Teams of five to fifteen people often schedule by feel and by last week. That works until it doesn't: a campaign, a delivery problem or a wave of colds, and the queue grows while everyone works flat out. The method below works for small teams and large contact centres alike, and you don't need tools that cost more than the team.
The short version: calculate staffing per hour and per channel, not as a daily average.
Why an average isn't enough
Tickets don't arrive evenly. A team taking 300 tickets a week averages 7.5 an hour during office hours. But Monday at 9 might bring 25 and Thursday at 15 might bring 3. Staff for 7.5 and you get a queue on Monday morning and idle time on Thursday afternoon. Customers only remember Monday.
Step one is to pull inflow per hour, weekday and channel for at least four weeks back. Every support tool has the report. The curve differs by business: e-commerce peaks on Sunday evenings and Mondays, service companies around invoice dates, SaaS after releases.
Phone and chat: use queue maths
In real-time channels, response time depends on how many people are free at the moment the customer gets in touch. Contact centres have used the Erlang C formula for this for about a century. It gives the probability that a customer has to wait, based on the load (contacts per hour × handling time) and the number of agents. You don't need to compute it by hand, because calculators and most scheduling tools have it built in. What it shows:
- Small teams get hit hardest. Ten calls an hour at six minutes each is a load of one person. But for eight out of ten calls to be answered within a minute, you need three people.
- The response time target drives cost. Going from two minutes to 30 seconds can mean one more person in a small group. Choose the target deliberately.
- The formula assumes infinite patience. Real customers hang up, so track the abandonment rate as a separate metric.
For chat, count the concurrency you actually measure, not the one the vendor promises. Two or three chats per agent is common, and quality drops above three.
Email and forms: throughput against a target
Nobody hangs up on an email, so the maths is simpler:
| Step | Calculate | Example |
|---|---|---|
| Inflow per day | Tickets per weekday | 60 emails on a Monday |
| Handling time | Minutes per ticket incl. lookups | 8 minutes |
| Work per day | Inflow × handling time | 480 min = 8 hours |
| Capacity per person | Productive hours after shrinkage | 5.5 hours |
| Need | Work ÷ capacity | 1.5 people on Monday |
The response time target decides how much can carry over. A four-hour target means the hourly curve applies to email too. A next-business-day target lets Monday's peak spread into Tuesday.
Shrinkage: why schedules are wrong without it
Shrinkage is the paid time when agents can't take tickets: breaks, meetings, training, system trouble, sick leave and holiday. Someone scheduled for eight hours doesn't take tickets for eight hours. Measure your own shrinkage for a month rather than guessing, and split it into planned (schedule it outside the peaks) and unplanned (keep it as a margin). In small teams, one sick person can be a fifth of your capacity.
Self-service and AI: subtract what you've measured
Self-service removes simple contacts, but only when it actually works. Gartner found that only 14 percent of customer service issues are fully resolved in self-service. So subtract what you've measured it resolves per ticket type, not what it ought to. Also expect the remaining tickets to be harder and longer, so staffing falls less than ticket count.
Out-of-hours expectations are usually better met with an AI chat that answers from reviewed knowledge and hands over the next morning than with evening shifts. Test in one channel before building shift work. If you want to put rough numbers on the trade-off, an ROI calculator is a good starting point.
Checklist
- Pull inflow per hour, weekday and channel for four weeks
- Measure handling time per ticket type, using the median if a few tickets are very long
- Run real-time channels through an Erlang calculator, and test a stricter and a milder target
- Calculate email and forms as daily throughput against the target
- Measure shrinkage and add it on
- Subtract measured self-service resolution, and allow for longer handling of what's left
- Fit the schedule to the curve and review response time and abandonment per hour weekly
Recalculate every quarter and before every known peak.
Originally published at https://www.supportifier.se/en/blog/customer-service-staffing-calculation
Top comments (0)