What helped the customer choose you before they clicked?

A source report answers only part of the question

Imagine a buyer hears about your business in a community, watches a demonstration, asks an AI assistant about alternatives and types your address a week later. The website visit is observable. Much of the decision process may not be. Do not assign all the credit to the final visit, or assume every unseen influence came from AI.

Google Analytics distinguishes first-user, session and event-scoped channel dimensions. These describe different parts of acquisition and conversion reporting. Choose the scope that matches the question and label it in your scorecard. Current Google documentation also lists an AI Assistant channel, while Google AI Overviews and AI Mode remain under Organic Search; check the dimensions and coverage in your own property.

Keep three kinds of evidence side by side

Observed: recorded referrals, campaign parameters, enquiries, booked meetings and purchases. Reported: the customer tells you which sources influenced them. Inferred: branded searches and qualified demand rise after a campaign, but you cannot isolate its contribution. These are useful evidence types with different limits.

Use a compact scorecard: channel or initiative; spend; relevant visibility; observed enquiries; qualified enquiries; acquired customers; customer-reported influences; next decision. Deduplicate customers in your CRM or order system. Do not add platform conversion totals together and call the result unique customers.

1. Define a qualified enquiry in writing.
2. Use consistent campaign names on links you control.
3. Ask “Where did you first hear about us?” and “What helped you decide?” as separate questions.
4. Match enquiries to outcomes using an approved internal identifier.
5. Keep reported influences as multiple values, not forced last-click credit.

A simple worked example

Illustrative data: 12 enquiries become six qualified opportunities and two customers. Analytics records eight organic-search sessions and four direct sessions for those enquiries. Five respondents mention a video, three mention a recommendation and two mention an AI answer. These influence counts can overlap and are not ten additional leads.

You can say two customers were acquired from this group and describe the recorded or reported paths. You cannot conclude the video generated five incremental enquiries, or allocate a precise percentage of revenue to AI. The example is deliberately small because many service businesses make decisions with small samples.

Use experiments when the decision warrants them

For a larger investment, consider a properly designed holdout, randomized test or phased rollout with comparable groups. Define the outcome, exposure window and minimum detectable change before interpreting results. Before-and-after comparisons can be distorted by seasonality, sales response times and changes to the offer.

When volume is too low for a convincing causal test, combine customer conversations with operational data and label the conclusion as directional. A good report improves the next decision; it does not hide uncertainty behind a complicated attribution model.

Try it with your business

1. Create columns for observed, customer-reported and inferred evidence.
2. Reconcile last month’s qualified enquiries with booked calls and sales.
3. Review five customer journeys with sales.
4. Choose one tracking gap and one business decision the scorecard will support.

Use the planning worksheet and save it as a PDF →

Use AI to help organize the work

Provide only material you are allowed to use. This prompt helps structure your evidence; it does not give the tool access to your analytics or verify facts on its own.

Review this anonymized marketing scorecard. Separate direct observations, customer-reported influences and hypotheses. Flag duplicate conversions and mismatched reporting windows. Explain what we can and cannot conclude, then recommend one measurement improvement. Do not assign unsupported revenue credit.

Measure: Deduplicated qualified opportunities, acquired customers and cost per acquired customer.

Sources & further reading

Official references checked October 26, 2026. Product reporting and requirements can change.

• Google Analytics: default channel groups and scope