A compliance team knew, roughly, when the pain was coming. The renewal calendar wasn’t a secret — certain months always brought a surge of licences up for renewal. What the calendar didn’t tell them was the thing that actually mattered: would the current review desk clear that surge inside the service-level targets, or would applications pile up until something breached?

Every year the answer arrived the hard way, in the middle of the wave, when it was too late to add capacity. They needed to know before, in terms they could act on: how many reviewers, for how long.

Forecasting the wave from the book

First we forecast the wave itself. Because the team held the active portfolio with expiry dates, the volume was largely deterministic: the model bucketed upcoming expiries by month and licence class, applied each class’s historical renewal rate, and produced expected renewals per period — with a confidence band. The surge that used to be a vague “March is bad” became a number, per category, weeks ahead.

Sizing the desk as a queue

Then that forecast fed the desk model. A renewal desk behaves like a queue — renewals arrive, each takes time to review, and a pool of reviewers works them down — so the peak period’s expected load went straight into Erlang-C queueing math. Out came the operational picture: how utilised the team would be, how long the slowest renewals would wait, whether that wait crossed the SLA line, and the reviewer headcount to keep it from doing so.

“A wave is coming” is a worry. “You’ll be two reviewers short in March” is a plan.

Category by category

Renewals aren’t uniform, so the model was run across categories — by licence class and region — each with its own load and its own service rate. That surfaced where the pressure would concentrate, rather than averaging it away, and produced a recommended seat count per category to bring p90 wait times back inside target.

From scramble to schedule

Instead of discovering the shortfall mid-surge, the team went into the renewal season with a staffing plan sized to the wave — reviewers allocated where the queue math said they were needed, and SLA-breach risk quantified rather than guessed. The annual scramble became a scheduling exercise.

Seeing the surge on a calendar was never the hard part. Translating it into “this many people, here, or these targets break” was — and that translation is exactly what treating the desk as a queue delivers.

Key takeaways

  • A known renewal surge was converted from a worry into a concrete capacity plan.
  • The desk was modelled as a queue: load and throughput in, wait times and SLA risk out.
  • Running per category surfaced where pressure would concentrate, not an average.
  • The team entered renewal season with reviewers sized to hold p90 waits inside SLA.