Skip to main content
By the end of this guide, you have answered a product question from your data, broken the answer down in a chart, and saved it to a Pulse that refreshes and posts to Slack. The guide has these steps:
  1. Ask a question in plain language.
  2. Break down and visualize the answer.
  3. Save or share the result as a Pulse.
  4. Act on what you learn: test a finding with an experiment, or watch the sessions behind it.
Do it with AI. The Confidence Agent writes SQL against the data warehouse and turns the results into charts. Open the agent in Confidence and start with one prompt:

​
Before You Begin

​
Ask a Question

Start with a broad question about a product outcome. The agent identifies the relevant data, queries the data warehouse, and returns an answer. To learn more, see Insights.
Ask the Confidence Agent a question such as:
You can also ask the agent from Slack. See Use Confidence in Slack.
Review the sources and the query in the answer before you use it to make a consequential decision.

​
Break Down and Visualize

Ask follow-up questions to narrow the answer. The agent returns charts and tables.
Ask for a breakdown or comparison:
The agent combines warehouse data with Confidence context, such as projects, rollouts, and experiment results. This helps you read a metric movement alongside what the team shipped. It doesn’t establish that a change caused the movement.

​
Save or Share

Select Add to pulse to save a number, chart, or table from the conversation to a Pulse. A Pulse refreshes on a schedule, hourly, daily, or weekly, and you can send each refresh to Slack. To send updates to Slack:
  1. On the Pulse, open Subscribe and schedule.
  2. Enable Slack, then add the destination channel ID or choose to receive a direct message.
  3. Select Save notifications.
  4. Configure the refresh schedule separately. Without a schedule, Confidence sends the Slack update only when you refresh the Pulse manually.
To post to a channel, first add Confidence to the Slack channel.

​
Act on What You Learn

Exploration finds patterns. It can’t prove what caused them.
  • Test a finding. If you suspect a change drives a metric, test it. See Run an A/B Test.
  • Watch sessions. Record and replay sessions to see the user behavior behind a number. See Set Up Session Recordings.