Agentic product analytics platform that runs on your data warehouse
Ask for a funnel, a retention curve, or a cohort in plain language and get the answer in Slack or in the app. Mitzu reads the events already sitting in Snowflake, BigQuery, or Databricks — no second copy of your data, no per-event bill.
Conversion Funnel
User Conversion Flow
1. Page Visit
2. Sign Up
3. Activated
4. Converted
10,000
Total Users
9.9%
Conversion
+12%
vs Last Month
Trusted by leading companies
What is a product analytics platform?
A product analytics platform reads event data — page views, sign-ups, feature clicks, purchases, cancellations — and turns it into the four analyses product teams run constantly: funnels, to see where users fall out of a flow; retention, to see whether they come back; segmentation, to see which groups behave differently; and journeys, to see the paths they actually take. Cohort definitions sit underneath all four, so the same group of users can be followed across every question.
First-generation platforms bundled a fifth job in with those four: storage. Before you could analyse an event you had to ship it to the vendor through their SDK, pay to keep it there by volume, and accept that the analysis would only ever see what the vendor ingested. Billing records, CRM fields, support tickets and dbt models stayed on the other side of that wall.
Mitzu separates the two jobs. Your events stay in your warehouse. Mitzu reads them where they are and runs the same four analyses on top, joined natively to everything else you already model. The result is a product analytics platform without the ingestion pipeline — and, because an agent drives the analysis, without the query-building step either.
Warehouse readiness is the qualifier, not a nice-to-have. Mitzu needs event data already landing in Snowflake, BigQuery, Databricks, Redshift, ClickHouse or an equivalent. If your events are locked inside a tool that will not export them, this is not the right platform yet.
Where the answers come from
Step 1
The Configuration Agent builds your semantic layer
Point Mitzu at the warehouse and the Configuration Agent scans it: event tables, dimension tables, user and account identifiers, join relationships. It recognises the common shapes — Segment, Snowplow, Firebase, GA4, and custom event schemas — and samples real property values so filters suggest what is actually in your data. Your analyst reviews and adjusts. Nobody hand-writes YAML.
Step 2
The Analytics Agent assembles the analysis
Ask a question in the app or in Slack. The Analytics Agent turns it into an analysis specification: for a funnel, the first event, the subsequent steps, the conversion window and the breakdown; for retention, the cohort definition, the return event and the time granularity. A single question can fan out into dozens of tool calls when the answer needs an investigation rather than a chart.
Step 3
A deterministic engine writes the SQL
The agent never authors SQL. Specifications go to Mitzu's deterministic query engine — the same code that has generated SQL for the product for years — so a funnel always enforces its conversion window and a retention chart always buckets its cohorts correctly. The same specification always produces the same SQL — run against whatever your warehouse holds at the time — and that SQL is there for your analyst to read.
Identify drop-offs and optimize conversion
Build conversion funnels in seconds and understand where users drop off. Analyze funnels by any user property or behavior to identify optimization opportunities.
- Visual funnel builder - no SQL required
- Breakdown funnels by any user property
- Time-to-convert analysis
- Compare funnels across user segments
Funnel Analysis
Signup to Conversion
Overall Conversion
24%
Time to Convert
3.2 days
Measure and improve user retention
Understand how well you retain users over time. Identify what actions drive retention and build features that keep users coming back.
- Cohort-based retention heatmaps
- N-day, unbounded, and bracket retention
- Compare retention across segments
- Identify retention drivers automatically
Retention Cohorts
Weekly analysis
| Cohort | D0 | D7 | D14 | D21 | D28 | D35 |
|---|---|---|---|---|---|---|
| Week 1 | 100% | 72% | 58% | 51% | 48% | 46% |
| Week 2 | 100% | 69% | 55% | 49% | 45% | - |
| Week 3 | 100% | 74% | 61% | 54% | - | - |
| Week 4 | 100% | 71% | 57% | - | - | - |
| Week 5 | 100% | 68% | - | - | - | - |
Segment users by any property or behavior
Create powerful user segments based on demographics, behavior, or any custom property. Understand how different user groups interact with your product.
