Skip to main content
Help & Documentation

Discover

AI-generated questions, insights, and chart suggestions for the dataset you just loaded.

10 min readUpdated November 5, 2026
Note: Summary: Discover is the first page you see after connecting data: the questions worth asking, the patterns Golden already found, ready-made charts, and a quality check, all before you build anything.

Discover reads your dataset and hands you a starting point. It's for anyone who just connected data and wants to know what's in it, and what's worth a closer look, in the first minute.

The Discover page for an E-Commerce Sales dataset: a summary line of 10,194 rows, 11 dimensions, 4 geographic, 4 measures and 2 date fields, tags for Time-series, Geographic, Comparison and E-commerce, an Ask box, two Key questions cards, the first Quick insights, and the Data details panel on the right showing Quality as Reliable and a Profile of 21 fields

When to use it

  • A new export lands on your desk. A sales ops analyst uploads 10,000 order rows. Discover tells them Discount is the top driver of Profit / Sales before they've opened a chart.
  • Someone asks "anything interesting in here?" A team lead skims Quick insights, finds that Texas Binders lose 162.6% on Profit / Sales, and clicks Create dashboard on that card.
  • You need a first dashboard fast. A marketing manager ticks five Suggested visualizations and clicks Create Dashboard.
  • You don't trust the file yet. The Quality check says whether insights from this data are reliable, and the Profile shows every field's unique values and range.

If you're staring at a fresh dataset and wondering where to start, start here.

Getting started

  1. Connect or upload data (or pick a sample from Home).
  2. Open the analysis and click Discover in the left rail.
  3. Read the summary, then scroll: Key questions, Quick insights, Suggested visualizations.

The left rail also links straight to Key Questions, Insights and Suggested Visualizations, each as a full page.

The header

The top of the page names the dataset and counts its rows, dimensions, geographic fields, measures and date fields. Tags like Time-series, Geographic and E-commerce say what kind of data Golden thinks it is, with a one-line description underneath.

  • Chat with your data opens the chat panel.
  • Refresh reruns the analysis, for example after you change the data in Prepare.
  • The Ask box takes a question, or a description of the dashboard you want.

Key questions

Key questions are the questions your data can answer well. Each card tells you what kind it is:

CardWhat it does
Recommended dashboardA ready analysis Golden can build for this data, like "What sells together?" for order data.
Custom time analysisFinds the periods where a measure swung most and breaks each swing down. Pick the measure and the time grain.
Custom geography analysisShows where a measure is concentrated and how that shifts. Pick the measure, geography level and time grain.
_N_ insights combinedA question built from several insights, with the supporting numbers listed under it.

Click Explore or Explore the answer to open the answer. Use the arrows to page through, or See all key questions for the full grid.

The Key questions page with 7 cards: a Recommended dashboard card for What sells together?, a Custom time analysis card with Profit as the measure and quarter as the time grain, a Custom geography analysis card with Profit, Postcode and year, and three insights-combined cards such as Why does Furniture struggle to generate meaningful profit?

Quick insights

Quick insights are patterns Golden found on its own: the top driver of a measure, a seasonal swing, a combination that's worse together than apart, a level that runs at a loss. Each card says what it found in plain numbers and shows a small chart or table to back it up.

Click Create dashboard on a card to turn that insight into a dashboard. See all insights opens the full list.

Four Quick insights cards: Discount is the top driver of Profit / Sales with influence bars, Volume peaks in Nov and Dec with a monthly bar chart, State/Province Texas + Sub Category Binders compounds into the worst Profit / Sales with a table, and Discount 80% runs Profit/Sales at a loss with red bars

Suggested visualizations

Golden suggests a handful of charts for your data, each with the question it answers and the fields it puts on each axis.

Suggested visualizations: Monthly Sales Trend, Top Sub-Categories by Sales, Profit by Region, Profit vs Sales by Category and Top States by Sales, each with a preview, its question and its fields

Tick the checkbox on the charts you want, or click Select all. A bar appears at the bottom:

  • Add to Analyze adds each one as a sheet you can keep editing.
  • Create Dashboard puts them all on a new dashboard.
All five suggestions selected, with the bar at the bottom reading 5 visualizations selected, Add to Analyze and Create Dashboard

Data details

The Data details panel on the right is your sanity check.

  • Quality gives a verdict. Reliable means no issues were found and insights from this data are reliable.
  • Profile counts rows, dimensions, measures, dates and geographic fields, then lists every field with its unique count.

Click a field to expand it. You'll see its Field Type, number format, a short description, Total, Unique, Range and sample values. Click the chart icon on a field to analyze it in a new sheet. Collapse the panel with the icon at its top right.

The Data details panel: Quality marked Reliable, the Profile tiles, and the Discount field expanded to show Field Type Measure, Number Format Standard with 2 decimals, a description, Total 10,194, Unique 12, Range 0.00 to 0.80 and sample values

Do it from chat

The Ask box at the top of Discover goes to the same analyst as chat. Ask a question, or describe the dashboard you want:

Which regions lose money on discounted orders?
Build a dashboard of monthly sales and profit by category

FAQ

Does Discover change my data? No. It only reads it. Changes happen in Prepare.

Why did my key questions change? Golden keeps looking while you're on the page ("Looking for more…"), and Refresh reruns the analysis. Questions and suggestions can differ between runs.

Can I edit a suggested chart before using it? Add it with Add to Analyze, then change anything on the sheet.

What if Quality isn't Reliable? The panel lists what it found. Fix those fields in Prepare, then click Refresh.