Skip to main content
Collection visualizations
Visualizations turn your collections into charts (distributions, trends, comparisons) without exporting anything to a spreadsheet, computed by the same exact aggregation engine your agents use over the structured-data layer. They live on a collection’s Visualize tab.

​
What the Visualize tab looks like

Open a collection and switch to Visualize. You’ll see:
  • a row of KPI cards (single headline numbers) at the top,
  • below them, a grid of charts (bar, pie, line),
  • and an Add chart button that opens the chart builder.
On a brand-new collection, a few charts are suggested automatically from your schema (a total-records KPI, a KPI summing the first numeric column, a bar chart counting by the first text column, and a line over the first date column) so the tab is never empty. Everything else, you build yourself.

​
The four building blocks

Every chart is defined by four choices. Understanding them is enough to build any view: A chart is therefore read as <aggregation> of <measure>, grouped by <dimension>, shown as <chart type>, e.g. sum of amount by stage, as a bar chart.

​
Build a chart, step by step

The chart builder: chart type, group by, measure
Let’s build “Won amount by stage” on a Sales CRM collection whose Deals have stage (text), amount (number) and is_won (boolean) columns.
1

Open the builder

On the Visualize tab, click Add chart. The builder appears above the KPIs.
2

Pick the chart type

Choose Bar, we’re comparing a value across categories.
3

Choose the dimension

In Group by, select stage. Each bar will be one stage.
4

Choose the measure and aggregation

In Measure, select amount (instead of Count of rows). In Aggregation, select Sum. Each bar’s height is now the total amount for that stage.
5

Add it

Click Add chart. The chart is created, sorted from highest to lowest, and saved with the collection, it will still be there after a refresh, for everyone who can see the collection.
To remove a chart, hover it and click the × in its corner.

​
Recipes

Concrete settings for common questions. Read each row as the four choices in the builder:
Not sure which columns exist or how they’re named? Use the Ask bar on the Data tab first (“what are my stages and how many deals in each?”), then translate the answer into a chart.

​
How the numbers are computed

The theory behind the builder, so results never surprise you:
  • Exact, server-side aggregation. Charts don’t sample the visible page, they aggregate the whole collection in the database (SQL GROUP BY / equivalent), the same engine agents use. A “sum of amount by stage” chart and an agent answering “total by stage” always match.
  • One dimension at a time. A chart groups by a single column. For a two-dimension cross-tab, build two charts or filter the data.
  • Sum/Average/Min/Max need numbers. These aggregations only apply to numeric columns; on text/date columns, use Count of rows. Empty cells are skipped by sums and averages.
  • Dates are bucketed by month. A Line over a date/datetime column groups points by month and orders them chronologically, so a daily timestamp column still gives a clean monthly trend.
  • Top values first. Bar and pie charts sort by value (largest first) and show the top slices, so the most important categories are always visible.
  • Empty groups are labelled. Rows with no value for the grouping column are gathered under an “(empty)” label rather than dropped.

​
Best Practices

Charts group by exact values. Normalize category columns (consistent casing and wording) before importing, or you will see duplicate bars for the same logical category.
Sums and averages skip empty cells. If a numeric column has many gaps, count-based views give a more honest picture of your data.
Since charts are derived from your schema, a collection mixing unrelated record types produces confusing visualizations. One topic per collection keeps dashboards readable.

​
Next Steps

Working with Data

Clean your data to improve your charts

Using Collections with Agents

Let agents run the same aggregations conversationally