Carman Creative
back to work

// S02 — Product / Creative Operations

CQAP

Review once. Decide once. Reuse the decision.

ProjectSelf-initiated prototype
RoleConcept, product design, build
Year2026
StatusAlpha — working prototype, not a shipped product
StackNext.js, deployed on Vercel
Scroll / Case study
CQAP — cover

// The brief

Creative review at enterprise scale is where quality is won or lost, and it is almost always run out of inboxes and memory. CQAP is a prototype for the alternative: a structured review that captures the reasoning, not just the verdict.

// The challenge

In a large organisation, the same creative question gets answered again and again. Someone decides how a template should version, or which identity a piece of collateral carries — and that decision lives in an email thread until the person who made it moves on. Reviewers spend their time re-deciding settled questions instead of catching the things that actually matter.

Built on my own time as a prototype, out of years of running creative review at volume. The screens show a real review pass on a real piece — a Carelon Palliative Care welcome guide — from package validation through to an approved decision.

// The work

cqap.vercel.app
CQAP — the workspace: one review active, one completed, and the standards learned so far
The workspace with work in it: one review in progress, one just approved and already available as a prior decision, and the standards the system has learned along the way.
CQAP — validation and context: the package is ready to review
Validation before judgment. File integrity is checked and the context — brand, line of business, channel, piece type — is confirmed before anyone forms an opinion about the work.
CQAP — the review workspace with two reviewer findings accepted
The review workspace. Findings are placed on the proof as numbered markers and written up on the right, each tied to a supporting standard — here, a Carelon Palliative Care welcome guide, two minor findings, both accepted.
CQAP — the review workspace before any findings
When the evidence isn't strong enough, CQAP stays quiet — the panel says so in its own words rather than manufacturing a finding.
CQAP — review complete: approved, saved as a review memory
A decision becomes a record. The review is saved as a memory, its accepted findings keep their evidence, and the workspace updates.

// The mandate

Prototype a review tool where the reasoning survives the review: classify work on the way in, capture the decision on the way out, and make both searchable.

// John's role

  • ↳Framed the problem as institutional memory rather than workflow speed
  • ↳Designed the review model — classify on intake, capture the decision on completion
  • ↳Designed the workspace, the searchable intelligence layer and the memory view
  • ↳Built the prototype in Next.js
  • ↳Wrote the product language, including the measurement principle below

The people behind the work

  • ↳Self-initiated — concept, design and build by John Carman

// Key decisions

01

Memory is the product

The obvious version of this tool is a faster approval queue. The useful version is one where a completed review becomes a reusable record — 'Decisions that can be reused' is the actual heading on the memory view, because that is the thing worth building.

02

Classify on the way in

Work is identified by brand, line of business, channel and piece type before review begins. Most review disagreement is really ambiguity about which standard applies, and this settles it first.

03

Measure the system, never the reviewer

The Impact view states it in the interface: these metrics describe organisational activity, not individual reviewer productivity. A governance tool that becomes a surveillance tool stops getting honest input, and that has to be a stated principle rather than a good intention.

04

Approved work becomes reference

Strong examples are promoted into the intelligence layer rather than just being cleared. Governance that only ever says no teaches nothing; showing the best prior answer is how standards actually spread.

// The outcome

  • ↳A working Alpha with the full loop in place — package validation, a review workspace with placed findings, a decision that becomes a reusable record
  • ↳A stated measurement principle that keeps the tool pointed at the system rather than at people
  • ↳TODO: Confirm — no pilot, adoption or outcome data yet; this is a prototype and is presented as one

// Reflection

Building it clarified what I had been doing manually for years. The hard part of creative review was never the reviewing — it was that the reasoning evaporated the moment the job closed, so the same ground got re-covered every quarter. Once memory became the product rather than a feature, most of the other design questions answered themselves.

previousPremier Friends Club
All projects
nextCarman OS