Forecasting ticket volume without a data science team
Ticket volume is more predictable than most teams think — you don't need a data-science team to forecast it, just a spreadsheet and the discipline to stop staffing to the average.
You already know Monday is going to be rough. You knew it last Monday, and the one before that. Yet the roster you built assumes an ordinary day, because "ordinary" is the number your reporting tool hands you — tickets per day, averaged. So Monday lands, two people are out, the queue backs up by ten, and everyone calls it an unusually bad day. It wasn't unusual. It was Monday.
Forecasting inbound volume has a reputation as a data-science problem: a warehouse, a modelling team, a Python notebook. It isn't. You need a spreadsheet, a year of history, and the discipline to stop staffing to the average. The pressure to get it right is only climbing, because customers have quietly reset what they'll tolerate.
The average is the wrong number
The mean is a comfortable lie. It describes a typical interval while telling you nothing about the intervals that actually break you. If inbound swings from 40 on a quiet Thursday to 140 on a post-release Monday, the average — call it 75 — describes a day that almost never happens. Staff to it and you are overstaffed in the troughs, underwater on the peaks, and wrong most days in between.
Where this bites is occupancy: the share of an agent's paid time spent actually handling contacts. Staff to the mean and, on your busy intervals, occupancy pins to the ceiling.
That 83.3% is the number practitioners themselves pick as a sensible maximum, and it is already high. Push past it and you don't buy more throughput — you get a slower, more error-prone team, because there is no recovery time between contacts.
Above ninety percent, burnout stops being a risk and becomes the plan. And burnout carries a price tag you can read straight off the staffing sheet.
Agents who spend every interval underwater don't stay. Turnover near a third of the floor each year isn't only a hiring cost; it drags first-contact resolution and CSAT down with it, which pushes volume back up, which raises occupancy again. Staffing to the average is how that loop starts.
Staffing to the average leaves you overstaffed when it's quiet and underwater when it counts — wrong in both directions, most days of the year.
Seasonality is the pattern hiding in plain sight
Here is the good news: ticket volume is one of the most predictable things a support team owns. It moves on rhythms you already feel — the day-of-week curve, the time-of-day curve, the monthly billing spike, the seasonal swell, the reliable bump after every release. A data scientist would call this seasonal decomposition. In a spreadsheet it is three columns.
Take a year of daily volume. Compute a trailing average to get the baseline trend. Then, for each day of the week, divide its actual volume by that baseline to get a seasonal index — Monday might run 1.4x, Saturday 0.5x. Do the same by hour, and by month if you bill or renew on a cycle. Multiply your trend forecast by the relevant indices and you have an interval-level forecast that respects the peaks instead of averaging them away. Overlay your known calendar — launches, marketing sends, invoice runs — because those aren't noise. They are the events you can see coming.
This won't be perfect. It doesn't need to be. A forecast that's roughly right about when the peaks fall beats a precise average that pretends they don't exist.
Staff to the shape, not the mean
Once you can see the curve, staffing becomes a coverage question rather than a headcount one. Cover the peaks to the occupancy target you actually want to hold — not the ceiling — and let the troughs absorb the slack with training, QA, and knowledge-base upkeep, which is real work that flexes around live volume. Build in shrinkage honestly: the schedule is made of people, not seat-hours.
And the shape now runs around the clock. When most of your customers expect help to be there whenever they need it, the overnight trough still needs a floor under it. Forecasting tells you how thin you can safely go — not whether you can go dark.
None of this requires a model you can't explain to the person building the roster. It requires looking at the history you already have and refusing to flatten it. If you want a defensible occupancy target to build against, start with the benchmarks; if the vocabulary — occupancy, shrinkage, seasonal index — is new to the team, the glossary is the faster on-ramp.