Analytics is most useful when it changes a specific source or operating practice. The dashboard and specialized views can reveal conversation volume, revenue and cart behavior, support demand, recurring questions, page context, topics, satisfaction, tickets, contacts, and operator performance depending on access and available data.

This guide provides a recurring review routine that moves from a business question to evidence, diagnosis, correction, test, and follow-up. It avoids interpreting a single metric without period, sample, or conversation context.

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: Analytics read access is required for the full area and Analytics write for dashboard customization. Agent report is a separate permission available to supervisors by default.

• Define the review period, company, market, and business question.
• Record important configuration or deployment dates that may affect results.
• Choose an owner who can investigate conversations and coordinate source changes.

Step-by-step workflow

1. Frame one review question

   Start with a decision such as reducing repeated support questions, improving product discovery, understanding a conversion change, or reviewing operator handoffs. Select the same reporting period across the views you compare.

   Avoid opening every dashboard card without a question. A focused review produces a clear owner and smaller corrective action.

2. Read the overview with context

   Review the configured KPI cards and charts, totals, change percentages, currency, and samples. Administrators can add, remove, and reorder current metrics or charts in edit mode; keep a stable set for recurring business reviews.

   Use clickable cards to open conversation, cart, conversion, interaction, offensive, or satisfaction detail where available. A satisfaction dash with zero sample means no ratings, not a measured zero score.

3. Use specialized evidence

   Open Top questions for recurring intents, Top URLs for page context, Topic map for themes, Conversation insights for analyzed conversations, and Reporting for scheduled reports. Use Agent report for live operator supervision when the role has that separate permission.

   Read enough sample conversations to distinguish a real pattern from an outlier. Note the shopper question, Agent response, product or source, action, handoff, and outcome.

4. Diagnose the source and change it

   Classify the pattern as Knowledge, Catalog, Guidance, Tools, Escalation, Inbox process, interface, installation, or measurement. Change the smallest authoritative source and avoid broad rules that hide a narrow data issue.

   Assign a reviewer and write the expected impact. If evidence is uncertain, run a controlled test or collect more data instead of presenting the hypothesis as a fact.

5. Test and measure the follow-up

   Add the observed shopper question to Playground or a Batch testing dataset and verify the correction. Check nearby scenarios for regression. Then monitor the same metric and specialized view over a future period with an appropriate sample.

   Use Reporting or the send-report action when stakeholders need a recurring or point-in-time readout. Include the period, source, change date, and limitations.

Permissions and important caveats

• Metric interpretation depends on captured events, integration quality, period, currency, filters, and sample size.
• Analytics can support a hypothesis but does not prove causality on its own.
• Supervisors see Agent report without full Analytics by default; operators see neither Analytics section by default.
• Subscription state can limit data access or actions.

Verify the result

Use this checklist before considering the work complete:

• The company, market, period, currency, filters, and sample are recorded.
• At least one aggregate view and representative conversations support the diagnosis.
• The correction changed the authoritative source and passed regression tests.
• The follow-up metric, date, and owner are defined before closing the review.

Troubleshooting

Analytics access is denied

Ask an administrator to review analytics or Agent report permissions. A supervisor is intentionally limited to Agent report unless a custom role grants more.

A dashboard card has no useful data

Check period, event availability, sample, integration, and whether the KPI is relevant to the company. Replace it only with administrator write access and preserve a stable review set.

A metric moved after several changes

Use deployment and configuration dates plus conversation evidence to separate hypotheses. Do not attribute the movement to one change without a controlled comparison.

Related guides

• Explore Analytics and turn findings into improvements
• Test your Agent in Playground
• Work from the Inbox
• Manage products, filters, collections, and promotions