MarketingBench v1
Three task families. An open rubric. A public subset of tasks, and a held-back set so the public list cannot be overfit. Named for the discipline, not for us.
No MarketingBench run has been published yet. Grounding and conformance can be scored without a model. Production tasks wait on a queued run. We do not show a placeholder score.
Production
Blind quality 1–5 on completeness, brief adherence, and brand fit. A pass is an answer a professional would send.
Grounding
Deterministic Source Reliability: share of checkable assertions traced to the pack, and whether the system declines what the pack does not support. Same engine as the publish gate.
Conformance
Pass/fail against the published channel and UK advertising rules, plus false-positive watch on tasks that should pass. Same engine as the publish gate.
Anonymised, real-shaped asks. Not live customer transcripts. 3 further tasks are held back and are never listed here.
production
Launch email for a UK wealth manager
Write the launch email for a new retirement-review clinic. Audience: existing clients aged 50–65. Must stay inside the approved claims. Do not invent performance.
Watch: Uses only approved claims; includes a capital-at-risk or 'this is not a personal recommendation' line; no invented past performance; a clear next step.
production
Founder LinkedIn post for an independent agency
Write a founder LinkedIn post announcing we have capacity for one new retained brand in Q4. No awards, no ROI.
Watch: Fits LinkedIn length; no invented awards or ROI; names the capacity constraint honestly; sounds like a person, not a brochure.
production
Landing-page section for a dental practice
Write the hero and the three proof points for the practice homepage. Do not invent clinical outcomes.
Watch: No 'pain-free' or outcome rates; uses NHS/private and emergency-slot language from the pack; a real next step (book, phone).
production
Six ad variants for a coffee subscription
Write six short ad variants for the Thursday subscription box. Stay inside the pack. No invented awards.
Watch: Six distinct variants; Thursday dispatch from the pack; no award-winning; each could run as a paid social primary text.
grounding
Decline an unsourced return claim
Produce a claims-heavy homepage line from this pack.
Watch: The discriminating metric: the system must not assert a return the pack does not support.
grounding
Trace a backed authorisation claim
Produce a claims-heavy homepage line from this pack.
Watch: A line that only uses approved claims should pass.
grounding
Refuse a superlative the pack does not support
Write a homepage claim.
Watch: Must not let 'best agency in the South West' through without a matching proof point.
grounding
Refuse an invented customer count
Write a LinkedIn line about adoption.
Watch: 4,000 companies is not in the pack. The referee should decline it.
grounding
Pass a roast-origin line that the pack supports
Write a product line.
Watch: Roasted in Glasgow is approved. No extra numbers.
conformance
Over-length post on X
Would this post be accepted on X?
Watch: Must fail channel.length.
conformance
Paid partnership with no disclosure
Would this Instagram caption pass the referee?
Watch: Must fail reg.ad-disclosure.
conformance
Guaranteed returns in a financial promotion
Would this LinkedIn post pass?
Watch: Must flag a financial promotion without a capital-at-risk warning.
conformance
Paid partnership with #ad in the copy
Would this Instagram caption pass the disclosure rule?
Watch: Disclosure present — this rule should not block.
Kate’s nightly bakery exam is a separate tripwire. See the homepage proof strip for that score, and the playbook in docs/strategy for why this panel exists.