWhether volume is growing enough to justify more people.
Gather the right data
Start with history. You need, for each interval, the number of calls offered to your team and, ideally, how long they took. "Offered" means all calls that arrived, including those abandoned, not only the calls answered, since answered calls understate demand.
- At least eight to twelve weeks of data, more if you have seasonality.
- Intervals of thirty minutes or one hour.
- Calls offered, answered, abandoned, and average handle time.
- A list of unusual events in that period, such as outages, campaigns and holidays.
Clean the data. Remove or flag exceptional days, such as an outage that produced three times the normal calls, so they do not distort the pattern. Record the reason so you can use it later if something similar happens.
A simple five-step forecasting method
From history to forecast
- Build the weekly patternAverage calls for each weekday and interval across recent weeks.
- Adjust for trendIf volume has been growing or falling, lift or lower the pattern to match the recent weeks.
- Add known eventsPromotions, product launches, billing dates, school terms and holidays.
- Add handle timeMultiply calls by average handle time to get workload.
- Review weeklyCompare forecast with actual, learn and adjust.
Step 1: the weekly pattern
Make a grid with days across and half-hours down. In each cell, take the average of the last several weeks for that day and time. A Monday at 9:00 has its own shape, quite different from a Saturday at 9:00. This weekly shape is the backbone of your forecast.
Step 2: trend
Compare the latest few weeks with the older weeks. If calls are rising steadily, scale the pattern up. Resist reacting to one unusual week; look for sustained change. Remember that a change to your website, product or marketing often shows up in call volume.
Step 3: events and special days
Make a calendar of anything that is likely to change volume. Holidays often reduce calls on the day and increase them just after. A billing run can increase calls for several days. A new feature or an outage can create spikes. Where you have seen the effect before, add the same uplift. Where you have not, make a cautious guess and mark it for review.
| Event type | Typical effect | How to handle |
|---|---|---|
| Public holiday | Quiet on the day, busier before and after | Use the same holiday last year, if you have it |
| Billing or statement date | Raised volume for several days | Add an uplift to those days |
| Campaign or launch | Short spike, plus follow-up calls | Ask marketing for dates and expected reach |
| Outage or incident | Sudden spike, then slow decline | Staff up, set an IVR message and review afterwards |
| Seasonal period | Gradual rise and fall | Compare with last year's pattern |
Convert calls into workload
Calls are not the whole story. Ten long calls need more people than ten short ones. Workload is calls multiplied by average handle time, including talk time and after-call work. If you handle a mix of call types, forecast the main types separately where you can, since a billing call may take much longer than a store hours enquiry.
To get people from workload, you need to allow for the time agents are not available: breaks, training, meetings, sickness and leave. This allowance is often called shrinkage. A reasonable shrinkage figure is something you measure from your own team's history, not something you copy from a book. See Occupancy and Utilization for how busy agents should be.
A hypothetical weekly pattern
Imagine a six-person dental practice team that gets most calls on Monday mornings and after lunch each day. Looking at ten weeks of history, the manager sees that Monday between 9:00 and 10:30 is consistently the busiest period, Wednesday afternoons are quiet, and Friday after 16:00 drops sharply. Starting from that grid, the manager adds a note for the day that reminder messages go out, because the calls rise the following morning, and another for school holidays, when appointment requests rise.
The manager then checks the first forecast against the week's actual calls. Monday was close, Tuesday was lower than expected, and the reminder day produced a bigger spike than the plan allowed for. The manager raises that uplift for next time. After a few weeks, the grid is dependable enough to plan breaks, training and leave around it. The method never needed software, only discipline.
Measure accuracy and improve
Compare forecast and actual every week, by interval and by day. Look at the pattern in the errors, not just the size. If you are always short on Monday mornings, the pattern for Mondays needs lifting. If events are repeatedly missed, improve the way you collect them.
Forecast checks
| Daily total | Interval pattern | |
|---|---|---|
| What it tells you | Whether your overall level is right | Whether the shape of the day is right |
| Typical error | Missing a trend or an event | Wrong peaks, misplaced breaks |
| Action | Adjust the level for the next week | Reshape the profile for that day |
Tips for very small teams
- Use hourly intervals if half-hours give too much noise.
- Do not chase small differences. For a team of five, a forecast within a call or two per interval is a fine result.
- Combine similar days if you have little history, such as Tuesday to Thursday.
- Use callbacks and voicemail to flatten peaks; see IVR Callback and Voicemail.
- Track missed calls; the cost of missed calls is explained in The Cost of Missed Calls.
Forecasting in Call
Call's Workforce area builds a half-hour forecast for a chosen day from the call history of your inbound queues. Open Calls › Workforce, pick the day with the date picker and press "Rebuild forecast". It uses the last eight weeks of history and a standard shrinkage allowance, and reports how many intervals it built, or says there is not enough history for that weekday yet. A queue with too little history is skipped. The "Required vs scheduled" view then compares the forecast with the shifts you have scheduled.
Rebuild the forecast after a busy period so that it follows what really happened. The Supervisor view shows calls through the day, so a spike can be told from a steady climb, and the Reports area can export figures for your own spreadsheet. The built-in forecast is a starting point; add your own knowledge of events and compare it with actual results.
Frequently asked questions
How much history do I need to forecast call volume?
Eight to twelve weeks gives a usable weekly pattern. If your business has strong seasons, you will also want data from the same time last year.
Should I forecast in half-hours or hours?
Half-hours are standard for larger teams. Very small teams can use hours, since half-hour counts can be too small to be reliable.
What about new products with no history?
Use a similar product as a model, start with cautious estimates and review them daily at first. Add more cover during launch.
How accurate should a forecast be?
Accuracy improves over time. Focus on getting the shape of the day right and on noticing large misses early, rather than on perfect numbers.
How do I include callbacks and abandoned calls?
Count calls offered, including abandoned ones, because they are real demand. Add callbacks as workload when they are performed.
Do I need special software?
Not to start. A spreadsheet with your call history and a weekly pattern works well. Software helps when you have many queues or need to publish schedules.