METHOD, read 13 Nov 2026
Weights fixed first, evidence second, scores last, and the scorer never knew the names
Download the scoresSee the index
The index answers one question: which tools label data as ready for an AI agent by rule, across the stack, and hand that label to the agent. Seven criteria and their weights were written down before any tool was researched or scored, and they have not changed since.
Every score from 0 to 10 rests on facts in the evidence files, each with a source URL, a quote and the date it was read. The page for each tool lists those facts, and the datasets page offers them for download.
The criteria and their weights
| Criterion | What is scored, with evidence | Weight |
|---|---|---|
| Labels by rule | Readiness labels (certified, healthy, free of sensitive data, ready for AI, or the tool's own words for them) applied from rules and applied again as metadata changes, with no person marking each asset | 25 |
| Signals a rule can read | Lineage, usage, ownership, health or quality, sensitivity, freshness and documentation, from the tool itself or from tools it reads as sources | 20 |
| The label reaches the agent | The label is served to AI agents over MCP, SDK or API at query time, can be filtered on, and is scoped to the task or the person | 20 |
| Reach across the stack | Warehouse, transformation and BI assets all carry the label, as documented | 15 |
| Re-checks itself | The label is evaluated again on a schedule or on change, and withdrawn when a condition breaks | 10 |
| Evidence beyond the vendor | Named customers, analyst coverage, partner pages | 5 |
| Price transparency | Published prices or a published pricing model | 5 |
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 scoring guide for each criterion
Scoring guide per criterion, written with the weights and before any research or score. It interprets the published anchors (0, 1-3, 4-6, 7 vendor-only ceiling, 8-10 second source) and changes no weight or criterion.
- Labels by rule
- Score how far readiness labels (certified, endorsed, verified, trusted, healthy, PII-free, AI-ready or the tool's own term) are applied by rules or policies the customer writes, without a person marking each asset, and re-applied as metadata changes. Manual certification only (a steward clicks 'certified'): 1-3. Rule-based tagging of a narrow kind (classification of PII only, or auto-tag on one signal): 3-5. Documented general rule engine that sets readiness labels from several conditions: 6-7. Above 7 needs a second source.
- Signals a rule can read
- Score the range of signals a readiness rule can read: lineage, usage or popularity, ownership, health or quality test results, sensitivity classification, freshness, documentation completeness. Count signals the tool produces itself or reads from named source tools. One or two signals: 1-3. Three to four: 4-5. Five or more documented as usable in rules or policies: 6-7.
- The label reaches the agent
- Score whether the readiness label is served to AI agents at query time over MCP, SDK or API, whether an agent or the call can filter on it, and whether what is served is scoped to the task or the person. A general metadata API only: 1-3. An MCP server or agent API that returns labels or certification status: 4-6. Label-aware filtering and persona or task scoping documented: 7.
- Reach across the stack
- Score whether warehouse tables and columns, transformation models (dbt or similar) and BI assets (dashboards, reports, fields, metrics) all carry the label, as documented. Warehouse only: 1-3. Warehouse plus one of transformation or BI: 4-5. All three layers documented: 6-7.
- Re-checks itself
- Score whether labels are re-evaluated on a schedule or on change and withdrawn or flagged when a condition breaks (for example a failing test removes certification). Static labels: 0-2. Alerts when a condition breaks but label unchanged: 3-5. Documented automatic re-evaluation and removal: 6-7.
- Evidence beyond the vendor
- Score named customers, analyst coverage (Gartner, Forrester, IDC etc.), partner pages listing the vendor. Vendor-published only: up to 7. A third-party page confirming a customer or analyst placement: 8-10.
- Price transparency
- Published dollar prices: 7-10 (8+ if a third party such as a cloud marketplace listing confirms). Published pricing model with no dollar figures: 3-5. Contact sales only: 0-1. Free open-source core with paid cloud counts as a published model; free with published paid tiers counts as prices.
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.
The kinds of evidence
- Vendor: the vendor's own documentation, pages or press release. It is written as the vendor's claim and scores at most 7 on its own.
- Third-party: a page by someone other than the vendor, such as a partner listing, an analyst note or a customer write-up. It lifts the ceiling above 7.
- Observed: something read directly in public code, a package registry or a release record. It also lifts the ceiling above 7.
- None found: a search of the documentation and pages read on the read date that found nothing. It scores 0 and is written as none found on that date, never as a missing feature.
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 | Rules | Signals | To agent | Reach | Re-check | Evidence | Price |
|---|---|---|---|---|---|---|---|
| A data team told to put an AI agent on the warehouse this quarter | 25 | 15 | 30 | 15 | 10 | 0 | 5 |
| A chief data officer who must tell the board which data is ready for AI | 25 | 20 | 5 | 20 | 10 | 15 | 5 |
| An analytics engineering team on dbt with thousands of models and no owner list | 25 | 30 | 10 | 15 | 15 | 0 | 5 |
| A company whose metrics disagree between dashboards | 15 | 25 | 10 | 30 | 15 | 0 | 5 |
| A regulated company that may let agents use certified data only | 30 | 15 | 25 | 10 | 15 | 5 | 0 |
- Agent this quarter: An agent going live this quarter needs readiness labels it can actually read at query time, set by rules rather than a review backlog.
- Board readiness: A board answer needs a rule-based label across the whole estate, backed by signals the CDO can name and a vendor others have used.
- dbt, no owner list: With no owner list, readiness has to be inferred from usage, lineage and test signals and re-applied as models change.
- Metrics that disagree: Disagreeing metrics are found where the label reaches the BI layer and is re-checked when the logic underneath changes.
- Certified data only: A certified-only policy holds when certification is applied by rule, withdrawn when a condition breaks, and enforced in what the agent is given.
How ties are broken
Ties on the total are broken by the higher score on the heaviest criterion, then by name. On the main ranking the heaviest criterion is labeling by rule, at 25.
Where a situation, a head-to-head or an alternatives list has a tie at the top, or at Euno's own position, that page says so in one sentence, names who is level and says which rule ordered them: the heaviest weight first, then name.
How the scoring was kept blind
The scorer received the weights, a written scoring guide for each criterion and a packet of facts in which tool names were replaced by codes. It had no web access and was told not to guess which tool was which.
A second scoring pass by a fresh reviewer, working under the same conditions, checked every cell against its cited facts and changed 11 of 119 cells, each for a stated evidence or guide reason. The codes were mapped back to names only after the last score was fixed.
What the scores do not cover
The index does not test a tool on live data. It reads what each vendor documents and what public code shows, so a feature that exists but is undocumented on the read date scores as none found.
Breadth facts such as module counts and integration counts carry at most 10 percent of any ranking and enter only through the systems a tool names.
The packet keeps customer names and categories as published, so a reader who knows the market could guess some tools. The scorer was told not to try.
Sources
- Ready Mark scores, CSV, read 13 Nov 2026
- Ready Mark evidence, CSV, read 13 Nov 2026
Last reviewed 13 Nov 2026.