Analytics governance, or why your reports disagree.

When a GA4 number and a CRM number diverge, the instinct is to look for broken tracking. That is rarely where it is. Tracking breaks loudly and produces zeros. Governance failures produce plausible numbers that are quietly measuring different things.

A form submission fires a GA4 event. The CRM records a contact. Those are different moments. One counts the click, the other counts the record that survived deduplication, spam filtering, and a sales rep deciding it was real. A gap between them is expected. A gap nobody can explain is the problem.

The question worth asking is not which number is right. It is: can you state, in one sentence, what each number counts. If you cannot, the reporting is decorative.

Definition drift. A conversion event was set up to mean one thing and now fires on three. Nobody changed the report, so the trend line looks like growth.

Double counting. The same lead enters twice, once through a form and once through a call, and only one system deduplicates. Both systems are correct. The totals will never reconcile.

Attribution window mismatch. GA4 attributes inside a lookback window. A CRM attributes to whatever was typed in a source field, often weeks later, often by a person. Comparing them is comparing two different questions.

Silent filtering. Bot traffic, internal IPs, spam submissions, and thresholded reports each remove rows. Each removal is defensible. Together they explain most unexplained gaps.

Diagnosis order matters more than diagnosis depth, because the first two checks resolve most cases in under an hour.

Start with definitions. Write down what each system counts as a lead, in plain words, before opening either interface. Most engagements end here, with two definitions that were never the same.

Then check deduplication. Take one real week and count the actual human beings who made contact. Compare that to both systems. The system further from the human count is the one with the reconciliation problem.

Then check the window. Align the date ranges and the attribution model before comparing anything. A GA4 report and a CRM export rarely default to the same period.

Only then look at the tag. Implementation is the last place to look, not the first, and it is where most audits waste their first day.

One number is designated the number that decides spend. Everything else is diagnostic. That single choice removes most reporting arguments, because the arguments were never about accuracy, they were about which report had authority.

Definitions live in writing, outside the tools, and a change to one requires the same approval as a change to budget. Events are named so a person who has never seen the account can guess what they mean.

The monthly report leads with booked work, not sessions. If the top line of your reporting is traffic, the reporting is measuring the agency's activity rather than the business.

The common failure is buying more measurement to fix a governance problem. Server side tagging, a new dashboard, an additional attribution tool. Each adds another system with its own definition of a lead, and the disagreement gets one participant louder.

The second failure is fixing it once. Definitions drift because businesses change: a new service line, a new form, a new intake process. Governance is a review cadence, not a project with an end date.