# Metacenta

> A Data & AI Technical Review: deterministic checks over a dbt project and repository metadata. The output is a written review, a per-finding worklist and a client portal. Scores cover testing, documentation, lineage, modelling, maintainability, source monitoring, execution, repository hygiene, ownership, actual usage, AI lifecycle, orchestration and the BI semantic layer.

- Reads dbt artefacts (manifest.json, and optionally run_results.json, sources.json, catalog.json) plus read-only repository metadata.
- Needs no access to a warehouse or to production data. A client may grant a temporary, read-only connection for a named question. What it reads depends on the service: for a Technical Review it reads metadata such as query history, never the contents of a table. profiles.yml is excluded by name because it holds credentials.
- Migration Assurance compares a data estate before and after a migration, asset by asset, from their definitions. A client may also choose to compare the data: they run a published read-only script and send a verdict per table pair, or the counts and relative changes it computed. It never returns a row, and sending nothing is a complete engagement. Fixed price: £3,500 for one migration, and £750 for each re-run before cutover; larger estates are scoped first.
- The whole review runs with zero LLM calls; an interpretive pass over the findings is opt-in and excludes owner identities.
- Where the build artefacts are supplied it also reports where build time goes: models rebuilt in full every run, and tests costing more than the asset they check.
- Findings are reviewed by a person before a client sees them. The review names what it could not see.
- Fixed price: £2,000 for a review of one dbt project and its repository; larger estates are scoped and priced before work starts. A reassessment after changes is £500. Invoiced after scoping; there is no subscription and no self-serve signup.
- Each page below links its markdown version: the page's URL plus `.md`, or the page's URL with `Accept: text/markdown`. Same content as the page. /llms-full.txt is every page in one file.

## Pages

- [Data & AI Technical Review](https://metacenta.com/index.md) — An example report, how a review works, and what it costs (~833 tokens)
- [Security and scope](https://metacenta.com/security.md) — What you send, how it's handled, what a review can't see, the fifteen areas and the deliverables (~2,526 tokens)
- [How a review is assessed](https://metacenta.com/standard.md) — The versioned check suite: dimensions, scoring, severity, confidence and the published rules we cite (~1,415 tokens)
- [Public dbt benchmark](https://metacenta.com/benchmark.md) — How often each check fails across public dbt projects on GitHub, with a free CSV (~1,623 tokens)
- [Pricing](https://metacenta.com/pricing.md) — Every price, what each service includes, and the tools and warehouses we work with (~1,202 tokens)
- [Migration Assurance](https://metacenta.com/migration.md) — A data estate compared before and after a migration; the data compared only if the client chooses to send reconciliation results (~1,093 tokens)
- [The rules a review checks](https://metacenta.com/rules.md) — One page per rule: what it looks for, why it matters, how to fix it, and when breaking it is fine (~3,387 tokens)
- [Sample review](https://metacenta.com/sample.md) — Two complete reviews as a client receives them: a 200-model dbt project we built, and dbt Labs' public jaffle_shop_duckdb (~14,393 tokens)
- [What we do with what you send](https://metacenta.com/privacy) — Artefact handling, retention, and what never leaves the engagement
- [Terms of engagement](https://metacenta.com/terms) — Scope, deliverables and the limits of the review