Reconciling conversion figures across Google Ads, GA4 and CRM

Name the compared events precisely

Record which action each figure represents. An analytics event, an event marked important and a conversion created for advertising may serve different purposes. CRM may count only a business-accepted enquiry. A report labelled “leads” does not explain those distinctions. Create a small dictionary covering trigger, counted object, source and status. Google documents differences between Analytics key events and advertising conversions; your business still needs to define and maintain the meaning of its own sales stages. Preserve the accepted definitions so that later analysts compare the same concepts rather than familiar-looking labels.

Align period and reporting clock

State date range, time zone, freshness and the event basis of each report. A person can click on one day, enquire later and be qualified in CRM later still. Reports may therefore place the same sequence in different periods. Compare an appropriately mature cohort alongside a calendar report. Record export dates and whether further events may arrive. A difference caused by reporting clocks is not repaired through a new tag configuration. It needs an understandable analysis rule and visible context in the report. Separate late arrival of data from genuine delay in buyer evaluation. Explore this further in the guide Conversion lag: manage B2B campaigns by signal maturity.

Which clock belongs to the decision?

A short-term operating check may ask about received events, while channel evaluation needs a mature acquisition cohort. Define the question before choosing a column rather than retrospectively declaring the most favourable view correct.

Distinguish assignment from counting

Check which contacts each system assigns to advertising and how repeated actions by the same person or company are counted. An account opportunity is not one website visit. Several members of a buying committee do not automatically represent several new sales opportunities. Record counting and deduplication rules before adding totals. Models may legitimately offer different perspectives. They do not justify reporting the largest available number as total performance. For a budget decision, state which observation and assignment rule support the conclusion, and disclose events that cannot be linked reliably under the approved measurement process.

Inspect the route with a controlled case

Choose an approved test case and record expected stages from website event through technical processing. Use suitable testing procedures and respect data and access rules. Check trigger, repeated transmission, processing status and actual goal configuration. A positive technical signal does not establish business-qualified demand. Conversely, absent assignment does not automatically mean that the person failed to perform the action. Record what was observed directly and which connection remains assumed. The issue then becomes actionable for an accountable specialist rather than a general complaint that the dashboard does not match another system.

Separate expected differences from defects

Classify findings as different definition, clock, assignment, coverage or a verified technical fault. Label a suspected cause as a hypothesis and state the additional check that could confirm it. Avoid turning every unexplained difference into the same blanket reimplementation project. Technical corrections need their own acceptance test. Changed reporting definitions need a version note. Disclose an unresolved remainder so that a later configuration change does not silently present it as improved campaign performance. A plausible explanation should remain open to challenge until the relevant observation supports it rather than merely sounding reasonable.

Publish a comprehensible reconciliation note

The shared report contains starting counts, their meanings, established differences and open investigation tasks. Assign an owner and identify the decision restricted pending each clarification. A dashboard should not imply greater precision than the reconciliation permits. Avoid one adjustment formula for causes that differ. Where reports answer legitimately different questions, retain them side by side. Marketing, analytics and sales operations should be able to understand the explanation without replaying the original analyst conversation every time. Preserve the evidence behind the classifications so that a future owner can reproduce the same bounded conclusion.

How is unknown origin retained?

Show the unknown group separately with its definition. Do not allocate it arbitrarily across known channels. Improved later assignment needs defensible permitted evidence and a documented change to the analysis rather than an unexplained balancing adjustment.

Fictional example: three figures with different jobs

An explicitly fictional team sees twelve form events, nine attributed ad conversions and four accepted CRM enquiries. These invented figures illustrate the method and are not benchmarks. Reconciliation establishes that the first counts technical actions, the second channel attribution and the third a business stage. The team also finds one duplicated form transmission. It repairs that defect without promising equality of the three figures afterward. It records remaining differences and revisits young contacts later. None of the totals independently establishes caused revenue or a complete buyer journey, even when their definitions and technical handling are correct.

Plan changes and repeat verification

Version event definitions, reporting scope and technical assignment together. A changed CRM stage or new conversion action can interrupt comparability. Repeat a controlled case after such changes and mark the transition in reporting. Monitor how well differences can be explained, not just their size. A coincidentally small discrepancy does not prove quality. A larger, understood difference may be correct. Useful measurement supports bounded accountable decisions and keeps uncertainty visible rather than forcing all platforms into one number after the fact. Review the reconciliation when assumptions change, not merely when a dashboard begins to look untidy. Explore this further in the guide Paid media tracking acceptance: a verifiable measurement contract.