Rule: concede the credence-good problem; make marketing as inspectable as possible

Rule: Do not claim marketing work is straightforwardly inspectable. Concede the credence-good problem ([[marketing-services-as-credence-good-for-gc]]) and then engineer mechanisms that make the inspectability problem smaller, not absent.

Mechanisms Candid commits to:

• Published process — the actual sequence of work Candid performs is documented and shareable, not opaque.
• Published reporting — what gets measured, at what cadence, with what tooling.
• Named-deliverable contracts — every workstream has a specific artifact attached, not a vague retainer.
• Refusal to quote cold — Candid produces a paid audit before quoting an engagement (parallels [[pioneer-craftsmen-kitchener-design-build-reference-case]]'s paid-design model).
• Visible artifacts — websites, written case studies, deployment logs, before/after CWV runs — the durable footprint Crawford ([[crawford-real-work-vs-marketing-probabilistic-deterministic]]) trained the GC to expect.

Why: Vendors who say "trust us, the work is real" are arguing against information the buyer already has (FTC HomeAdvisor record — [[ftc-homeadvisor-angi-jan-2023-7-2m-order]]; Vermont AG settlement — [[vermont-ag-angi-100k-settlement-oct-2025]]). Conceding the problem flips the credibility dynamic.

How to apply: Every Candid proposal includes a "how to inspect this engagement" section. Every long-form Candid content piece references the inspectability discipline at least once.