FUNNEL COUNTS AND DENOMINATORS WORKSHEET Companion to: Read funnel results with the right counts and denominators All filled examples below are fictional. Blank fields are for your own records. This is an editorial worksheet, not an analytics product export or benchmark. REPORT DEFINITION — FILL THIS BEFORE THE COUNTS Report name: Report owner: Question this report should answer: Time zone: Entry window start (inclusive, full date and time): Entry window end (exclusive, full date and time): Group inclusion rule: Group exclusion rule: Visitor identifier definition and limitations: Visit/session definition: Submission attempt definition: New saved request definition and unique identifier: Repeat-attempt reconciliation rule: Follow-up qualification criteria: Follow-up duration and start point: Report cutoff (full date, time and time zone): Source/export date and record owner: Known coverage gaps or unresolved joins: COUNT SHEET Enter a count only when supported. Keep the state and reason separate. Suggested states: measured / missing / unknown / not applicable. A measured 0 means no qualifying events in the defined source and window. Measure | Count | State | Source and counting rule Distinct visitor IDs | | | Visits | | | Visitor IDs with an attempt | | | Submission attempts | | | New-save attempts | | | Repeat confirmations | | | Known failed attempts | | | Attempts with unknown outcomes | | | Distinct confirmed request IDs | | | Visitor IDs with a new request | | | Visits with a new request | | | Requests with a qualified conversation | | | Requests with completed talk: not a fit | | | Requests with follow-up still pending | | | Requests with missing follow-up status | | | Qualified requests: agreed step | | | Qualified requests: no agreement | | | RECONCILIATION Use mutually exclusive categories that cover the whole defined group. Attempt total: New-save attempts + repeat confirmations + known failures + unknown outcomes: Do those two totals match? If not, document the unresolved difference: Confirmed request total: Qualified + completed not-fit + pending + missing follow-up status: Do those two totals match? If not, document the unresolved difference: Qualified request total: Agreed next step + no documented agreement within the observation window: Do those two totals match? If not, document the unresolved difference: These categories fit the example only. If your workflow permits several conversations or outcomes per request, define a single status rule or retain separate event counts. Do not sum overlapping categories as a partition. RATE RECORD — COPY FOR EACH QUESTION Question / metric name: Numerator definition and unit: Numerator count and state: Denominator definition and unit: Denominator count and state: Same cohort and compatible observation windows? Evidence: For a percentage: is the numerator a subset of the denominator? Evidence: Excluded records and reasons: Unresolved records and effect on interpretation: Calculation (numerator divided by denominator; multiply by 100 for percent): Result, unit and rounding: Count pair to display alongside result: What this rate does not establish: Do not divide by zero. Mark an undefined rate as not applicable and explain. Do not calculate a rate from missing inputs. Leave it uncomputed with a reason. If units differ, name the result as an average or ratio, not a percentage of people. A visitor ID is not proof of one person across all their devices. FICTIONAL RECONCILED EXAMPLE — DO NOT MIX WITH YOUR COUNTS Entry window: 2026-09-21 00:00 inclusive to 2026-09-28 00:00 exclusive. Time zone: America/Chicago. Follow-up: first seven elapsed days after each confirmed request was saved. Report cutoff: 2026-10-05 00:00 America/Chicago. Counts below describe this made-up, complete visit log only. Two attempt outcomes and two request follow-up statuses remain unresolved as labeled. Group IDs Visits Attempts New Repeat Failed Unknown No attempt 54 64 0 0 0 0 0 One request each + retries 16 22 22 16 6 0 0 Two distinct requests each 4 8 8 8 0 0 0 One known failed attempt 4 4 4 0 0 4 0 One unknown outcome 2 2 2 0 0 0 2 TOTAL 80 100 36 24 6 4 2 Each fictional visit has at most one attempt. The 24 new saves happen in 24 different visits. Six retries happen in six additional visits and confirm existing request IDs; they never create another request or reset its clock. Attempt check: 24 new + 6 repeated + 4 failed + 2 unknown = 36. IDs with attempts: 16 + 4 + 4 + 2 = 26. IDs with confirmed new requests: 16 + 4 = 20. IDs with an attempt / all visitor IDs: 26 / 80 = 32.5%. IDs with new requests / all visitor IDs: 20 / 80 = 25%. Visits with new requests / all visits: 24 / 100 = 24%. New-save attempts / all attempts: 24 / 36 = 66.7% (rounded). New requests / all visitor IDs: 24 / 80 = 0.30 requests per visitor ID. The final calculation is an average, not the percent of visitors who submit. Qualification rule: completed project conversation confirming both service fit and an achievable timeframe, linked to the original request ID. Count each request once even if several conversations occur. Confirmed request follow-up: 12 qualified + 6 not a fit + 4 pending + 2 missing = 24. Qualified request follow-up: 5 agreed next step + 7 without that agreement = 12. Observed qualified share / all confirmed requests: 12 / 24 = 50%. Qualified share / requests with completed conversations: 12 / 18 = 66.7% (rounded). Agreed next step / qualified requests: 5 / 12 = 41.7% (rounded). The completed-conversation denominator is 18 request IDs. It excludes 4 pending requests and 2 requests with missing follow-up status. Agreement does not establish attendance, payment or delivery. Unknown attempt outcomes can revise confirmed totals. Record revisions; do not silently turn unknowns into failures or overwrite a dated report. FINAL REVIEW Can another person trace one request from its attempt to its follow-up? Does each rate display its numerator, denominator, units and window? Are missing/unknown/excluded records visible? Are original and revised report versions kept distinct? Reviewer and review date: Next record or definition to resolve: