Generate and view insight reports

A project's Reports tab shows the aggregated AI analysis across every in-scope conversation, turning individual meetings into trends you can act on.

Open a project's Reports tab

1. In the sidebar, open Data Lake.
2. Pick your project from the list.
3. The Reports tab is the default tab on a project page.

Reports show one card per active Insight Definition. Each card displays the metric or chart appropriate to the definition's type, and the cards refresh over WebSocket as new conversations are processed — there is no manual refresh required.

Display options

At the top of the Reports tab:

• Display dropdown — switch between Graphs (chart-first) and List (table-first) views.
• Download — exports the project's data as CSV (see Reports and exports).

The Download button is enabled only when at least one definition exists. Empty projects show the tooltip "No insights defined for this project."

What reports look like per definition type

Visualisations depend on the definition type:

Reports in Harmony are tailored to the type of Insight Definition you've configured, with each definition focusing on a specific way of analyzing your conversations. Understanding how each definition type presents its results helps you interpret the data most effectively.

If your insight definition is a Summary, Open Question, Markdown, Competitors Analysis, or SPIN Improvement, the report will present a detailed table, where each conversation that matches your project’s criteria appears in a separate row. Within this row, you’ll see the AI-generated response associated with that conversation. This format helps you quickly scan across multiple meetings and read the AI’s output for each, making it easier to spot patterns, quotations, and themes at a glance.

For insights based on Yes/No Questions, the report aggregates how often each response (“Yes” or “No”) occurred across all relevant conversations. You’ll see both the raw counts and visual charts (such as bar or pie charts) that illustrate these proportions. In addition, one of the answers may appear highlighted, depending on how the definition was configured—making it clear whether your team is tracking toward a positive or negative metric.

When using the Discount definition type, the report focuses on identifying in how many conversations a discount was discussed or detected, calculating the overall detection rate. If actual discount values were captured during the analysis, those amounts will also be summarized, helping you understand the prevalence and magnitude of discounts offered across meetings.

A Buyer Engagement report interprets overall client or prospect engagement throughout all conversations in your project. It displays an average engagement score and shows how the conversations are spread between high, moderate, and low engagement categories. This distribution helps sales and customer teams quickly identify if they are succeeding at keeping buyers involved.

For definitions using a Scale Rating, the report shows the average rating given across all analyzed meetings and also breaks down the ratings across your custom scale (for instance, rating responses from 1 to 5). This way, you can immediately tell not just the consensus score, but also see how often various scores occur.

Checklist insights are displayed by showing the proportion of times each checklist item was marked as “pass,” “fail,” or “unsure.” This allows you to evaluate processes or compliance steps, seeing which checklist items are most frequently completed, skipped, or left uncertain.

If your report is driven by a Group definition, it will display how conversations are distributed among the named categories or buckets you’ve configured. You’ll be able to see which groups or segmentation criteria are most commonly represented in your data, helping you segment your outcomes further.

Lastly, both Funnel and SPIN Distribution definition types generate stage-by-stage visualizations of your sales, support, or engagement pipeline. The data for these definitions is presented as sequential charts—either bar charts or spider/radar charts depending on the definition—making it easy to follow progression and drop-off rates at each defined stage.

It’s worth noting that Harmony customizes the charts and metrics for each definition type, so you always get the most meaningful visual or numeric summary for your data. There isn’t a global “heatmap” or aggregate chart across all definition types; instead, each card and report is purpose-built to illuminate the particular question or trend you’re analyzing.

When you want to examine one specific definition in greater detail, simply click on its insight card. This action opens the dedicated detail view for that insight. At the top, you’ll find a summary tailored to the insight’s type—such as the relevant average, breakdown, or group leader—giving you immediate context. The detail view also provides powerful filtering tools: you can apply conversation filters (like by team, date, or tag), as well as filters that are unique to the result type—such as showing only meetings where the Yes/No answer was “Yes,” or only those where a checklist item failed. Below the filters, a paginated list appears, displaying each relevant conversation and the AI’s direct answer, with a link back to the meeting for evidence. If you need to work with this data outside of Harmony, there’s an option to export these filtered results as a CSV, scoped precisely to your current selection.

URL state is preserved on the detail view, so you can bookmark a filtered breakdown.

Cross-project filters on the Reports tab

The filter bar at the top of the Reports tab lets you narrow what's shown across the project's cards:

• Date range — compare performance over a chosen window.
• Teams — when the project has multiple teams, isolate one.
• Tags and Meeting Status — focus on subsets of conversations.

Filters combine, so you can get as specific as you need.

Updating reports after definition or criteria changes

Past conversations are not retroactively re-analysed when you:

• Add a new definition to the project.
• Edit an existing definition's configuration.
• Add or change a project criterion.
• Change the project's Language or Project Knowledge.

To apply those changes to past conversations, run a Revision from the Revisions tab — see Reports and exports. Revisions are gated by the reports-rerun-tab workspace feature flag and require the pipeline-rerun:execute:all permission to trigger.

Troubleshooting

The report is empty. Reports require at least one definition under Configuration and at least one completed meeting from an assigned team. If you just added a new definition, past conversations won't be analysed automatically — start a Revision (see Reports and exports).

Data seems outdated. Check the date range in the filter bar. If you changed definitions or criteria, run a Revision so past conversations are re-analysed with the new logic.

Cards aren't refreshing. The Reports tab updates over WebSocket. If your network blocks WebSocket connections, refresh the page manually.

A specific conversation is missing from the report. Open that conversation and check whether it falls into the project's team scope and whether all attached Project Criteria evaluated as required. If a criterion answered "No" or "N/A", the conversation will not appear.