Which metrics should decide what an agency gets paid once clients stop buying hours? Not ROAS and CPA on their own. The metrics that hold up in a fee negotiation are the ones a client's finance director already reports to the board: incremental revenue, contribution margin, new customers and retention. Platform metrics become the diagnostics underneath.
The question is no longer theoretical. Campaign reported a WFA and Agency Mania Solutions study of 69 multinationals, with a combined $147bn in marketing spend, showing that labour-based remuneration has fallen from 54% of models in 2011 to 17% in 2026. Fifty-eight per cent of those brands expect to increase performance-based pay.
Why is time-based agency pay disappearing?
Because AI broke the link between hours and output. When a first draft takes minutes, billing for the time it used to take is a hard sell. Laura Forcetti of the WFA described AI as "making time an increasingly weak proxy for value". According to the study, 20% of brands are already changing remuneration models in response to AI, and 61% intend to.
The money is moving in two directions. Fixed-fee and output models rose from 20% to 35% of arrangements. Hybrid labour-and-performance models more than doubled, from 9% to 23%. "The future of agency compensation is not paying for effort," Forcetti said.
For agencies, that is both threat and opportunity. Effort was always a poor way to price an idea. But outcome pay only works if both sides agree on what an outcome is.
Why aren't ROAS and CPA enough to set agency fees on?
ROAS and CPA are the metrics agencies know best, which is exactly why they make dangerous fee triggers.
- They are usually platform-reported. The platform selling the media is also marking its own homework.
- They ignore incrementality. A campaign can post a superb ROAS by retargeting people who were going to buy anyway.
- They reward the bottom of the funnel. Pay an agency on CPA and it will rationally shift spend towards cheap conversions and away from building future demand.
- They say nothing about profit. A heavily discounted sale and a full-price sale look identical in ROAS.
None of that makes them useless. It makes them diagnostics. You want them on the dashboard, not in the contract.
Which metrics should outcome-based agency pay be built on?
The test is simple: would the client's CFO recognise the number without a glossary? Here is how the usual candidates stack up.
| Metric | What it tells the client | Agency control | Best use in a contract |
|---|---|---|---|
| ROAS / CPA (platform-reported) | Efficiency inside one channel | High | Diagnostic only |
| Incremental revenue (from MMM or holdout tests) | Sales that would not have happened without the work | Medium | Primary outcome metric |
| Contribution margin | Whether growth is profitable | Low to medium | Guardrail alongside revenue |
| New-customer growth | Whether the brand is expanding or just recycling buyers | Medium | Secondary outcome metric |
| Retention or customer lifetime value | Long-term value of the customers acquired | Low to medium | Longer-cycle bonus |
| Leading indicators (creative pre-test scores, test cadence, brief turnaround) | Quality of the agency's own inputs | High | Small, fast-cycle component |
Notice the pattern. The metrics clients value most are the ones agencies control least. That is the central design problem, and it is why pure outcome pay rarely survives contact with reality.
Why does the shift to performance pay keep stalling?
Everyone agrees on the direction. Few are moving fast. Adweek reported Bob Lord, president of Horizon Media Holdings, calling FTE billing the "old system" and questioning whether it "can actually be broken without some kind of external force".
The holding companies are no quicker. Digiday reported WPP chief executive Cindy Rose saying outcome-based pay will take "a few years" to spread across the client base. At the time, only one WPP client, Jaguar Land Rover, had adopted it, while WPP's first-half revenue less pass-through costs fell 4.7% to about £5bn.
The reasons are structural. Outcome pay moves risk onto the agency for things it does not control: pricing, stock, distribution, the client's website. Procurement teams want comparable bids, and outcomes are hard to compare. And if nobody agreed the measurement method in advance, every bonus conversation turns into an argument about attribution.
Why value is what the client feels it is
Value perception theory describes something agency people learn the hard way. Value is not what you deliver; it is what the buyer believes they received, set against what they gave up. Put more bluntly: value is not what you say it is; it's what they feel it is.
Under hourly billing, the timesheet did the persuading. It was visible proof of effort. Under outcome pay, that prop disappears, and the agency has to make value visible in the client's own language. A ROAS chart is the agency's story. Incremental revenue in the CFO's board pack is the client's feeling.
The WFA study adds a twist. Brands rated briefing, mutual respect and trust, and high-quality feedback as the most important drivers of good agency performance, and saw financial incentives as a weaker driver. The contract does not create perceived value on its own. The relationship does. As Bruno Gralpois of Agency Mania Solutions put it: "Becoming better partners isn't a mindset shift, it's an operational one."
How do you move a client to outcome-based pay without betting the agency?
- Protect the base. Keep a fixed fee that covers your cost of delivery. Put only a capped share, say a fifth, at risk in the first year.
- Pick one north-star metric the CFO already reports. Usually incremental revenue or new customers, with contribution margin as a guardrail.
- Agree the measurement method before the work starts. Marketing mix modelling, geo holdouts or matched-market tests, plus the baseline you are measuring against.
- Add two or three leading indicators you control. Creative tested before launch, test cadence, brief turnaround. They pay out faster and keep the team motivated.
- Write down what sits outside your control. Price changes, stock-outs and site outages should adjust the target, not your fee.
- Review quarterly, settle annually. Short enough to course-correct, long enough for real outcomes to show.
Evidence is what makes all six steps work. Defy's predictions are validated against real ad performance, creative is scored before media spend and every AI conclusion can be checked against the posts behind it, so the outcome conversation starts from evidence rather than opinion. You can see how those pieces fit together in the Defy platform.
