Short answer: Define the chain before judging the channel. Track how a person discovered the business, what they did on the site, which lead action occurred, whether the lead was qualified and which business outcome can be verified.
Start with the business decision
A report should help someone decide what to protect, improve, test or stop. That decision determines which measures matter. A campaign manager may need query and cost evidence each week. An owner may need qualified opportunities and business outcomes over a longer period.
Do not begin with every metric a platform can export. Begin with the action the organization is trying to create and the evidence available at each stage.
Layer one: discovery
Discovery measures whether the business appeared in a relevant context. Examples include search impressions, paid reach, social distribution, Maps visibility and known AI/search referrals. These measures help diagnose coverage and demand, but they do not show that a person engaged or became a customer.
Layer two: engagement
Engagement shows what happened after discovery. Useful signals may include landing-page views, navigation to a service page, meaningful content interaction or progress toward a contact path. Avoid treating every scroll or short interaction as a conversion.
The page should also be reviewed directly. Analytics may show that people leave, but a rendered page inspection can reveal why: unclear copy, mismatched intent, a weak mobile layout, a broken form or a slow primary element.
Layer three: lead action
A lead action is a defined step such as a completed form, connected call, audit request or scheduled conversation. Track success only after the action completes. A form view, button click or dial attempt may be useful as a diagnostic event, but it is not the same as a submitted or qualified lead.
Keep personal information out of analytics event parameters. Store the lead securely in the appropriate business system and send only non-sensitive event context to analytics.
Layer four: qualification
Qualification asks whether the inquiry fits the service, market, budget, timing or other real business criteria. Without this feedback, ad and analytics platforms may optimize toward the easiest form completion rather than the most useful opportunity.
Create a small set of qualification outcomes that the team can apply consistently. Examples might include qualified, unqualified, duplicate, spam or unable to determine. The labels should reflect the actual sales process rather than a generic template.
Layer five: business outcome
The strongest available evidence may be a sales opportunity, signed engagement, purchase or revenue record. Not every organization can connect this stage immediately. When it cannot, the report should say so and avoid presenting lead volume as revenue proof.
When CRM or sales data is connected, preserve the original source and landing context carefully. Campaign parameters, click identifiers and first landing details can help, but attribution still has limitations across devices, privacy controls and longer decision cycles.
| Layer | Question | Example evidence | Common mistake |
|---|---|---|---|
| Discovery | Did the business appear? | Impressions, reach, Maps pattern, known referrer | Calling visibility a lead |
| Engagement | Did the visit find something useful? | Page path, useful interaction, conversion progress | Counting every interaction as success |
| Lead action | Did the contact action complete? | Submitted form, connected call, audit request | Using button clicks as completed leads |
| Qualification | Did the inquiry fit? | Sales or intake disposition | Optimizing toward cheap spam |
| Business outcome | What happened downstream? | Opportunity, sale or verified revenue | Claiming revenue without a record |
Define events before implementation
For each event, record its name, trigger, owner, expected parameters, sensitive-data restrictions and test method. Use a success event for a completed action and separate diagnostic events for earlier steps. This makes QA and reporting easier to interpret.
Google provides recommended event conventions for analytics implementations. The event name is less important than using it consistently and verifying the exact trigger. See the Google Analytics event reference.
Clean the reporting view
Spam referrals, non-production hostnames, internal testing and duplicate events can distort the headline. Keep a raw view for auditability and build a documented reporting segment that removes known non-business traffic. Do not quietly delete inconvenient data.
Record the exclusion rule, why it exists and when it was last reviewed. A filter that once removed obvious spam may later hide legitimate traffic if it is too broad.
Use the report to make one decision
Every reporting cycle should end with a small set of actions. Protect what is working. Investigate a specific gap. Run one controlled test. Repair measurement where certainty is too low. Stop work that has no business role.
The report is complete when the reader understands the evidence, the uncertainty and the next decision without needing to decode the platform.