Skip to main content
Insights helps you explore product and business data in plain language. Ask a question, inspect the answer, and continue the investigation in the same conversation.

​
How Insights works

The Confidence agent identifies relevant data, queries the data warehouse, and returns an answer that you can inspect and question further. Start with a broad question. You can then chart how monthly active users changed, break the result down by market or device, and compare the affected segment with a rollout or experiment. Insights combines warehouse data with context already held in Confidence, including projects, rollouts, and experiment results. This context helps you interpret a movement alongside what the team shipped without copying data into a separate analytics store. You can ask questions in Confidence or from Slack.

​
Key capabilities

Explore your data

Ask questions in plain language, inspect the result, and request charts, comparisons, or further breakdowns.

Connect data to delivery

Interpret changes alongside Confidence projects, rollouts, and experiment results.

Keep track with Pulses

Save results, refresh them on a schedule, and share updated summaries in Slack.

Control available data

Choose which fact tables the agent can use and add instructions for how the agent should interpret the data.

​
Explore your data

Start with a question such as:
  • How did activation change last month?
  • Which markets contributed most to the change?
  • How does the affected segment compare with the current rollout?
  • What did recent experiments show for the same metric?
The agent keeps the conversation context, so each follow-up can build on the previous result. When the answer includes a query, chart, or table, review the sources and query before using it to make a consequential decision. Insights conversation with a chart showing daily transactions by product category.

​
Connect data to delivery

A metric movement means something different during a launch, halfway through a rollout, or after an experiment. Insights can use Confidence context alongside warehouse data to help you investigate those connections in one conversation. For example, if activation drops in one market, you can chart the change, check whether a rollout reached that market at the same time, and review the related experiment result. This process can help narrow the investigation, but it does not establish that the rollout caused the change.

​
Keep track with Pulses

A Pulse turns a result into a recurring view. Save a number, chart, or table from a conversation, then configure it to refresh hourly, daily, or weekly. You can review refreshed results in Confidence and send a fresh summary to a Slack channel or direct message after a scheduled refresh. Use the update to continue the investigation when something changes. Open Subscribe and schedule on a Pulse. Enable Slack, add the destination channel ID, or select Send me a direct message, then select Save notifications. Configure the refresh schedule separately. Without a schedule, Slack updates send when you manually refresh the Pulse. Insights Pulses with recurring charts and summaries in Confidence.

​
Control available data

Insights uses selected fact tables, rather than every table in your warehouse. In Admin > Data agent, connect fact tables and choose which ones are available to the agent. Insights fact table controls in the Data agent admin settings. Start with product events, business measurements, and aggregations that have clear definitions. Leave out modeling tables, unfinished analysis, and other sources that could be misleading outside their original context.

​
Share your analytics best practices

Admins can also add instructions that explain how the agent should use the data. For example, you can define which partition represents the current state, provide your organization’s definition of an active user, or mark domains that the agent should not use for forecasting. Insights guardrails defining questions the data agent should not answer.
Insights can help locate and explain patterns, but it cannot establish causation, correct source data, or resolve an ambiguous business definition. Verify consequential findings with the relevant data owner. Questions about causal impact may require an experiment.

​
Get started

Select the fact tables eligible for Insights. Then ask a question about your data in the Confidence agent.

Fact tables

Configure the warehouse data sources available in Confidence.

Experiments

Use controlled experiments to test whether a change caused an observed effect.

The product loop

Learn how evidence feeds the next product decision.