Analytics

Define the measurement window#

Open a site's Analytics page and choose 7, 30, 90 days, or a custom start/end date. Select the primary metric, optional channel/source filter, and Compare with previous period when you need a like-for-like baseline.

Filters apply to the report and URL so a view can be reproduced. Use equivalent campaign length, weekdays, seasonality, and publication state when interpreting a comparison. A large percentage from one or two events is not reliable evidence.

Understand the main metrics#

Message volume and average messages per conversation are engagement measures, not additional conversation starts. Returning-visitor, unanswered, handoff, booking, and qualified-lead measures describe different parts of the journey.

Read trends, funnel, and time#

Use the daily trend to find a change point, then compare it with releases, campaigns, source updates, provider incidents, and publishing changes. The funnel separates three questions: did visitors start a conversation, did conversations create leads, and did visits create leads overall?

Hourly and weekday views show when recorded activity occurred. They are useful for staffing and campaign timing, but timestamps, audience geography, and small samples can distort a simple “best time” conclusion.

Inspect acquisition and channels#

Top sources use referrer/UTM information when available. Direct/none includes traffic whose source is absent or hidden. Browser privacy, link apps, redirects, consent controls, and cross-device journeys make attribution incomplete; do not force unattributed visits into a campaign.

Channel performance compares recorded text, voice, and real-time activity. Messaging connector conversations contribute to the broader operating picture, but not every external platform exposes the same page-view or attribution baseline, so cross-channel conversion rates are not perfectly equivalent.

Find knowledge and conversation gaps#

The questions table shows frequently asked topics and unresolved/gap counts. What your agent grounded on shows recent answers plus recorded knowledge sources. Use both with the actual Conversations inbox:

1. Find a high-volume or high-gap question.
2. Read several complete threads.
3. Decide whether the problem is missing knowledge, conflict, persona, language, UI, or visitor intent.
4. Make one controlled change.
5. Rerun acceptance tests and compare a new period.

A metric indicates where to investigate; it does not explain the cause by itself.

Manage leads and follow-up#

The lead area shows captured leads, supports safe export/delete actions, and can create reviewed follow-up. Filtered lead attribution can use newer lead events while the lead inbox still shows all captured lead records; read the note shown by the dashboard before assuming counts should match exactly.

Export the report#

Use Export to create Markdown (.md), Excel (.xlsx), or a print/PDF view for the current filters. The PDF option opens a print-optimized page; choose Save as PDF in the browser. Record the filter/date context with the exported file and protect any visitor/lead data it contains.

Exports are snapshots. They do not continue updating and should not be treated as a separate source of truth after dashboard filters or underlying events change.

Privacy and data quality#

Collect only the analytics needed for a defined purpose and reflect relevant tracking in visitor disclosures. Consent blocking, privacy controls, offline delivery, duplicate suppression, provider gaps, and older events can all produce incomplete data. Use the dashboard's last-event timestamp to distinguish “zero” from “not recently receiving events.”

Troubleshooting#