Customer service staffing: when enquiries really arrive and how many people you need

Most customer service schedules follow the rhythm of the company rather than the rhythm of customer questions. One person starts at eight, another finishes at four, and lunch falls when it falls. Customers do not consult your internal schedule—they write when they need something. If these two curves are two hours apart, you are overstaffed for most of the day and have a queue for two hours that nobody knows about.

Build an hourly distribution of enquiries before creating the rota

An hourly distribution is one simple table: the number of new cases that arrive in every hour of every day of the week. Not the number of messages or website visits—the number of moments when someone starts a conversation, submits a form or requests contact. That figure determines how many people need to be available, because every new case requires someone’s attention from the beginning.

1. Collect data from the past 8–12 weeks.A shorter period leaves too much to chance; a longer one mixes different seasons and promotions.
2. Count new cases, not messages.A conversation containing thirty messages is still one case to handle.
3. Break the data down by hour and day of the week.Use a seven-by-twenty-four grid. A daily split alone will show you nothing.
4. Add each channel separately.Chat, callbacks and forms have different distributions and different handling costs.
5. Set unusual days aside.An outage, a newsletter send or Black Friday—analyse them separately.
6. Plot the current rota on the same grid.Only placing both curves together shows where there is a gap and where there is surplus capacity.

Averaging is the biggest trap. The average number of enquiries per hour across a month is a figure you never experience in practice—for half the day it was quieter, and for two hours it was much busier. Planning around the average guarantees that you will always be short-staffed at the peak, even though ‘everything adds up’ in the report.

There is no point looking for somebody else’s benchmark for hourly distribution. It depends so heavily on the industry, traffic sources and customer groups that the only reliable data is your own from the past several weeks.

Peaks are rarely where you expect them

Intuition suggests that most questions arrive in the middle of the working day. In practice, the distribution usually has two peaks, both in places the rota does not cover. The table below collects patterns worth checking first—not so that you accept them on faith, but so that you know what to look for in your own data.

↔ Scroll the table sideways to see all columns.

Campaign effects deserve separate attention because they are the only peaks you know about in advance. A mailing or promotion launch is a planned event, yet it rarely leads to a staffing change that day. If marketing knows about a campaign a week in advance, customer service can know just as early. Before Black Friday, use our seven-day plan for getting support ready for a campaign, which covers promotion terms, saved replies and a test of the customer path alongside the rota.

It is also worth separating peaks by channel. Live chat usually has a sharp curve tied to website visit times, forms are distributed evenly around the clock, and callback requests accumulate at the edges of the working day. Three curves on one chart usually cancel each other out and produce a picture that tells you nothing.

People per hour—estimating without call-centre formulas

Contact-centre literature offers queueing models that calculate staffing from probability distributions. They make sense for hundreds of contacts a day and teams of dozens. For a dozen or a few dozen cases a day, they are excessive: they require assumptions you cannot verify and offer false precision. A simpler calculation is enough as long as you track three figures.

• Number of new cases at the peakUse the worst hour in a typical week, not the average across the whole day.
• Handling time per caseFrom the first response to closure, including the note and any handover.
• An operator’s real availabilityA clock hour is not an hour spent on conversations. Leave a margin.
• Number of cases handled concurrentlySeveral chat threads can be handled at once, but not every type of case and not all the time.

The calculation works like this: multiply the number of cases in the peak hour by the average handling time per case, divide by the time an operator can actually spend on conversations during an hour, then divide the result by the number of cases one person can handle concurrently. The last step is the one most often forgotten—and the one you must not stretch, because concurrency is bought at the expense of response quality. How to work out how many chats one person can run in parallel, and test that number, is covered in our guide to concurrent chat limits.

In this example, two people is the minimum, not a comfortable level—cases do not arrive evenly every five minutes, but in bursts. A practical rule: round a result just above a whole number up; if it is just below, check whether a second person can do something during that hour that can be interrupted without loss.

Handling time per case is the only one of the three figures you can materially influence without hiring. Good routing to the right department, an organised contact history and relieving the team of repeated questions through a website knowledge base and FAQ all shorten it. You should separately ensure that shorter handling does not come at the expense of customer waiting time—we explain how to distinguish the two in our article on first-response time in chat.

