Ticket-volume forecasting
Start with a rolling 8-week average of weekly ticket volume by category. Apply a peak factor (typically 1.4–1.7) to the peak hour to get from weekly average to hourly peak. That number, divided by tickets-per-agent-per-hour, gives your hour-by-hour requirement. The forecast is not exact, but it is closer than the alternative. Over a quarter, the forecast and the actual usually agree within 10% — well within the noise of staffing decisions.
AHT — average handle time
Most teams underestimate AHT by ignoring wrap-up time. Measure end-to-end: open → resolve → wrap-up notes. AHT × forecast volume = required agent-hours. Round up, not down. For a typical IT helpdesk, AHT runs 12–25 minutes per ticket; for a customer-service helpdesk, 8–15 minutes; for a regulated case workflow, 45 minutes to several hours. The category-level AHT matters more than the overall — a single average hides the variation that drives staffing.
Shrinkage is the missing 20%
Annual leave, sickness, training, breaks and meetings — typically 20–25% of paid hours. If your forecast says you need 5 agents, schedule 6. Without a shrinkage allowance, every plan breaks on the first quiet week of January when half the team is on leave. Build the shrinkage into the forecast at the start, not after the staffing gaps show up.
Worked example — a 200-employee IT helpdesk
200 employees generate roughly 0.4 tickets per employee per week = 80 tickets per week. AHT is 18 minutes including wrap-up = 0.3 agent-hours per ticket. Total agent-hours required per week = 24. After 25% shrinkage, gross agent-hours = 32. At 35 hours per agent per week, that is 0.9 FTE — round up to 1 FTE plus on-call cover from a senior engineer for peaks. The same maths at 600 employees gives 2.7 FTE; round to 3.
Channel mix matters
Phone, chat, email and self-service portal each have different AHT profiles and different peak patterns. Chat is high-volume, short-AHT, peak-heavy. Phone is lower-volume, higher-AHT, more even through the day. Email is asynchronous and absorbs peaks. The staffing model should consider channel-by-channel demand, not just total tickets, because the peak-hour requirement differs by channel.
Self-service and deflection
A well-built self-service portal deflects 30–50% of repeat questions in most environments. The deflection is the highest-leverage staffing lever available — every deflected ticket is an avoided 18-minute interaction. The investment in the portal usually pays back in staffing within a year, and the deflected tickets are also the lowest-value tickets for the agents to handle, so morale improves alongside the headline number.
When to hire vs when to optimise
If the forecast says you are short-staffed by 0.5 FTE for two consecutive quarters, hire. If the forecast says you are short-staffed at peak hour but fine on average, optimise the shift pattern. If the forecast says the AHT is climbing across all categories, investigate the cause (often a knowledge-base gap or a recent system change) before hiring.
Helpdesk staffing is forecastable inside 10% with three numbers and a peak factor. Beats vibes every time — and the conversation with finance about headcount is materially easier with a defensible model behind it.