- Behavioral segmentation
- Property-based segments
- Dynamic cohorts that update in real-time
- Cross-segment comparison
User Segments
By engagement level
Monitor your product in real-time
Track key metrics and events as they happen. Identify trends, anomalies, and patterns that impact your product's performance instantly.
- Live event streaming dashboard
- Customizable metric alerts
- Anomaly detection
- Export to any downstream tool
Real-time Metrics
Live dashboard
Active Users
2,847
Events/min
15.2K
Avg Session
4m 32s
Recent Activity
The analyses, and the pages that explain them
Funnel analysis
Define a multi-step flow, set a conversion window, and see where users fall out. Break the funnel down by any event or user property to find the segment that converts worst.
Read moreRetention analysis
N-day, unbounded and bracket retention against a cohort you define. Compare curves across acquisition channels, plans or onboarding paths to see which behaviour actually predicts a user staying.
Read moreCohorts
Group users by what they did, not just what they are — everyone who hit activation in their first week, everyone who churned after a price change. Cohorts live in Mitzu's semantic layer and stay current as new events land.
Read moreSegmentation
Count events or users over time, split by any property, with filter values sampled from the warehouse so you pick real values instead of guessing at region codes and plan names.
Read moreUser journeys
Start from an event and watch the paths that follow it, ranked by volume. Useful for the flows nobody designed on purpose — the loops, the backtracks, the exits.
Read moreB2B account analytics
Roll behaviour up from users to accounts, teams or workspaces. Because the events sit next to your billing and CRM tables, account health can be measured against contract value without an export.
Read moreUser lookup
Open a single user and read their event timeline end to end. The fastest way to answer a support escalation or check what a churned account actually did before it left.
Read moreDashboards
Save any answer the agent produces onto a dashboard, and every saved insight becomes context the agent draws on for the next question.
Read moreWarehouse-native vs first-generation product analytics
| Dimension | Mitzu (warehouse-native) | First-generation platforms |
|---|---|---|
| Where your events live | In your warehouse. Mitzu queries them in place. | In the vendor's storage, shipped there by their SDK. |
| What you pay for | Seats. Warehouse compute is yours to tune. | Event volume, so the bill grows as the product does. |
| Joins to billing, CRM and support data | Native — those tables are already in the warehouse. | Only what you re-ingest as events or user properties. |
| What the analytics agent can see | Everything modelled in the warehouse, dbt included. | Only what the vendor ingested. |
| Product analytics methodology | Funnels, retention, cohorts and journeys in the engine. | Mature and well established — this is what they built. |
| Data residency and egress | No customer data leaves the warehouse to be analysed. | Events are copied to the vendor's region and retained there. |
First-generation platforms are good at product analytics. They invented most of the methodology, and their own analytics agents answer behavioural questions well. The limit is architectural: an agent can only reason over data the vendor ingested, so the questions that need a join — feature usage against contract value, retention against support volume, activation against ad spend — stop at the edge of the silo.
Mitzu keeps the methodology and moves the boundary. Because queries run in your warehouse, the same agent that builds a funnel can join that funnel to any table your data team already models, and the numbers reconcile with the ones your analysts publish because they come from the same tables.
50%
Faster insights
100%
Data accuracy
0
Data movement
<5min
Setup time
Connect to your data warehouse in minutes
Get instant answers with AI Agents
Skip the SQL. Ask questions in plain English and get instant, accurate insights directly from your data warehouse. Our AI understands your metrics and delivers actionable answers.
- Natural language queries - no SQL required
- Context-aware answers based on your data model
- Automated anomaly detection and alerts
- Follow-up questions for deeper analysis
Mitzu AI Agent
Ask anything about your data
Users who use the 'Dashboard' feature within their first week have 3.2x higher 30-day retention. Only 34% of new users currently discover it.
Enterprise users activate at 78% vs 52% for SMB. The biggest drop-off for SMB is at the 'Invite Team' step - consider simplifying this flow.
Frequently asked questions
Ready to understand your users better?
Get started with warehouse-native product analytics. Connect your data warehouse and start analyzing user behavior in minutes.