A one-person team and working in duty blocks

In a small company, customer service is usually handled by someone who has four other responsibilities. Expecting them to remain continuously available for eight hours is unrealistic and costly elsewhere: every interruption of focused work to answer a chat costs more than the answer itself. The solution is not responding faster, but replacing continuous availability with duty blocks.

• Set two or three duty blocks each day and place them precisely at the peak hours in your own distribution.
• Do not schedule anything else during a duty block—it is work, not a background activity.
• Outside those blocks, configure availability so that chat switches to offline mode with a form.
• State exactly when the case will be picked up in the offline message instead of saying ‘we will be in touch soon’.
• Begin the first block of the day with cases left overnight, not the current chat queue.
• Leave one short closing block at the end of the day so that nothing carries over unanswered.
• Decide what happens on the sole team member’s day off—a substitute or an explicitly closed channel.

This change looks like reduced availability, but usually improves actual service time. A channel open for eight hours with a forty-minute response time is worse than one open for three hours with a two-minute response time. Customers tolerate a closed door much better than an open door with nobody behind it.

It helps to understand how the system defines availability. In Meetlead, an operator is considered available only when the panel has an active connection, the role is appropriate and the current time falls within the configured schedule. A logged-in panel running in the background is not enough for the channel to begin accepting conversations—and rightly so, because the widget does not promise a presence that is not there.

Once there are several people, the same mechanism works in reverse: instead of everyone working at once throughout the day, create overlapping shifts with extra coverage at the peak. Conversations are then routed to departments using round-robin or current-workload strategies, so nobody picks up three difficult cases at once simply because they clicked first.

Growing the team looks like the obvious solution, and it becomes one only after a new person has been onboarded well. We describe a first-week plan in our article on onboarding a new customer service operator.

When work is intermittent, the entire arrangement depends on notifications—and those are what fail most often. We discuss them in our article on customer service notifications.

The edges of the day and weekends: offline mode instead of silence

The edges of the day are the hours when enquiries still arrive but the team is no longer there. A case then sits until morning without any feedback. The point is not to extend working hours—it is to give the customer an honest answer instead of silence at the edges of the day.

• Honest informationA message with a realistic return time, not an assurance of a quick response.
• A form that creates somethingAn offline message should create a case rather than enter a general inbox.
• Callback requestThe customer chooses from available slots instead of waiting for an unexpected call.
• Morning queueOvernight cases have their own block at the start of the day rather than competing with current chats.

A callback is the strongest tool here because it moves contact from an hour when nobody is present to one when somebody is. Server-generated slots prevent customers from selecting a time you cannot cover, and availability outside working hours is configured separately—so you can accept requests in the evening and call back in the morning. We explain how to organise this in practice in our article on website callbacks.

Cases from the edges of the day and weekends should be kept in one place with a clear owner and status rather than in a shared inbox. Otherwise, Monday morning begins with working out who has already read what. We show how to build this flow without an elaborate system in our article on a customer-service ticketing system.

For a team of several people, some of these mechanisms are excessive. We describe the minimum sufficient set in our guide to customer service in a small business.

An absence affecting the entire team is a separate case—a holiday or illness can close the channel for a week in a small team. We cover a plan for this situation in our article on customer service during absences.

Review staffing every month

The distribution of enquiries changes with the season, a shift in traffic sources, expansion into a new market or a website change. A rota set once stops fitting after a few months, usually without a clear signal. A short monthly review is enough to catch it.

1. Refresh the hourly distribution.Use the same seven-by-twenty-four grid with data from the past four weeks.
2. Overlay the actual staffing.Not the planned staffing, but the level resulting from holidays, sickness and shifts actually worked.
3. Find the month’s three worst hours.The hours with the largest difference between the number of cases and available people.
4. Check how many cases arrived outside the schedule.If the share grows month by month, the boundaries of the working day need adjustment.
5. Verify handling time per case.An increase of a few minutes changes required staffing more than a rise in volume.
6. Change one thing in the rota.One change a month lets you say what worked. Three changes at once do not.

This review takes a few dozen minutes and marks the point where staffing stops being a matter of instinct. It does not require forecasting models—only that somebody place two curves side by side once a month and make one decision.