THE SUPPLY DROP
PRODUCTION READINESS SCORECARD
Twenty-five questions that separate a pipeline from a habit
- Format
- Scored assessment, PDF and on this page
- Length
- 25 questions
- Gate
- None
- Licence
- Use it, commercially
Twenty-five questions across five areas, scored zero to five, that tell you whether a team or a studio is actually ready to run generative production at volume, or is running demos and hoping.
Use it on a supplier you are evaluating, or on yourself before you commit to a campaign you cannot deliver.
DOWNLOAD THE PDFNo email, no form. The whole thing is on this page too.
HOW TO USE IT
- 01Score each row 0 to 5. Zero means it does not exist. Five means it is written down, in use, and somebody owns it.
- 02A row nobody can answer scores zero. "We sort of do that" scores one.
- 03Total out of 125 and read the band table at the end.
A. Locks and references
- Master plates existA frozen, versioned reference frame per product that must not change
- Plates verified against a source of truthPrint artwork or the physical object, not the client’s website
- Trained identities rather than re-uploaded referencesFor any face appearing across more than a handful of assets
- Set specification written before generationKey direction, temperature, surface, lens language, prop rule
- Version tags and dates on every reference artefactSo a packaging change cuts a new plate rather than editing the old one
B. Control gates
- Type gateEvery printed word read at full resolution against the artwork
- Silhouette gateOverlay on the plate at 40% opacity, with a stated pass criterion
- Light gateOne key direction and temperature held across the set
- Claims gateNo invented certification, award, ingredient or percentage survives
- Failures return to sourceA failed frame is regenerated, never retouched into passing
C. Measurement
- Cost per accepted asset is measuredNot cost per generation, and not per campaign average
- The ledger is written automaticallyBy the pipeline, not by a person at eleven at night
- Acceptance rates known per shot typeA fraction, not a reassurance
- Credit ceiling per asset encodedAnd it halts the run rather than logging a warning
- Operator and review time countedIncluding salaried time, which is capacity
D. Compliance
- Model licences cleared before renderingFor the specific model and version, not in general
- Consent files complete before the first renderIncluding derivative training, scope, territory and term
- Disclosure decided at brief stageSet to the stricter of platform policy, UK position and EU Article 50
- On-asset labelling plus provenance metadataNot metadata alone
- Record kept per assetModel, version, sign-off, disclosure, placement
E. Handover and continuity
- Client owns the plates and trained identitiesSo the second campaign starts from a lock
- Naming convention in forceAugust’s output is findable in November
- The pipeline is documented, not rememberedIt survives the operator leaving
- Systems run in the client’s accountsA system you cannot operate without the supplier is a dependency
- A named person can stop a runWith the authority to reject, at a defined point
Reading the score
| SCORE | VERDICT | WHAT TO DO |
|---|---|---|
| 105 to 125 | Production-ready | Commit to volume. Spot-check the measurement rows quarterly. |
| 80 to 104 | Works, with known gaps | Run the campaign. Fix the lowest-scoring area before the next one. |
| 50 to 79 | Demo-grade | One campaign is survivable. Volume is not. Build locks and gates first. |
| Below 50 | Not ready | Do not commit to a campaign that depends on consistency. Start with section A. |
LICENCE
Free to use, copy, adapt and put your own name on. No attribution required, though a link back is appreciated. Published by TaleCrafters, talecrafters.studio.
THE WRITING BEHIND IT
- How to Choose an AI Video Model in 2026 (Without Rewriting Your Pipeline Next Quarter)Everyone asks which model is best. The useful question is which one your pipeline can survive being wrong about.
- Switching AI Video Models Without Losing Your LookA model change is a re-grade, not a re-render. Treat it as the second and you will pay for the campaign twice.
- How to Choose an AI Video Production Agency: Twelve Questions and the Answers That DisqualifyEverybody has a showreel. Ask what got cut from it and how many renders it took.
- How We Keep a Product Consistent Across 100 AI-Generated ShotsShot one is easy. Shot one hundred is the job. This is the method, including the numbers we plan against.
- Cost Per Accepted Asset: Measuring Generative Production ProperlyThe metric that makes generative production explainable to the person who signs the invoice, and the arithmetic behind it.
- How to Brief an AI Video Production StudioMost generative briefs fail on the same missing paragraph. Here is what to put in one, and what a studio should ask you back.
- What Creative Automation Actually Removes From a Marketing WeekThe hours automation removes are not the ones anybody expects. They are also not the ones any vendor demo shows you.
YOUR MOVE
Using this on a live job?
If the checklist turns up something you would rather not run yourself, TaleCrafters does this for a living. Bring the brief and the deadline.