What is included
Enterprise data engineering discipline, scoped for SMB surface area. Every layer of the stack is owned by one engineer, not outsourced to a vendor per step.
Source connectors and ETL
Pull from the systems the team already pays for: Shopify, Stripe, Xero, HubSpot, Klaviyo, Google Analytics, bespoke SQL. Scheduled, idempotent, monitored.
Managed warehouse
Postgres or BigQuery depending on volume. One schema per tenant, partitioned for growth. No shared tables across clients.
Tables the team can read
Versioned SQL models for the business entities that matter: orders, customers, products, sessions, cohorts. Column names that match the operator's language.
Dashboards that answer one question each
Metabase or Looker Studio, focused boards per role: founder, head of ops, head of marketing. Every chart carries the SQL and a refresh timestamp.
Data-quality checks
Row counts, freshness, referential integrity, outlier detection. Failures go to the operator channel before the dashboard lies.
Weekly data review
Thirty minutes to walk the dashboards, flag what is broken, decide what to build next. Not a monthly report; a working session.
First 90 days
From zero visibility to three trustworthy dashboards. Each milestone is a shipped artifact the operator can open.
Inventory and schema v0
- Map every source of truth the team uses
- Pick the three highest-value streams
- Provision warehouse and tenant schema
- Schema v0 committed to cohort repo
Ingestion and first dashboards
- Connectors for the top three sources live
- Modelled tables for orders and customers
- Two focused dashboards shipped
- Quality checks wired to alerts
Trust and expand
- One more data source integrated
- Third dashboard by role
- Backfill historical and reconcile totals
- Weekly review cadence locked in
Proof this is not greenfield
Engineer credentials plus runtime references. The SMB package is the downshift of patterns already running.