What cruise runs on.
Nine pillars.
Every criterion shows its evidence.

This card grades products, not companies. The scored unit is one product line inside one peer group — Reservations, inventory & core commerce; Revenue, pricing & decisioning; Guest engagement, CRM & loyalty; Onboard operations & commerce; Distribution & booking technology; Connectivity & ship–shore; or Experience & destination commerce. A vendor selling into two of those functions gets two cards, weighted for what each function actually needs: security counts for a quarter of a connectivity card's score and nothing on an onboard-operations card's AI criterion; AI counts for a fifth of a revenue-and-pricing card. No company-level grade is published, and rank runs only within a peer group.

A product is admitted to the scored field on one of two tests: dated 2024–26 evidence of confirmed production in a cruise commercial function at a brand the Cruise Commercial Excellence Index scores — confirmed by the client, or observable running on the client's own surface — or confirmed production at three or more cruise operators of any size. Confirmed means a go-live, not an announcement: a “selects” release or a partnership is a sales result, and this index reads it as a selection, not as evidence a system is running. Products that clear only that lower bar sit in a listed register and can be graded the edition a go-live is confirmed.

Two categories sit outside the scored field by rule, not by omission. In-house and consultancy co-built systems — Carnival's YODA, Royal Caribbean Group's own yield models, TUI Cruises' revenue-management system built with BCG X, Norwegian Cruise Line's Nora built with Slalom — are not products a buyer can license, so they are recorded in an unscored Build Register with their public evidence rather than graded. Infrastructure and tag management without a commercial decision attached to it — cloud migrations, tag managers, a biometric component inside another company's app — is excluded outright. Signalry / Shorelight, a guest-intelligence product Ethan Hawkes, an H&K partner, builds and has demonstrated to CEI-scored brands, is excluded for conflict rather than scored.

The gaps are disclosed rather than smoothed over. Enterprise cruise technology sells under confidentiality, so a vendor whose clients will not be named on the record reads lower here than its installed base may deserve — that is a statement about the record, not the product. Corporate-level evidence counts on only three criteria across the whole method, so a large parent's financial strength does not lift a thin product-level record. And because production requires a named, checkable client, the two places this edition found the most visible AI activity — horizontal platforms with the loudest agent counts, and operators' own in-house systems — sit partly or wholly outside what a scored card can capture, in the listed register or the Build Register instead.

Every criterion state cites the sources it rests on, and every source declares an evidence mode: direct primary (95), official observed surface (85), multi-source primary (80), official source snapshot (78), independent reported (70), high-confidence proxy (65), moderate-confidence proxy (45). The mode values describe evidence reliability; they do not set any score.

Coverage is the share of a card's applicable criterion points with public evidence either way (met, partly met or not met). Under 60% a card is not graded; from 60% to 79% its grade is provisional and published with a range, reading every unevidenced criterion first as zero and then as partly met.

Confidence (method counted_citations_v2) keeps the CEI's four components and weights — breadth 25%, specificity 25%, recency 30%, primary coverage 20% — recalibrated for criterion-level citation: full breadth at 12 distinct sources; a source cited on three or more cards is not specific; recency is the mean freshness of the sources cited (100 within a year of the check, 70 within two, 40 within three or undated, 10 older); primary coverage is the share of applicable criteria resting on primary evidence.

Fragility is the share of a card's earned points that rest on a single source or on proxy evidence: low under 25, moderate 25–49, high 50 and over. It answers how much of a grade would move if one citation fell.

Evidence provenance follows the CEI rule: a card's profile is the set of its per-pillar evidence modes, and a profile carried by more than three other cards is marked shared.

Verification runs in both directions, because a card can be wrong either way. A refuting skeptic re-opened every cited source and attacked every point a card earned; an upgrading skeptic searched for evidence behind every criterion that scored nothing. A reconciler applied a change only when the evidence held up on its own inspection, and where it could not settle a disagreement it took the lower state and marked the cell contested, with both sources shown.

• Not a measure of capability hidden by confidentiality. A vendor whose clients will not be named reads lower on the public record than its installed base may deserve; the card says so, and a client's confirmation corrects it.
• Not a vendor-selection recommendation. Grades describe the public record of one product in one function; they do not replace a buyer's due diligence.
• An outage or breach is attributed to a product only when the client, the provider, a regulator, a court or two independent reports say so, and only final or confirmed events appear as red flags.
• In-house and consultancy-built systems at cruise lines are recorded in the Build Register, not graded.
• No vendor was surveyed or interviewed for this first edition, and no confidential information entered any card.