A single 'average response time' can hide several different waits. A customer acknowledgement, a qualification decision and a finished quote are different events. If their clocks are mixed, the number cannot tell the team where work is actually waiting.
The measurement sheet below separates those events. It can be used in Excel, Google Sheets or a CRM export and does not assume that every business needs the same target.
Define four clocks before measuring
First-human-response time runs from enquiry received to the first substantive human acknowledgement. Do not count an automatic receipt message unless it answers a real question or states a reviewed next step. Qualification time ends when the team records fit, missing information and ownership.
Quote cycle time runs from the agreed starting event—usually qualification-ready—to quote sent. Follow-up delay runs from the dated next action to the completed follow-up. Store each timestamp separately instead of compressing the process into one duration.
Use a timestamp contract
For every eligible enquiry, retain a stable ID, source, received time with timezone, first-human-response time, qualification-ready time, quote-sent time, owner, work type, market and outcome. Record the business-hours calendar used for the calculation.
Choose one rule for messages received outside working hours. Either measure elapsed clock time or pause the clock until the next working period; never switch methods between records. Preserve the raw timestamps so the calculation can be audited later.
Report median, P90 and SLA attainment together
Median is the middle duration after records are sorted. P90 is the duration within which 90 percent of eligible records were handled. SLA attainment is eligible records completed within the team's declared target divided by all eligible records.
Keep the arithmetic mean as a secondary diagnostic, not the headline. One abandoned or weekend-spanning enquiry can pull the mean away from the typical case. Median describes the middle; P90 exposes the slow tail; SLA attainment shows consistency against the operating promise.
Make exclusions visible
Test submissions, duplicates, spam and withdrawn enquiries may be excluded only with an explicit reason. Never delete a slow record merely because it makes the report look worse. A missing timestamp is its own quality metric: missing eligible records divided by all eligible records.
Segment only where the operating path differs—such as web versus shared inbox, service versus project work, business hours versus after-hours, or UK versus US routing. Small segments should be shown as small samples rather than treated as reliable benchmarks.
Turn the report into a weekly operating loop
Review the oldest open enquiries first, then records without an owner, records without a dated next action and records beyond the declared target. Sample the slowest cases and classify the cause: missing input, unclear ownership, capacity, approval, supplier dependency or integration failure.
Automate only the repeated, observable cause. Capture and routing can often be automated; pricing, technical feasibility and commercial commitments remain reviewed human decisions. Compare later periods only after the event definitions and calendar are stable.