Measurement / calculator / Free to use

Conversion Delivery Reconciliation Calculator

Reconcile eligible, submitted, accepted, rejected and pending events. Expose unsent and unclassified gaps without confusing delivery coverage with ad attribution.

Growthcraft Editorial · 2026-09-17. AI-assisted research and implementation. Examples are synthetic; Akshay's personal review is not claimed.

Enable JavaScript to change inputs in the interactive calculator. The complete formulas, default example and limitations are available below.

Prepare comparable counts

Use whole-number counts from 0 to 1,000,000,000 for one event-time cohort, destination and snapshot cutoff. Eligibility is determined before submission by your approved business and data-use rules. Every submitted event must belong to that eligible cohort. Accepted, rejected and pending must be mutually exclusive current states, not cumulative attempts or repeated polls.

Keep the acceptance boundary explicit: accepted for destination ingestion does not mean matched to an ad user or reported as a conversion. Missing evidence belongs in unclassified, not pending. Where a provider supplies only request-level totals, do not invent event states. Use this calculator only when your aggregate counts are supported and deduplicated.

Formulas and zero-denominator handling

  • Not submitted = eligible E − submitted S.
  • Unclassified U = S − accepted A − rejected R − pending P.
  • Accepted coverage = A / E. Submission coverage = S / E.
  • Terminal acceptance = A / (A + R); it excludes every unresolved event.
  • Classified share of submissions = (A + R + P) / S.
  • Unresolved = (E − S) + P + U. Rejected is a known state, not an unresolved one.

The population identity E = (E − S) + A + R + P + U must hold. Rates with zero denominators are undefined and exported as null. The interface rounds percentages to two decimals; JSON preserves the underlying ratios. No statistical interval, projection, revenue value or causal effect is calculated.

Synthetic example and interpretation

For E = 1,000, S = 900, A = 720, R = 90 and P = 60, the tool returns 100 unsent, 30 unclassified and 190 unresolved. Accepted coverage is 72%; terminal acceptance is 88.89%; submission coverage is 90%; and 96.67% of submissions have a classified state.

The 16.89 percentage-point difference between the two acceptance percentages reflects different denominators, not a performance improvement. Begin with the 30 unclassified submissions because the record of delivery itself is incomplete. Then investigate the unsent queue and the age of pending events. Explicit rejections require reason-level review even when the ledger fully reconciles.

What this tool cannot verify

The calculator validates arithmetic, not identity, consent, processing maturity or provider semantics. It cannot identify which people matched, whether accepted records were attributed, or how much revenue an integration lost. A request record_count is not inherently a successful-event count; inspect the provider's definitions. Google Data Manager diagnostics.

Copy and download retain your non-sensitive scope label, inputs, formulas and raw results. Nothing is sent to an AI API by this calculator. Editing any input clears stale results. Reloading discards the input state.

Turn the numbers into an investigation

Use the ownership worksheet, the incident-review prompt, and the tested technical guide.

Companion resources

Back to strategic resources