CRM Sync — Feature Specification

Accessibility & Machine-Index Score — the measurable foundation of the accessibility offer

Document ID: CRM-FEAT-003 Version: 1.0 Date: 2026-07-06 Status: Published Classification: Public Parent: CRM-FUNC-SPEC-001

1. Executive Summary

The core offer of CRM Sync is accessibility: making connected commerce data usable by every enterprise team and by AI agents, through one governed gateway. Accessibility is only a credible promise if it is measured. This feature specifies the Accessibility & Machine-Index Score — a repeatable, automated quality gate that scores every published surface on two dimensions:

• Accessibility Score (human reach) — conformance of a surface to the automatable items of the Webflow Accessibility Checklist (WCAG-aligned): language, structure, landmarks, alternative text, keyboard and form semantics, media captioning.
• Machine-Index Score (agent reach) — how readable a surface is to search and answer engines, per the automatable items of the Webflow AEO Checklist: crawlability, structured data, metadata, semantic sectioning, scannable formats, freshness and authorship signals.

These are two faces of one objective. A surface that a screen-reader user cannot navigate and a surface that an LLM cannot parse are the same failure — data that is present but not accessible. This feature makes that failure visible, ranked, and fixable before publication.

2. Why this is foundational, not cosmetic

The platform already governs access to data at the data plane (scoped, revocable, PII-free by construction). This feature governs accessibility of the rendered surface — the last mile where humans and agents actually consume the data. It is the presentation-layer complement to the access-governance substrate: the same discipline (measurable, repeatable, fail-toward-safe), applied to reach.

3. Functional Requirements

FR-AXI-01 — Two-dimensional score. Every evaluated surface receives an Accessibility Score and a Machine-Index Score, each 0–100 with a letter grade (A ≥90 · B ≥80 · C ≥70 · D ≥60 · F <60).

FR-AXI-02 — Standards-anchored checks. Scored checks are drawn only from the automatable items of the two published checklists (§1). Each check records its source criterion (WCAG or AEO reference) so a score is defensible and auditable, not a black box.

FR-AXI-03 — Weighted, severity-aware scoring. Each check carries a severity (critical / high / medium / low) that weights its contribution. A score is the weighted proportion of applicable checks that pass — so a missing image alternative (critical) moves the score more than a missing social-preview tag (low).

FR-AXI-04 — Scope awareness (page vs. fragment). Whole-page requirements (document title, canonical URL, structured data) are evaluated only on full pages. Embeddable fragments are not penalized for lacking them; a fragment therefore receives a meaningful Accessibility Score and a Machine-Index result of n/a, because answer-engine indexing is a whole-page property.

FR-AXI-05 — Ranked remediation. Output includes a Top Gaps list ordered by severity × prevalence across surfaces — a fix-first worklist, not an undifferentiated dump.

FR-AXI-06 — Honest boundary (manual-review carve-out). Items that cannot be decided from static markup — colour contrast, visible focus, target size, caption content quality, answer accuracy in live AI engines, E-E-A-T truthfulness — are listed as manual review and never scored. The score never overstates what automation can prove.

FR-AXI-07 — Continuous evaluation. The score runs in continuous integration on every change, and can also be run on demand against local files or against live URLs (real rendered HTML). It emits a machine-readable report suitable for dashboards and trend tracking.

FR-AXI-08 — Promotable gate. The score begins as a non-gating baseline (reported, not enforced) and is promotable to a required release gate at chosen thresholds (default: Accessibility ≥ 80, Machine-Index ≥ 70). This mirrors the platform's forward-deploy posture: measure first, tighten to fail-closed as the baseline is raised — never a hard stop dropped without warning.

4. Scoring Model (summary)

score(dimension) = Σ weight(passing applicable checks)
                   ─────────────────────────────────── × 100
                   Σ weight(applicable checks)

weight:  critical 3 · high 2 · medium 1 · low 0.5
grade:   A ≥90 · B ≥80 · C ≥70 · D ≥60 · F <60
n/a:     no applicable checks in that dimension (expected for fragments on Machine-Index)

An overall figure per dimension is the weight-aggregate across all evaluated surfaces, so a portfolio can be tracked as a single number over time.

5. What each dimension checks (automatable set)

Accessibility (human reach) — document language; unique, descriptive title; skip-to-content link; landmark regions; zoom not disabled; image alternative text; decorative graphics hidden from assistive tech; no autofocus; no autoplaying media; captions/subtitles on video; logical heading order; a single top-level heading; no empty links; descriptive link text; table header cells; labelled form fields.

Machine-Index (agent / answer-engine reach) — indexable (not blocked); meta description; descriptive title; canonical URL; structured data (JSON-LD) with a recognized type; FAQ content carrying FAQ structured data; social/preview metadata; semantic sectioning; descriptive subheadings; scannable formats (lists / tables / disclosure); internal linking; freshness (dated content); author / authority attribution.

Manual review (informational, unscored) — colour and border/icon contrast; visible focus states; minimum target size; caption/audio-description content quality; harmful-motion review; plain-language review; live answer-engine accuracy; BLUF summary quality; E-E-A-T verification; citation share in AI answers.

6. Relationship to the platform

• Access-governance parity. The data plane governs who may reach the data; this feature governs whether the delivered surface is reachable — by people and by agents. Together they make "accessible, governed data" a testable claim end to end.
• Forward-deploy alignment. Non-destructive and additive: surfaces are scored as they are, gaps are surfaced as a worklist, and enforcement is tightened only as the baseline improves.
• AEO substrate. The Machine-Index dimension is the acceptance test for the semantic-structure work (semantic sectioning, structured data, scannable formats) that makes connected content answerable by AI engines — the machine half of accessibility.

7. Non-Goals

• Not a certification of legal WCAG conformance; it is an automated pre-flight that raises the floor and routes the remainder to manual review (FR-AXI-06).
• Not a live AI-answer-quality monitor; it measures machine readability, the precondition for good AI answers, not the answers themselves.
• Ships no customer data and requires no credentials to run against public surfaces — it reads only rendered markup.

CRM Sync is a governed Shopify ⇄ Xano ⇄ Webflow gateway. This feature makes the accessibility of every delivered surface measurable, ranked, and enforceable — the foundation under the promise that connected data is genuinely accessible to enterprise teams and to AI agents alike.