When to hire: check the capacity inputs first

Put matching labels on both sides

Name the role or team, period and included work. A fictional boundary: "Remote agreement support in September; separately staffed projects excluded from both availability and demand."

Available hours are the time left after the commitments excluded under that boundary. Workload is demand assigned to that remaining time. Count a commitment once. Either exclude project bookings and project demand from both sides, or include both. Subtracting the bookings and then adding their demand creates a false shortfall.

Trace availability to actual staffing, leave, training, site visits and management duties. Keep overtime separate from the planned schedule. Trace demand to support, escalation, documentation, vendor coordination and other duties in scope. Adam Hannemann's time-tracking discussion concerns that effort visibility.^[1] A ticket count cannot substitute for hours.

A short calculation, then a separate coverage check

In a fictional month, 600 available hours minus 610 workload hours gives −10 hours of aggregate headroom. If a proposed improvement releases 11 recurring hours, the scenario becomes +1 hour. Eight setup hours leave −7 in the first period.

Those figures answer one question about total effort. They leave peak coverage, skills, absence resilience and client growth unresolved. Ask when commitments were missed, which capability was unavailable and whether the same constraint persisted across comparable periods. A project spike deserves a different plan from recurring specialist demand.

The MSP KB describes duties across MSP roles.^[2] Write the proposed role around the missing duty. A general support hire might reduce routine work while leaving an owner-bound dispatch process unchanged.

Where to find the evidence

Use PSA time records and booking calendars, not only the dashboard's utilization tile. Autotask, ConnectWise PSA and HaloPSA are examples; verify your version's export fields and exclusions. Autotask documents separate ticket/task references, work dates, hours and nonbillable status.^[3] Include non-invoiced work when it is inside the review scope, and check that permissions have not filtered it out.

RMM alerts can help find repeated faults to investigate. An alert count has no dependable conversion to technician hours; join the resulting work to service records and account for work done outside tickets.

Keep scenarios out of today's availability

Existing automation already affects current workload, so it receives zero new scenario credit. For a proposed change, confirm the work still exists in the baseline and another initiative has not claimed it. Deduct review, exceptions and maintenance from the saving, then charge setup once in the first period.

Missing workload is unknown, not zero. Better time capture can increase recorded hours without increasing demand. Record that measurement change before comparing months. A promising scenario still needs a test and a usable time window before it becomes coverage.