Define a consistent HVAC quote-conversion cohort before reporting

Write the metric definition before reading the percentage

Choose the sales question first. For a sent-quote cohort, define the sending period, unique commercial unit being counted, included quote types, revision treatment, approval evidence and observation cutoff. Ask AI to organize a definition card and identify records that do not fit. Let the reporting owner decide those rules rather than silently accepting a model’s interpretation of conversion.

This is different from assigning opportunity stages. It is a reproducible counting method for a specific report. Record whether the unit is a quote version, an opportunity or another approved entity. A revised PDF should not automatically become another potential sale simply because it has a later sent date.

Compare two fictional denominators

A fictional cohort contains 20 unique opportunities first quoted in a defined week. At the stated observation cutoff, eight have verified approvals, six have verified declines and six remain open. Approved divided by sent is eight divided by 20, or 40%. Approved divided by decided is eight divided by 14, approximately 57.1%. Neither figure can be substituted for the other without changing the question.

Keep the six open cases visible. Do not label them lost just to close the denominator, and do not remove them from a sent-cohort metric because their eventual outcome is unknown. Where two quote revisions belong to one opportunity, apply the approved counting rule and retain the revision links for audit.

Check comparability before describing a change

Give each row a stable identifier, cohort inclusion reason, source dates, observed outcome and unresolved issue. Recalculate totals outside the assistant. Show exclusions and missing outcomes alongside the percentage, including how you treated withdrawn or superseded quotes. Preserve the observation cutoff so a later approval does not appear to have been known earlier.

When comparing periods, use the same definition and comparable observation windows. A younger cohort has had less time to reach a decision. Changes in customer mix, scope or pricing also limit interpretation. Do not attribute a conversion change to AI messaging from this table alone. The assistant may draft a qualified explanation but must not update pipeline outcomes or invent approval evidence.

Review checklist

• Define the counted entity and revision rule.
• State sending period and observation cutoff.
• Show pending cases, exclusions and missing evidence.
• Use matching definitions and observation windows for comparisons.

Source context and limits

Jobber documents quotes and their workflow states. Our fictional cohort definitions and arithmetic are an independent reporting method, not a claim about the default calculation in any Jobber report.