Your AI data analyst
The analyst that shows its work.
Symbric Analyst answers business questions with the number, its source, its freshness, and the checks it passed. When Symbric cannot stand behind one, it says so.
Book a Chatsymbric / your agent
Churn by plan, last quarter
Customer grain, calendar quarters. Reconciled against Stripe.
% monthly churn
Monthly churn8.9%
- Source
- postgres.subscriptions, per customer
- Fresh
- 6:04am today
- Asked in
- your agent
- Checks
- 3 of 3 passed
- Stripe reconciliation, within 0.5%
- DAU never exceeds WAU
- Golden query matches
Certified
01
A general AI agent will answer anything. Nothing tells you if it's right.
It never shows how it landed on the number: which metric definition, which time window, which table, which checks, or what query it wrote. What comes back is a number nobody should run the business on.
-
Before
User engaged with a feature96%
What the team published. A human caught it weeks later.
Unverified
-
After Symbric
User engaged with a feature91.6%
- Taxonomy check re-run, 1,464 of 1,464 rows
- Definition computable from code
Root cause captured: the event fired on every data load, so it measured availability rather than engagement.
Certified
Based on real tests with a consumer app, July 2026.
02
What arrives with every number.
- Source
- Which table the number came from, and what it counts
- Fresh
- When the data behind it last landed
- Asked in
- The tool your team already works in
- Checks
- What it had to pass before it was allowed out
Certified means the number named its source, showed when its data last landed, and passed every check it was given.
03
The questions your business runs on.
Revenue by segment, last quarter
% of quarter revenue, change against the prior quarter
Enterprise leads the mix and took share from both smaller segments.
- Source
- postgres.orders, joined to accounts
- Fresh
- 6:04am today
- Checks
- 3 of 3 passed
Certified
RevOps asks
- Revenue by product, segment, and region
- Sales attribution
- Forecast accuracy against actuals
Funnel conversion by step, last 30 days
% of visits reaching each step, and distinct users
The drop is all at the first step: after signup starts, over half continues at each stage.
- Source
- posthog.events, funnel_step
- Fresh
- 6:04am today
- Checks
- 3 of 3 passed
Certified
Growth asks
- Funnel conversion by step and segment
- CAC payback and blended acquisition cost
- Monthly churn and retention cohorts
Feature adoption after the last release
% of active accounts using the feature, at day 30
Adoption cleared the prior release in week one, then flattened after day 14.
- Source
- posthog.events, event_share
- Fresh
- 6:04am today
- Checks
- 3 of 3 passed
Certified
Product asks
- Feature adoption after a release
- Activation and time to first value
- Usage by segment and platform
Win rate by segment, this quarter
% of opportunities won, against last quarter
Mid-market gained the most. Enterprise is the only segment falling.
- Source
- postgres.opportunities, stage history
- Fresh
- 6:04am today
- Checks
- 3 of 3 passed
Certified
Sales / GTM asks
- Pipeline coverage by stage
- Win rate by segment and source
- Quota attainment by team
Renewal rate by cohort, last quarter
% of cohort accounts renewed, and accounts in each cohort
Each newer cohort renews better than the one before it.
- Source
- postgres.subscriptions, per cohort
- Fresh
- 6:04am today
- Checks
- 3 of 3 passed
Certified
Customer satisfaction asks
- NPS and CSAT trends
- Ticket volume and time to resolution
- Renewal rate by cohort
p95 latency by endpoint, last 7 days
p95 milliseconds, daily. The target is 400ms
Search climbed all week and crossed its 400ms target yesterday.
- Source
- postgres.request_logs, per endpoint
- Fresh
- 6:04am today
- Checks
- 3 of 3 passed
Certified
Engineering asks
- p95 latency by endpoint
- Crash-free sessions by release
- Error rates and uptime
Answered from the systems your team already runs.
04
It builds itself from the systems you already run.
Chat interfaceClaude Code, Cursor, ChatGPT, any MCP client
Symbric Analyst
Answers and chartscertified, built per question
Orchestration and checksplans the query, checks the result
Context and semantic layerdefinitions, entities, join paths
Your codebase and warehouseapp code, schema, migrations, models
-
Step 01
It builds your semantic layer.
Symbric derives your metrics, entities, and joins from the systems you already run: your application code, your schema, your warehouse. There is no modeling project to staff.
symbric/contract.ymlbilling-migration #412
metric monthly_churn: grain: customer - source: events.subscription_cancelled + source: billing.subscriptions checks: [grain, stripe_reconciliation] -
Step 02
Every definition is tested.
Grain, invariants, and reconciliation against numbers your team already trusts. A failed check blocks the answer.
eval monthly_churn3 passed, 1 blocked
- graincustomer levelPASS
- stripe_reconciliationwithin 0.5%PASS
- cardinalityDAU never exceeds WAUPASS
- golden_queryexpected 8.9%, got 4.1%BLOCKED
-
Step 03
Queries are planned before they run.
The contract resolves joins, filters, and time windows before anything runs, so the question you asked is the question that gets answered.
Churn by plan, last quarter
- resolvemonthly_churn from the contract
- graincustomer level, one row per account
- joincustomers to subscriptions, 1:n verified
- window2026 Q2, calendar quarters
- verify3 checks before return
-
Step 04
Every answer shows its work.
The result returns with a chart, its source, how fresh it is, and the checks it passed.
Churn by plan, last quarter
% monthly churn
-
Step 05
It keeps up with the business.
Definitions follow the code that changed them. Drift is caught on a cadence, before somebody happens to notice, and the contract updates itself on the pull request.
05
More decisions made on numbers that hold up.
-
Before your first data hire
Get the answers a data function delivers while you decide when to build one, each with its proof attached.
-
Teams with a full roadmap
Recurring questions get answered to your team's standard, so your analysts stay on the work only they can do.
-
Teams shipping fast
Your product changes weekly and the definitions move with it, so Monday's number still means what it meant last quarter.
06
The work of a data function, at the volume you already have.
Without Symbric
- Questions waiting behind a roadmap
- Hours spent reconciling two numbers
- Headcount added to answer questions
With Symbric
- Answers on the day the question is asked
- One definition, wherever the question is asked
- Volume answered without a req
07
What it reads, and where it answers.
Application stack
- Postgres
- Supabase
- PostHog
- Stripe
Warehouse
- Snowflake
- Databricks
- BigQuery
- dbt
Agents
- Claude Code
- Cursor
- ChatGPT
- Any MCP client
- Built from the systems you already run.
- Read-only against your data. It never writes to your database or warehouse.
- We never train models on your data. Model calls run under your own key.