Imagine two production partners deliver the same approved product film. Both versions are accurate, licensed and ready to run. One supplier needs eight working days. The other uses AI-assisted pre-production, automation and a better asset system to finish in three. If procurement rewards only headcount and time, the more capable supplier is forced to charge less precisely because it performed better.
That is not simply a negotiation problem. It means the unit of purchase is wrong. A brand is not buying visible busyness; it is buying usable marketing assets, controlled risk and the ability to learn from the work.
The market signal: compensation is moving away from labour
On 29 September 2026, the World Federation of Advertisers published research based on 69 multinational companies representing $147 billion in global marketing spend. The share using labour-based models as their standard fell from 54% in 2011 to 19% in 2026. Fixed-fee or output models rose from 20% to 33%. Production was particularly suited to this change: 61% used fixed-fee or output-based compensation as the standard model.
The study does not prove that every market or every brand has changed. It does reveal a structural pressure. When a deliverable can be defined and accepted, hours are becoming a weaker proxy for value. Only 20% of respondents had already changed commercial models in response to AI, while 61% intended to do so. Production technology is moving faster than production contracts.
The IPA Pricing Playbook, published on 12 March 2026, offers a useful distinction between input-, output-, outcome- and hybrid pricing. Its more important point is that no single model wins everywhere. The right choice depends on risk appetite, scope stability, operational flexibility, commercial fluency and whether outcomes can be measured with confidence.
“Pay for outputs” cannot mean “count the files”
A 30-second master exported in 16:9, 1:1 and 9:16 is not automatically three equally valuable videos. File-count pricing can reward meaningless fragmentation: more crops, more exports and more versions, even when none creates a new answer for the customer.
ONCE proposes a stricter unit: the Qualified Output Unit (QOU). A file counts as a QOU only after it clears five gates:
- Defined job: the audience, channel and stage of the customer journey are explicit;
- Approved truth: product claims, demonstrations, copy and subtitles have brand approval;
- Usable specification: aspect ratio, duration, encoding, language and platform requirements are complete;
- Clear rights: talent, music, footage, territory and usage period are documented;
- Final acceptance: reviewers, revision limits, approved version and source-file scope are recorded.
This separates “we generated a file” from “we delivered an asset that can do its job.” It also allows origination, substantial re-editing, language adaptation and simple transcoding to carry different complexity weights instead of pretending that every export is equivalent.
A video quote buyers can actually compare has four layers
A practical commercial model is:
Fair production price = base operating layer + Σ(QOU × complexity weight) + rights and risk + optional learning or outcome incentive
1. Base operating layer
This covers brand onboarding, clarification, creative and production planning, producer coordination, review systems and archive governance. It should be visible rather than hidden across dozens of deliverables. A structured executable creative brief is part of that operating infrastructure: it reduces rework before cameras, editors or models start producing.
2. Qualified output layer
Separate original concepts and hero masters from recuts, language adaptations, aspect-ratio adaptations and technical exports. Give each QOU type an acceptance test and a complexity weight. The supplier can then retain some of the value created by better tools, while the buyer gains predictable unit economics and does not pay extra for avoidable inefficiency.
3. Rights and risk layer
Talent and music licences, media scope, territories, usage windows, product shipping, reshoot responsibility and regulatory review should be explicit. A low “price per video” often excludes these obligations; a higher one may already carry them. Until those boundaries match, the two quotes are not comparable.
4. Learning and incentive layer
If a production partner also designs variant tests, reads performance data and turns evidence into the next brief, charge for that learning loop. Add a limited outcome bonus only when attribution is credible, data is shared and the supplier controls enough of the variables that drive the result.
Why pure performance pricing is usually a poor fit for production
Video can influence attention, comprehension and conversion, but sales also depend on targeting, media spend, offer, price, inventory, reviews, landing-page quality and fulfilment. A production company rarely controls all of them. Tying the entire fee to ROAS or revenue transfers unrelated risk to the supplier and can encourage short-term click tactics.
The WFA findings reflect that caution. Across most disciplines, 67% to 80% of respondents said performance-based remuneration represented less than 20% of total agency compensation. Moving closer to outcomes does not require eliminating the base fee. A healthier hybrid keeps the operating system available, pays for accepted QOUs, exposes rights and risk, and reserves bonuses for the part of performance both sides can credibly influence.
Choose the model by the shape of the work
| Buying situation | Better-fit model | Protection to include |
|---|---|---|
| Exploration with an unstable scope | Time or stage fee with a budget cap | Decision gates and a clear exit from discovery |
| One-off project with defined deliverables | Fixed project fee or QOU pricing | Acceptance, revisions and rights agreed first |
| Predictable monthly content volume | Base subscription plus QOU credits | Rollover, complexity and rush rules |
| Measurable outcome under shared control | Base fee plus a modest outcome bonus | Data source, baseline, window and anomaly rules |
How to compare a Shenzhen quote with a US or European quote
Do not begin with the day rate or the headline price per video. Give every bidder the same QOU sheet. Identify original masters versus adaptations, intended channels, languages and aspect ratios, revision rounds, usage periods and the source assets that must enter the archive. Then compare the four layers separately.
This matters for overseas brands whose products are made in China. A Shenzhen production team should not be valued only as a cheaper pool of labour. Its commercial advantage can come from direct access to samples, engineers and packaging variants; efficient batch capture; faster reshoots; and delivery of structured agent-ready footage that remains useful after the first edit.
Consider two proposals. A remote Western team may include local talent, high-cost studio time and international product logistics. A Shenzhen hybrid team may reduce physical handoffs and capture more product evidence near the supply chain. Neither is automatically better. The correct comparison is which proposal produces more accepted QOUs, with the required rights and fewer operational gaps, for the total budget.
The procurement change to make now
Before asking another supplier to cut its day rate, define what counts as one qualified output. Require every proposal to disclose the same operating, output, rights and learning layers. That single change makes hidden exclusions visible and stops the contract from rewarding slowness.
AI will not make professional production free. It will move value from time spent to qualified work completed. Brands that update the buying unit can share efficiency gains without pushing every new tool into a race to the lowest price. Production partners that can prove the quality, rights and reusability of each output can scale without pretending that hours are the product.
Source note: industry figures are drawn from the WFA report published 29 September 2026 and the IPA Pricing Playbook announcement published 12 March 2026. QOU and the four-layer quote are ONCE analytical frameworks, not industry standards or performance guarantees.
