# Weights fixed first, evidence second, scores last, and the scorer never knew the names

METHOD, read 13 Nov 2026

**The short answer.** Eight criteria and their weights were committed on 13 Nov 2026, before any vendor was scored. A subagent scored every tool from the evidence files without being told which tool is the client, and a second reviewed those scores blind. A vendor claim earns at most 7 of 10 until a second source confirms it.

Need to Know ranks tools by one question: which show where sensitive data surfaces, in which reports, to which viewers and to which agents, and control what an agent is told. The eight criteria below turn that question into scores, and the weights are the ones a governance lead in a regulated company would set before looking at any vendor.

The weights were written to a file in the repository first. The first score was committed after it, and the file has not changed since. The scores, the weights and the ranked views on this site are checked against that history on every build, and a build that disagrees does not ship.

## The criteria and their weights

| Criterion | What is scored, with evidence | Weight |
| --- | --- | --- |
| Follows sensitive data into BI | From a warehouse column through transformation to the BI field and the report, across BI tools, as documented | 25 |
| Controls what an agent is told | Context and access scoped to the person asking and applied on each agent call | 20 |
| Shows who actually saw it | Viewers per report and per agent, from usage or access logs | 15 |
| Carries classification downstream | Takes sensitivity labels from where they already exist (the warehouse, a classifier, a security tool) and propagates them along lineage | 15 |
| Acts on it | Alerts, tickets, blocking or masking | 10 |
| Coverage of the stack | Warehouses, transformation, BI and security tools, as documented | 5 |
| Evidence beyond the vendor | Named customers, analyst coverage, partner pages | 5 |
| Price transparency | Published prices or a published pricing model | 5 |

Weights fixed 13 Nov 2026, before the first score, and unchanged since.

## What each score means

- **0:** none found on the read date
- **7:** the most a vendor claim or vendor documentation alone can earn: detailed, specific, covering the whole criterion
- **1-3:** mentioned, partial, or limited to one system or one level
- **4-6:** documented and working, with gaps the documentation shows
- **8-10:** detailed and confirmed by a second source (third-party write-up, or observed directly in public code, schema or release record)

## The rules every score follows

- Every vendor is scored 0 to 10 on every criterion, whole numbers, from the evidence file only.
- A vendor claim scores at most 7 until a second source confirms it (kind third-party or observed).
- None found scores 0 and is written "none found on" the read date, never as a missing feature.
- Breadth facts (module counts, integration counts, headcount, funding) carry at most 10 percent of any ranking; integration counts inform coverage only through the systems they name.
- The total is the weighted sum of the scores divided by 10 (0 to 100 scale, published to one decimal).
- Ties break on the higher score on the heaviest criterion, then by name.
- Once the first score exists, no weight and no criterion changes. A score changes only when a sourced fact changes, logged with the fact.

## How ties are broken

Tied totals are broken by the higher score on the heaviest criterion, then by the tool name. Where two criteria share the top weight, as Follows sensitive data into BI and Acts on it do in the lean team situation (20 each), the heaviest is the first in the published criteria order, which is Follows sensitive data into BI for the site weights.

## The five situations and their own weights

Each situation page ranks the same scores under weights a buyer in that situation would use. These were fixed and committed with the site weights, before any score.

| Situation | Into BI | Agent told | Who saw it | Label carried | Acts | Stack | Evidence | Price |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| [A governance lead who must show where a regulated field ends up](/best-for/regulated-field/) | 35 | 5 | 20 | 20 | 10 | 5 | 5 | 0 |
| [A company connecting AI agents to its analytics data this year](/best-for/agents-on-analytics/) | 15 | 40 | 15 | 10 | 10 | 5 | 0 | 5 |
| [A Microsoft estate: Fabric, Power BI and Purview](/best-for/microsoft-estate/) | 30 | 10 | 15 | 20 | 15 | 10 | 0 | 0 |
| [A mixed stack: Snowflake or Databricks with dbt and two or more BI tools](/best-for/mixed-stack/) | 30 | 10 | 10 | 20 | 5 | 20 | 0 | 5 |
| [A lean data team with no governance hire](/best-for/lean-team/) | 20 | 10 | 10 | 15 | 20 | 5 | 5 | 15 |

Situation weights, each summing to 100. Column order is the published criteria order.

- **A regulated field:** An auditor asks for the path of the field into each report and the people who saw it, with its classification carried along.
- **Agents on analytics:** Before an agent connects, someone must control what it is told for each person it acts for, and see what it was given.
- **A Microsoft estate:** In a Microsoft estate sensitivity labels already exist; the question is how far they travel into Power BI and what happens when they do.
- **A mixed stack:** Across several BI tools, the value is in following a column through dbt into every BI tool in use, with the label carried.
- **A lean team:** With no governance hire, the tool has to act on its own findings and the cost has to be knowable up front.

## How the scoring was kept blind

Each vendor was researched to the same depth: product pages, documentation, release notes, a trust or security page and at least two third-party sources, every fact stored with its quote, its source URL and the date it was read. A second pass re-read six quotes per vendor and re-searched every none-found fact.

A subagent was then given only the weights, a scoring guide and an anonymized evidence packet, and was not told which vendor is the client. A second subagent with fresh context reviewed every cell the same way. Each cell cites the facts it rests on, and a script checks that a score above 7 cites a fact from a second source.

## What the scores do not cover

The packets kept customer names and product categories as the vendors publish them, so a reader who knows the market could guess some vendors. The scorers were told not to try and could not browse.

A score is a reading of public documentation on 13 Nov 2026. It is not a test of a running product, and a tool can do more than its documentation says. Where nothing was found the cell reads none found on that date, and it scores 0.

Breadth facts such as integration counts, headcount and funding carry at most 10 percent of any ranking. Euno's pricing page and home page describe two pricing models and neither gives a dollar price, so price transparency is scored on what each page says.

## Sources

- Need to Know scores, CSV: https://aidatagovernancetools.com/datasets/need-to-know-scores.csv (read 13 Nov 2026)
- Need to Know evidence, CSV: https://aidatagovernancetools.com/datasets/need-to-know-evidence.csv (read 13 Nov 2026)

Last reviewed 13 Nov 2026. Canonical: https://aidatagovernancetools.com/method/
