Product analytics that knows what you shipped.
Ask a question about your product in plain language. The Confidence Agent finds the right data in your warehouse, writes the query, and answers with a chart, its sources, and the rollouts and experiments that explain the change.
Add image here (HERO)
A plain-language question about activation in Germany, answered with a chart and a note about a rollout that started the same day.
The Confidence Agent is used by thousands of people at Spotify every month.
Dashboards answer the questions you planned for
Dashboards are good for questions you ask every week. They break down when you need a cut nobody built. If monthly active users drop in one market and your dashboard can't split that market by device or plan, you're stuck waiting for an analyst or writing the query yourself.
Even when you get the number, it's missing context. A standalone analytics tool can query your warehouse, but it doesn't know you rolled out a new onboarding flow last Tuesday. You still have to connect the chart to what you shipped, across separate tools.
Add image here (ILL-1)
Illustration: a dashboard with a falling line and a question mark, separated from a release timeline in another tool.
How does Product Analytics answer questions?
You ask a question in plain language, such as “How did monthly active users change in France last month?” The Confidence Agent, the AI agent built into Confidence, finds the relevant tables, writes the query, and returns the answer as a chart you can explore.
Every answer shows its query and sources, so you can check how the number was calculated. Ask a follow-up to break it down by age group, device, or plan. If the data needed to answer confidently doesn't exist, the agent tells you, so you can start collecting it. You can ask in Confidence or in Slack and get the same chart, query, and sources.
You test a hypothesis in one conversation, without building a new dashboard.
Add image here (A)
A question about monthly active users answered with a chart, with the SQL query and source tables shown below it.
Add image here (C)
The same activation question answered in a Slack thread.
How does Product Analytics explain why a metric moved?
A metric move means something different during a launch, halfway through a rollout, or right after an experiment. Product Analytics reads your warehouse alongside what Confidence already knows about your flags, rollouts, experiments, and product strategy.
When activation drops 8% in Germany, Confidence charts the drop and points out that a rollout reached 50% of German users the same day. It pulls the experiment readout for that change into the same answer. You can see how a drop lines up with a launch, or why an experiment moved only in one segment, without opening another tool.
You know what happened and have a hypothesis for how to fix it, without relying on someone to remember what shipped last week.
Add image here (B)
An activation chart for Germany with a marker for a rollout reaching 50% and an experiment readout attached.
Get your key metrics in Slack every day or week with Pulse
Some questions need an answer every day or week. A Pulse turns a useful result into a recurring check on a metric, such as monthly active users, activation, or rollout health. Each time it refreshes, Confidence posts it to your team in Slack.
A Pulse stays explorable after it lands. Reply in the thread to ask for a deeper breakdown, or to compare the metric with a rollout or experiment, and the agent answers right there.
A metric change starts a conversation in Slack and gets answered in the same thread.
Add image here (D)
A weekly Pulse on activation by market posted in Slack, with a teammate's follow-up question and the agent's reply in the thread.
You choose the data the agent can use
Connecting every table in your warehouse makes it harder for an agent to pick the right source. Spotify learned this building its internal agent: across more than 70,000 datasets, ad hoc tables and one-off analyses taught the agent the wrong patterns.
In Confidence, your data experts choose the fact tables Product Analytics can use. A fact table is a curated warehouse table of events or measurements with clear definitions. Start with product events, business metrics, and trusted aggregations, and leave out modeling tables and unfinished work. You can also give the agent instructions, such as “use the latest partition for current-state questions” or “follow our definition of an active user.”
Product Analytics works with BigQuery, Databricks, Redshift, and Snowflake, and reads your data in place, without copying it out of your warehouse.
The agent answers from data your team trusts, using your definitions.
Add image here (E)
Fact table settings in Confidence with three tables selected and a written instruction for the agent.
Does Product Analytics replace analysts?
No. Product Analytics helps you explore and decide where to look next. A chart can show a change, but it can't prove the cause, fix the source data, or settle what an ambiguous metric means. Important findings still need review from someone who knows the data, and often an experiment to confirm the cause. Confidence runs that experiment on the same platform.
Frequently asked questions
What do I need to get started?
Where does the data come from?
How do I know an answer is right?
What happens if the data doesn't exist?
Can I use Product Analytics in Slack?
How is a Pulse different from a scheduled report?
From “activation dropped” to a hypothesis in one conversation
A PM asks why activation dropped in Germany last week. Confidence charts an 8% drop starting Tuesday and notes that a rollout of the new onboarding flow reached 50% of German users that day. The experiment readout for the flow, pulled into the same answer, shows the treatment arm was flat. A follow-up shows the drop is almost entirely on desktop.
The team pauses the rollout on desktop and saves the question as a weekly Pulse in Slack, so the next change shows up in their channel.
Read more about Product Analytics
Introducing Product Analytics
Product Analytics in Confidence lets you ask questions of your data in plain language and get answers you can explore with more depth and flexibility than a dashboard.
Introducing Confidence Agent
Confidence Agent is an AI collaborator built into Confidence that works with your flags, experiments, metrics, and project documents.