Trust and FAQ

Short answers to common evaluation questions. Each answer links to the guide that documents it. Where something has not been decided or validated, this page says so.

Which edition should I start with?

Start with BigQuery if you want realistic retail data at order-item grain; choose Snowflake if your team already works in Snowflake and a small seed-based example is enough. The edition comparison lists data, account needs, costs, and models side by side. You do not need to run both.

Are the two editions equivalent?

No. They share a source → staging → marts layout and a gross completed_revenue definition, but they use different data and grains. Snowflake has no product or item rows, and its cancelled status is not BigQuery's returned status. See the mart contract.

What has been tested?

• Pull-request CI runs a credential-free dbt parse for each edition with a placeholder profile. The changelog records parse runs passing. Parsing checks project configuration only.
• Dependencies are pinned in edition-specific Python 3.11 hash locks, and CI checks direct requirements against them.
• No authenticated warehouse build or warehouse test run has been recorded for either edition. See the compatibility matrix.
• No independent first-run evaluation has been recorded yet. The first-run checklist is the template for one.

Will running it cost money?

It can. Everything runs in your own cloud account.

• BigQuery: queries run and are billed in your Google Cloud project, even though the public source tables are hosted by Google. The profile's 1 GB maximum_bytes_billed setting applies per query; it is not a total spending cap. BigQuery Sandbox allows limited exploration without billing. See BigQuery setup and cost checks.
• Snowflake: builds use warehouse compute in your account. Use a small development warehouse with a short auto-suspend interval. See Snowflake operations and cost checks.
• The cost control guardrails cover both clouds.

Is any of the data real?

• BigQuery reads Google's public TheLook eCommerce dataset, which describes a fictional retailer. The curated marts omit direct identifiers such as names, email addresses, and street addresses. The kit ships SQL only, not TheLook rows; do not redistribute its data or screenshots. Attribution is in NOTICE.
• Snowflake uses synthetic seed rows written for this repository, with placeholder names such as Ada Example.

Do not load real customer data into the example projects. See the BigQuery overview and Snowflake overview.

Do I need to share credentials?

No. Credentials stay in your local environment variables and an edition-local .dbt/profiles.yml that Git ignores. Pull-request CI uses placeholder profiles and no credentials. Optional warehouse CI uses your own repository secrets (Snowflake) or keyless Workload Identity Federation (BigQuery). See CI security review.

Can I use it commercially? What is the licence?

The kit is source-available under an evaluation licence (LICENSE). You may run and modify it free of charge for evaluation, learning, and non-production testing. Production use, client delivery, and redistribution need a separate written licence from the owner. TheLook data is not part of the kit. Your queries run under your own Google Cloud terms and any terms on the dataset's Marketplace listing, which the owner has not yet reviewed; review those listing terms yourself before relying on or sharing the data or results. The reasons and alternatives are in decision 0001.

Is there support or an update policy?

Support is best effort through GitHub issues, with no warranty, service level, or update commitment. See SUPPORT.md. Paid support and update entitlements have not been decided; see release readiness.

Is there a dashboard?

The BigQuery edition includes a Looker Studio walkthrough that maps the marts to sample charts. Dashboard queries are billed to the selected Google Cloud project. The Snowflake edition has no dashboard configuration.