Capacity and overbooking: hard limits, deliberate exceptions

What does the capacity model do?

It makes “how full is this session?” a question with exactly one answer, and makes every exception to a limit deliberate.

• Per-occurrence limits. Capacity lives on each scheduled occurrence, not just the parent session, so one date can differ from the rest of its series.
• Live counters. Booked, waiting, and attended counts are maintained on the occurrence and updated in the same transactions that change them. Dashboards read them directly; nothing is recounted or synced.
• Sub-capacities. Separate limits for trial places and make-up places let operators ring-fence capacity for prospects and catch-ups without eating regular places.
• An explicit overbooking allowance. Where no-shows are predictable, staff can set an overbooked capacity per occurrence. The system distinguishes policy (a configured allowance) from defect (a race), and prevents the second architecturally.
• Serialised checks. The capacity check and the write that consumes a place happen under a per-occurrence lock, the design detailed in checkout reservations. Demand beyond the limit flows into the waitlist atomically.

Why per-occurrence capacity matters

Real schedules are not uniform. A weekly class moves to a smaller room for one date; a workshop runs a double session before a holiday; a trial day needs five ring-fenced places. Systems that model capacity only at the class level force workarounds for all of these, and workarounds are where double-bookings live. Modelling capacity per occurrence, with its own counters and its own overrides, makes the exceptions ordinary configuration.

Why it matters commercially

Capacity is revenue with a ceiling. Undersell it and places go empty; oversell it and refunds and apologies follow. A model with hard per-occurrence limits, live counters, ring-fenced sub-capacities, and deliberate overbooking allowances lets operators run sessions at their true limit with confidence. The companion guide below covers the contention engineering that keeps those limits honest under high demand.