How Accurate Is GTM Capacity Forecasting?
Compare a frozen workload forecast with observed demand, handling time and available capacity without hiding differences between work types.
Preserve the forecast before the week starts
Choose the operational queue, forecast horizon and unit of work. Save the forecast version before actual demand is known. Record expected arrivals, planned handling time by work type and available working hours after known absences or commitments. This is an internal measurement design, not a published benchmark.
Keep subsequent forecast updates as separate versions. Comparing actual results only with the final revised forecast conceals the uncertainty the original plan faced. Record the information available when each version was made.
Separate demand error from handling-time error
Compare forecast and actual arrivals for each work type. Then compare planned and observed handling time using the same definition of active work. Waiting for an approver or a customer should not silently enter an active-handling measure.
Review changes in the work mix. A stable total count can conceal a shift toward more demanding requests. Keep exceptional cases visible and document how incomplete time records are treated, rather than estimating a precise average from unreliable entries.
Connect the forecast to usable capacity
Compare expected workload hours with the hours actually available for that queue. Record interruptions, competing commitments and unplanned absence separately from demand. Include backlog carried into and out of the period so completing old work is not mistaken for new arrivals.
Show signed differences to reveal consistent over- or under-forecasting, and absolute differences to show error size. Explain units and denominators. Percentage errors need special handling when the actual or forecast quantity is zero; show the underlying counts instead of a misleading percentage.
Test an adjustment against a later period
Illustrative review: the total request count is close to forecast, but launch-related requests take longer than routine updates. The planner separates those work types and revises the handling-time assumption rather than adding the same allowance to every task.
Evaluate the revised assumption in a later window using the same definitions. Record unusual launches or staffing changes that affect comparability. Use the result to improve planning and escalation thresholds, not to rank individuals from an aggregate queue estimate.