Tables

A table is the right visual when you need detail — the actual rows, the specific values, the items behind a number. Which accounts declined, and by how much. Every order over a threshold. One row per customer with the columns that matter. Where a card gives you a headline and a chart gives you a shape, a table lets you read and look things up.

Like every visual, you get one by describing it — in a conversation or a dashboard prompt.

Describe the rows and the columns

A table is defined by two things: what each row represents (the grain), and which columns you want to see. Name both, and the first result is usually the one you wanted. Here’s a real prompt and the table it produces — it’s live, so try sorting and filtering it:

You ask “One row per customer: email, number of orders, total revenue, and last order date. Tag each customer as loyal or one-time, sort by revenue, and highlight the segment.”

Notice how each part of the prompt became part of the table:

• “One row per customer” set the grain — not one row per order, not one per segment.
• The named columns appear in the order they were asked for.
• “Sort by revenue” put the biggest customers on top.
• “Highlight the segment” produced the color-coded segment column.
• Revenue came back formatted as currency and dates as dates — sensible formatting is the default, and you can override any of it (“show revenue in thousands”, “dates as ‘Jan 21’”).

(Customer emails in this sample are anonymized — that’s why they look like codes.)

What you can ask for

• Which columns — name them, in the order you want them. “Drop the last order date.” / “Add the customer’s first order date.”
• Sorting — “sort by revenue, highest first”, “order by decline, worst at the top”.
• How many rows — “top 20”, “only accounts below target”, “everything”.
• Filters — “only the loyal clients”, “orders over $1,000”.
• Formatting — currency, percentages, dates; “show revenue in thousands”.
• Highlighting — “highlight loyal clients in green”, “flag anything below target in red”. Color earns its keep in tables: it lets the eye find the rows that matter without reading every line.
• Totals — “add a total row at the bottom”.

Orders over $1,000 this month: order ID, customer, amount, status — sorted by amount, largest first, overdue ones highlighted.

Reading the result

Every table you get back is interactive — no need to re-ask for small explorations:

• Click a column header to sort by it; click again to reverse.
• Open a column’s menu to filter — by text, number range, or date.
• Drag column edges to resize.

These tweaks are for reading. To change what the table shows — different columns, different rows, a different grain — ask the agent.

Refining a table

As with every visual, you change it by asking, in small steps:

• “Sort by the change column instead.”
• “Only show the top 10.”
• “Add a column for last year’s value.”
• “Format the amounts as currency.”
• “Highlight customers who haven’t ordered since November.”

When a table isn’t the right visual

• If the point is a single headline number, use a card.
• If the point is a trend or a comparison you’d rather see than read, use a chart. A table of 24 monthly values is usually a line chart waiting to happen.

A table answers “show me the specifics” — the rows, in the order and detail you asked for.

Export

Download any table as CSV to take the underlying data into a spreadsheet or another tool.

Limits to know

• No row grouping. Tables are flat — there’s no expand/collapse or grouped subtotals. To summarize by a category, ask the agent to aggregate it in the analysis first (one row per group), rather than grouping rows in the table.
• Read-only. Tables are for reading and lookup; you can’t edit cell values. To change what a table shows, change the analysis behind it.