Start with a business question
Choose what you need to understand: whether visitors open the widget, whether enquiries contain useful context, or which questions need better knowledge. A dashboard is easier to act on when the team knows the decision behind each metric.
Conversation activity, captured leads, and completed sales are separate events. Compare them without treating a chat as a sale. Your team should review the lead requirement and record the eventual business outcome through its normal process.
Optional analytics depends on visitor choice
The widget offers a choice for optional analytics. Visitors who decline can still use chat. Global Privacy Control and Do Not Track signals disable that analytics. Blocking extensions, failed requests, and a changed browser identifier can also affect the report.
A smaller report than your hosting traffic count is therefore not automatically a data collection bug. Check consent, date filters, domain installation, and whether analytics is enabled before assuming records are missing. Never bypass a visitor choice just to make the dashboard busier.
Interpret geography honestly
Approximate country, state, and city come from trusted hosting information or an available licensed local GeoIP source. If those inputs are absent, the report shows Unknown. The application does not assign a city using a visitor name or device brand.
Locations are approximate and should not be treated as exact delivery addresses. City totals need parent geography to avoid confusing places with the same name. Use captured enquiry details, where authorised, for service-area decisions rather than inferring them from a chart.
Compare the right period and denominator
Filter a period appropriate to the question and compare equivalent reporting scopes. Optional visitor events and lead records need matching attribution to support a conversion metric. Do not combine all CRM leads with an opted-in visitor denominator and present the result as a universal conversion rate.
A short campaign or a small number of enquiries can produce volatile percentages. Review counts alongside rates and explain collection limitations when sharing a report. The dashboard should support investigation, not replace judgement about traffic quality or sales.
When analytics appears empty
- Check that the real website hostname is approved and the widget loads.
- Confirm optional analytics is enabled for the environment.
- Test with your own consent choice and review browser privacy signals.
- Clear restrictive report filters and select the relevant date range.
- Remember that Unknown geography is expected without reliable location inputs.
Retention and deletion have separate scopes
Optional analytics records are subject to configured retention and scheduled cleanup. Visitors can delete records associated with their current browser analytics identifier. That deletion does not automatically remove authorised conversation or lead records because those have different purposes.
Read the analytics privacy notice and the general policy before collecting real customer data. Do not put sensitive information in page paths or campaign labels. Reports and exports should be shared only with people authorised to review the business information.
Investigate a report using a controlled visit
Open your approved website in a test browser and make an explicit analytics choice. Ask a question and submit your own synthetic enquiry, then review the matching period in the dashboard. Keep the test separate from genuine customer results and do not present it as production performance evidence.
Repeat the visit with analytics declined and check that chat remains usable. The analytics report should not be made complete by overriding that choice. A browser with a supported privacy signal should be treated consistently even if it reduces the report count.
When a metric differs from another system, compare definitions first. Sessions, unique browser identifiers, page views, and CRM contacts are not interchangeable. Check filters and attribution before concluding that events were lost. Explain these limits when a colleague exports or shares the report, especially when they plan to use a percentage in marketing copy.
A controlled trial to adapt
Example to adapt: review two synthetic visits to the same approved page, one with optional analytics allowed and one declined. Both should permit chat, but their analytics treatment differs. Check the date filter and consent choice before interpreting the report. If geography is Unknown, do not replace it with a guessed location just to complete a chart. Record the collection scope when discussing a conversion percentage. A controlled test proves the tested workflow; it is not a real customer result or an independent accuracy audit.