What Google Ads conversion tracking tells you—and what it misses

Conversion tracking can answer a carefully defined question very well: did a configured action get recorded and attributed under the system's rules? Problems begin when we ask that number to answer a different question, such as “Was this a good customer?” or “Would the sale have happened without advertising?” A tag is a measuring instrument. It does not chair the revenue meeting.

Good measurement starts by naming the event, the business meaning and the decisions it should support. A completed form, a qualified enquiry and a paid contract belong to the same journey, but they are not interchangeable outcomes.

Begin with the event you actually need

Imagine a hypothetical payroll software campaign. The website records a form submission as a conversion. An enquiry from a job seeker and an enquiry from a finance manager can both trigger the same event. The tag may be working perfectly while the campaign is attracting the wrong people.

Create a measurement map before adjusting bids. For each event, write its trigger, owner, counting unit, expected frequency and business purpose. Define what counts as successful completion on the server or application, rather than assuming every button click represents a submitted enquiry. Include failed submissions, retries and duplicates in the test plan.

Then follow a test record through the entire path. Confirm that the website accepts the enquiry, the CRM creates or updates the intended record, and the responsible team can recognise it as a test. Exclude test activity from business analysis using a documented process. A reassuring tag preview is one checkpoint, not the finish line.

Primary is a bidding decision

Google's documentation distinguishes primary and secondary conversion actions. A primary action can influence bidding when its associated goal is selected by the campaign. Secondary actions are generally observed through “All conversions.” There is an important exception: a secondary action included in a custom goal can still be used for bidding when that custom goal is selected.[1]

This makes campaign-level goal inspection essential. A useful account naming convention does not guarantee the intended goal is selected everywhere. Review the actual settings, including custom goals, and ask whether the optimisation signal represents what the business wants more of.

Do not turn every observable interaction into an equally important success. A page view can help diagnose engagement without deserving the same business interpretation as a qualified sales enquiry. When both are counted together, the total becomes hard to interpret even if every tag is technically correct.

Attribution distributes credit

Attribution assigns credit to touchpoints according to a rule or model. Google Analytics documents different attribution models and explains how its data-driven approach uses available path data.[2] A model can be useful without making the report a complete history of human influence.

A buyer may hear about a company from a colleague, investigate on another device and later search for its name. Some of that journey may be visible, some may not. Consent choices, identifiers and the available integrations affect the observable evidence. Avoid treating “not present in this report” as “did not matter.”

Also distinguish attribution from an experiment designed to estimate incremental impact. Changing the credit distribution does not by itself create new customers. If a high-stakes budget decision depends on what would have happened without the campaign, consider a suitable controlled test and its practical constraints.

Bring the business outcome back

Offline feedback connects later business stages to advertising measurement. Google's enhanced conversions for leads documentation describes supplementing imported outcomes with appropriately handled first-party information.[3] That improves the connection between enquiry and later events; it does not make poor CRM stage definitions reliable.

Before implementation, agree which milestone deserves reporting, who records it and when it becomes stable. Document identifiers, timestamps, duplicate handling and consent requirements. Use current official integration guidance: upload routes and platform settings change, so a copied implementation recipe can become stale.

Inspect diagnostics and reconcile a known sample end to end. An accepted upload is not proof that every record matched, and a matched record is not proof that the salesperson classified it correctly. Keep operational ownership alongside technical ownership.

Reconcile differences instead of hiding them

Expect tools to disagree until you align their definitions. Compare date ranges, time zones, event names, attribution settings, counting rules and the unit of analysis. Then investigate remaining differences with examples. Do not force CRM revenue and platform conversion totals to match by quietly discarding inconvenient rows.

A short measurement contract helps: “This report counts unique qualified enquiries by acquisition month, with status current as of the stated review date.” That sentence often prevents more confusion than another dashboard tab.

The takeaway: use conversion tracking as evidence about a defined event. Add CRM outcomes for commercial meaning, document blind spots and test the whole path. Strong tracking gives you better questions. It does not remove the need to ask them.

Sources

[1] Google Ads: conversion goals ↗.
[2] Google Analytics: attribution ↗.
[3] Google Ads: enhanced conversions for leads ↗.
Platform documentation reviewed 26 October 2026. This is a decision framework, not an account-specific implementation guide.