Explore Analytics and turn findings into improvements
Read the current dashboard and specialized Analytics views, then convert evidence into targeted Agent changes.
Analytics helps administrators understand how shoppers use the Agent and where content, catalog, support, or deployment work is needed. The overview can combine conversation, revenue, cart, conversion, interaction, support, visitor, contact, ticket, product-page, moderation, and satisfaction metrics depending on the saved dashboard configuration and available data.
This guide explains how to choose a period, interpret the available views, drill into evidence, and close the loop with a controlled improvement. It also separates the full Analytics permission from the narrower Agent report available to supervisors.
How this fits into Humind
Catalog data is the factual base for product discovery and recommendation. Integration state, product visibility, metafield filters, collections, promotions, markets, and storefront deployment all influence what a shopper can see and what the Agent can explain.
Verify source data before compensating with Agent guidance. If a price, variant, collection, or attribute is wrong in the catalog, correct or resync that source first, then retest the exact product journey in Humind.
Before you start
Access: Full Analytics pages require analytics read access. The Agent report has its own permission and can be visible to supervisors without the rest of Analytics.
- Choose the company and reporting period that match the business question.
- Confirm that the Agent has been live long enough to produce a meaningful sample.
- Define the decision you want to make before changing dashboard cards or filters.
Step-by-step workflow
Start with the overview
Open Analytics and select the reporting period. Review totals and changes together: a percentage without its sample or a revenue result without the associated conversations can be misleading. Some cards open a drill-down such as all conversations, revenue conversations, support conversations, cart events, conversion, interaction, offensive conversations, or satisfaction ratings.
Administrators with write access can edit the dashboard, add or remove KPI cards and charts, and reorder them. Keep a stable core dashboard for regular reviews so period-over-period changes remain easy to compare.
Use specialized views
Top questions shows recurring shopper questions and detail. Top URLs explains where interaction occurs. Topic map groups conversation themes. Conversation insights summarizes analyzed conversations. Reporting manages scheduled analytics reports. Agent report focuses on operator supervision and handoff performance under its separate permission.
Use the view that matches the question instead of interpreting every issue from a single headline metric. For example, combine a support-volume change with question themes and sample conversations before changing escalation.
Inspect evidence, not only aggregates
Open relevant conversation drill-downs and read a representative sample. Classify each issue as missing or stale Knowledge, catalog quality, guidance, tool behavior, escalation, operator workflow, or installation. This prevents a broad prompt change from being used to fix a narrow data problem.
For satisfaction metrics, consider the number of ratings. Humind displays no percentage when a selected period has no responses, which is different from a measured zero score.
Make and validate one improvement
Assign an owner and make the smallest source-level change that addresses the observed problem. Test the affected scenario in Playground or Batch testing, then monitor the same Analytics view over an appropriate future period.
Record what changed and when. Without a change date and a stable comparison, later movement cannot be confidently attributed to the improvement.
Permissions and important caveats
- Metric availability and meaning depend on captured events, connected commerce data, selected period, and sample size.
- Operators do not receive full Analytics by default; supervisors receive only Agent report by default.
- Correlation in a dashboard does not prove that an Agent change caused the business outcome.
- A trial or subscription state can limit access to dashboard data or actions.
Verify the result
Use this checklist before considering the work complete:
- The selected company, currency context, and period are correct.
- The conclusion is supported by both an aggregate view and sample conversations.
- The improvement was made in the correct source area and passed a targeted test.
- A follow-up date and the metric or view to monitor are recorded.
Troubleshooting
Analytics is missing from navigation
Ask a company administrator to review analytics permissions. A supervisor may legitimately see only Agent report, while an operator has no Analytics access by default.
A metric is blank or shows no rating
Check the reporting period and sample. For satisfaction, no responses are displayed as unavailable rather than as zero. Also confirm the relevant survey or event collection was enabled during the period.
The dashboard and a conversation seem inconsistent
Confirm filters, period, company, currency, and event timing. Use the card drill-down or specialized view to inspect the exact records included before escalating the discrepancy.