Marketing asks
"What counts as revenue?"
Three teams have three answers. The truth lives in someone's Notion doc from 2024 and a CASE statement nobody wants to touch.
Private beta · BigQuery
Your warehouse, in plain English.
Truefold connects to BigQuery and generates the semantic layer your team has been maintaining by hand — dimensions, metrics, relationships, and the questions worth asking. Auto-updated, human-reviewed.
- Read-only access
- Your data never leaves BigQuery
- Set up in under 10 minutes
Input · BigQuery
orders
- tot_amt_usd_2
- cstmr_lftm_v
- prd_cat_id
customers
- cstmr_seg_v3
events
- utm_src
- utm_med
87 columns · 4 tables
Output · Semantic layer
Order TotalMETRIC
SUM(orders.tot_amt_usd_2)
Customer LTVMETRIC
approved · 94% confidence
Product CategoryDIMENSION
12 values · approved
Active CustomerDIMENSION
42 of 87 columns mapped
The translation tax
You're the only one who knows what the columns mean.
Every business question becomes a forensic exercise. Every new hire needs a glossary tour. Every dashboard quietly disagrees with another one.
Finance asks
"What's an active customer?"
Finance says 90-day. Marketing says 30-day. Your churn model uses both and nobody noticed.
Engineering ships
cust_lftm_v_2
Shipped two years ago by an engineer who left. Used in seven dashboards. Documented nowhere. You are the documentation.
What it does
One layer, four jobs.
Generates the semantic layer.
Connect BigQuery and Truefold auto-documents schema, dimensions, metrics, and relationships. Minutes, not months.
Suggests the questions worth asking.
Eight to twelve pre-built business questions per warehouse, tailored to what your data actually contains.
Watches for drift.
A continuous auditor scans three times a day. New column appears, metric breaks, definition drifts — you get a clear alert with the affected models named.
Speaks fluent SaaS.
Optional Tier-1 connectors pull directly from Shopify, Meta Ads, Google Ads, and Klaviyo. No warehouse required to start.
How it works
Connect once. Reviewed in a morning. Maintained forever.
- 01
Connect.
Read-only service account to BigQuery, or paste API keys for Tier-1 sources. ~5 minutes.
- 02
Generate.
Truefold scans your schema, samples values, drafts dimensions and metrics, and proposes the questions your team should be able to answer. ~10 minutes for most warehouses.
- 03
Review and publish.
Your team approves, edits, or rejects. The auditor takes over from there.
What you get
The output your team actually uses.
Truefold doesn't generate prose. It generates structured metadata your tools, dashboards, and AI agents can read directly.
Dimensionapproved · 96%
customer_status
Whether a customer can receive marketing or transactional messages.
Values · 4
subscribedunsubscribedbouncedcomplained
Mapped from
- shopify.customer.tags
- klaviyo.person.consent_status
Metricapproved · 94%
gross_revenue
Sum of line-item revenue across all shipped orders, gross of refunds.
Definition
SELECT
SUM(quantity * unit_price) AS gross_revenue
FROM shopify.order_line ol
JOIN shopify.order o ON o.id = ol.order_id
WHERE o.status = 'fulfilled'Alert · 2h ago
New column detected: order.discount_code_id
Affects 2 metrics and 1 dimension. Truefold drafted a mapping suggestion and is waiting on a review.
Affected
- metric · gross_revenue
- metric · discount_total
- dimension · campaign_source
Where we are
Honest about what's built and what isn't.
What works today
- BigQuery warehouses
- Direct connectors for Shopify, Meta Ads, Google Ads, Klaviyo
- Auto-generated dimensions and metrics
- Continuous drift detection (every 8 hours)
- MCP server for AI agents to query your layer
- Single-tenant, read-only by default
Not yet
- Snowflake, Redshift, Postgres warehouses (BigQuery-only for v1)
- Automatic schema-drift resolution (we alert; humans decide)
- Exhaustive value discovery (we sample — rare categorical values may be missed)
- Foreign-key inference is good but sometimes needs manual mapping
- SSO/SAML (planned for Team tier)
Who it's for
Built for the people who own the definitions.
Heads of Data · Analytics Leads
The one everyone pings.
You're the one Marketing pings, Finance escalates to, and Eng asks about. Truefold gives you a single layer to point everyone at — and an auditor that flags drift before someone files a ticket.
Analytics Engineers
The documentation layer you've been writing by hand.
dbt is your IDE. Truefold is the documentation layer you've been generating by hand. Export to dbt-compatible YAML; keep your existing models.
Finance · Marketing · Ops
Stop guessing what monthly_active means.
Stop guessing what monthly_active means in this dashboard versus that one. Browse the layer, read the definitions, query through your existing BI tool.
FAQ