Fixed-scope BigQuery, dbt and Looker stack
Your data modelled, tested and documented. One definition per metric.
Fixed-price analytics engineering sprint on Google Cloud: tested and documented dbt models in BigQuery, semantic layer exposed in Looker or in your BI tool, one definition per metric, ready for BI and for AI agents. 4 to 6 weeks, from $15,000 excl. VAT, quoted after a written brief.
The problem
When finance, sales and the executive team do not share the same revenue number, BI is not the culprit; the missing piece is a single place where "revenue" is defined. This engagement builds that place: tested and documented dbt models in BigQuery for facts and dimensions, a thin LookML layer on top (no business logic hidden in dashboards), a metrics glossary versioned with the code, and role-based access. It is also what gets your data ready for AI: an agent, Gemini or Claude through MCP, only answers correctly on single, tested, documented definitions, and this stack gives it those without a separate project. Delivered in four to six weeks depending on the number of domains, with automatic checks on every change and a written handover to your team.
Who it is for
- Data teams at scale-ups, mid-size companies and subsidiaries on BigQuery, with or without Looker, whose dashboards contradict each other
- Teams on Metabase or Looker Studio who want clean dbt modelling before choosing or changing their BI tool
- Finance leaders who want one official number per KPI, traceable to the query
- Teams preparing AI agents on their data who need a reliable semantic base
What you receive
- Written workshop defining the 15 to 30 metrics that matter
- dbt models (staging, intermediate, marts) with tests and documentation
- Semantic layer: LookML (generated views, explores per domain, single measures, drill-downs) or exposure for your BI tool
- Metrics glossary versioned in the repository
- Descriptions and metadata ready for AI agents (Conversational Analytics, MCP)
- Role- and row-level access control (access grants, user attributes)
- Automatic checks on every change (dbt tests, LookML validation)
- Handover: documentation, explainer videos, two weeks of written questions
Out of scope
- Operating source pipelines day to day (loading Shopify, Stripe, HubSpot or ERP exports into BigQuery is scoped in the quote; running them afterwards is the retainer's job)
- Dashboard-by-dashboard migration beyond two reference dashboards
- Looker licences and BigQuery costs
Prerequisites
- Source data in BigQuery, or exports and APIs we can load (Fivetran, Airbyte, native exports)
- One business owner per domain, reachable in writing
- A Git repository; a Looker instance if you have one, or one being provisioned
How it runs
01
Week 1: definitions
Inventory of metrics, discrepancies and sources. Glossary v1 approved in writing.
02
Weeks 2 to 4: modelling
Tested and documented dbt models, LookML layer, first domain shipped by end of week 2.
03
Week 5: reference dashboards
Two dashboards re-pointed to the layer, compared with the old figure, gaps explained.
04
Week 6: handover
Documentation, automatic checks, asynchronous training, extension plan for the next domains.
StackBigQuery · dbt · Looker · LookML · GitHub Actions · Terraform (optional)
Frequently asked questions
Why dbt and LookML, not LookML alone?
LookML is excellent at exposing and securing data, less so at transforming and testing it. dbt carries the transformations and tests, LookML stays thin and readable. It is also what lets you serve Looker Studio, notebooks or agents from the same definitions.
How many domains in the base engagement?
One complete domain (for example revenue and orders). Each additional domain is quoted separately.
What if we only have Metabase or Looker Studio, not Looker?
The dbt layer in BigQuery is built the same way and your BI tool connects to it. Moving to Looker later becomes a formality, if you decide to.
How does this get our data ready for AI?
An analytics agent has no common sense: it answers with whatever the semantic layer gives it. Tested models, one definition per metric, up-to-date descriptions: that is exactly what Gemini, Conversational Analytics or an MCP server need to answer correctly. The Analytics agents pilot comes next, on that base.
How does remote collaboration work?
Everything is written: brief, decisions, code changes, glossary. A weekly written status replaces the meeting; short videos explain what text cannot. No calls.
Reference implementations
Code you can read before you write to us.
Three complete, public stacks built the way we build them for clients: dbt on BigQuery, LookML, dashboard, linter and continuous integration. One per sector.
E-commerce
Margin, average order value, repeat rate, one revenue number. Public thelook_ecommerce dataset, 13 metrics, 3 explores.
Healthcare
30-day readmissions, length of stay, chronic conditions, patient field protection. Synthetic Medicare data in OMOP format, 17 metrics.
Fintech
Payment volume, chargebacks, churn, loan book and delinquency. Generated synthetic data, customer history, 56 metrics, 134 tests.
Guides
9 min read · 6 September 2026
Looker for e-commerce: margin, AOV, repeat, one revenue number
Ten e-commerce metrics defined once in dbt and exposed in Looker: net revenue, margin, AOV, repeat rate, returns, grain trap. With a public reference repo.
Read the guide →10 min read · 6 September 2026
Looker for fintech: TPV, chargebacks, churn, loan book
Settled vs attempted TPV, take rate, chargeback ratios against Visa and Mastercard thresholds, customer-month churn, point-in-time loan book, SCD2. Public repo.
Read the guide →9 min read · 6 September 2026
Looker for healthcare: readmissions, length of stay, patient data
OMOP in brief, 30-day readmission and length of stay defined once in dbt, prevalence by concept hierarchy, patient fields protected in Looker. Public repo.
Read the guide →Describe your context
Three lines are enough: your systems, what is broken, the deadline. Written answer within one business day.
Request a quote