GA4 key events vs conversions: which belongs in your client report?
Choose the right GA4 measure for website behavior or advertising performance, with a synthetic agency example and a practical report-mapping checklist.
For website behavior questions, use the relevant GA4 key-event report. For advertising performance, use conversion reporting designed for that purpose and check the selected conversion actions. Do not rename every key event as an advertising conversion. An agency reporting workflow should choose the measure from the client's decision before collecting numbers, then retain that choice in the report template.
Start with the decision, then choose the report
A client asking whether visitors complete an enquiry form needs a different view from a media buyer assessing the conversion actions used to judge campaigns. Both questions matter. Putting a single tile called conversions above both answers makes the report harder to trust.
Google distinguishes key events, which mark important business interactions, from conversions used to measure advertising performance. Its current guidance directs campaign-performance analysis toward conversion reporting in Analytics' Advertising section, while Reports and Explore support traffic acquisition, behavior and site-journey analysis. A key-event column from a behavioral report is therefore not an interchangeable substitute for an advertising conversion column. [1]
Croissant recommends writing the decision above each recurring report section: understand enquiry behavior, assess advertising performance, or review accepted enquiries in the CRM. The last question requires its own business records; neither an analytics event nor an attributed conversion by itself establishes that sales accepted a lead.
Give each measure a small, explicit contract
For each tile, record its business question, source property or account, report name, exact metric, selected action or event, date range, filters and owner. Include a plain-language unit such as recorded form key events or selected advertising conversions. Keep the source label available even if the presentation uses a shorter friendly name.
Next record the comparison settings rather than assuming they match. Google's attribution guidance separates key-event and conversion settings. It also explains that account and property time zones can differ. These are reasons to inspect the actual configuration, not evidence that any particular difference has already been explained. [2]
This contract is a proposed agency operating convention. It is not a requirement imposed by Google. Its purpose is to let a colleague reproduce the chosen view and recognize when a replacement export answers a different question.
Worked example: one form, two reporting questions
All figures and client details in this example are synthetic. A fictional agency prepares a report for Client A. Its website-behavior section shows 40 recorded generate_lead key events in an agreed period. A separately selected advertising conversion report shows 28 conversions for the campaign question. The report owner has not yet established why the values differ.
The unsafe draft says: Advertising generated 40 leads, and 12 were lost before reaching the campaign dashboard. That sentence invents both attribution and a loss mechanism. Subtracting the displayed values produces 12, but arithmetic does not establish a set of twelve identifiable missing enquiries.
A useful draft says: The selected website report recorded 40 generate_lead key events. The selected advertising report recorded 28 conversions under its configured scope. These measures serve different reporting questions; the owner must verify comparability before explaining their difference. The account lead can still discuss the checked website observations while holding the unsupported campaign conclusion.
Now suppose the client asks how many enquiries sales accepted. The workflow must request the authorized CRM measure and its acceptance rule. It must not relabel 28 as accepted leads merely because that number is smaller. Keep the three questions separate until evidence supports a relationship between them.
Implement the choice before asking AI to write
Begin with one client report and two named sections. Have the account lead approve each section's decision and have the reporting owner select the source view. Save an example export reference and the mapping contract beside the template. In a small agency one person may fill both roles, but both choices should remain visible.
Collect only the approved fields. Pass the resulting values, exact labels and unresolved questions to the drafting step. Use deterministic checks to reject a missing metric, an unexpected action or a changed report mapping. AI can summarize the accepted inputs; it should not choose a more flattering denominator or silently swap key events for conversions.
Test the mapping with fictional inputs: the correct metric, a similarly named metric from another report, an absent conversion action and an unavailable export. The expected result for an unrecognized mapping is a review task, not a polished replacement number. For general collection and discrepancy handling, use the existing agency reporting workflow guide. This article's narrower job is selecting the report that answers the question.
Keep configuration and business decisions with people
A reporting mismatch is not permission to change tracking or attribution. Google's documentation says some attribution changes affect historical and future reporting, while other settings apply going forward; conversion configuration can also affect campaign reporting and optimization. Have the responsible specialist review the specific setting before any change. [2]
Hold campaign recommendations when their supporting measure is unresolved. Record the question, evidence required, investigation owner and next review. Client distribution and budget changes need separate authorization. A good explanation can acknowledge uncertainty without making the whole report unusable.
A maintained spreadsheet mapping may be sufficient for a small, stable report. Consider a connected workflow when collecting multiple approved views and routing exceptions repeatedly consumes effort. Use the reporting demo to discuss that scope with Croissant. Bring the report questions and fictional sample rows through the contact form, leaving private client records and credentials out of the initial enquiry.
Implementation checklist
- Write the decision each report section must support.
- Choose key-event or conversion reporting deliberately; keep the exact metric label.
- Save source account, report, selected event or action, period, filters and owner.
- Check attribution settings and time zones before interpreting a difference.
- Test wrong-metric and unavailable-source cases using synthetic data.
- Hold unsupported explanations and route them to the reporting owner.
- Require separate approval for configuration changes, client distribution and budget decisions.
The sources behind the sauce.
Primary research and official documentation behind the cited points. The workflow examples and checklists are Croissant’s own guidance.