Outcome-Based Agency Pricing Research Note

Outcome-Based Agency Pricing Research Note

Outcome-based pricing ties some part of an agency's compensation to an agreed result instead of billing only for time or deliverables. It can align incentives, but only when the parties define the outcome, contribution, baseline, measurement period, data source, risks, and dispute process before work begins.

The attribution problem

Marketing results rarely have one cause. Revenue can move because of product quality, price, inventory, discounts, seasonality, competitors, distribution, prior brand investment, economic conditions, or changes made by other teams.

An agency can control research quality, creative choices, campaign operations, testing, and response time. It cannot normally control inventory, site uptime, fulfillment, the client's sales process, or the market. A contract that pays on revenue without allocating these factors creates a hidden transfer of risk rather than a clean alignment of incentives.

The United Kingdom government's risk allocation and pricing guidance recommends placing risk with the party best able to manage it. The same principle applies to agency work. The agency should not guarantee factors controlled by the client, and the client should not pay a bonus for growth that would have occurred without the intervention.

A workable structure

A practical model often combines a base fee with a variable component. The base covers committed capability, access, research, operations, and minimum delivery. The variable component rewards an incremental result above an agreed baseline.

The measurement design should state the unit of analysis, attribution method, exclusions, observation window, source system, data latency, refund treatment, and treatment of repeat customers. It should also state what happens when campaigns cannot run, tracking breaks, inventory disappears, or the client changes the offer.

Controlled experiments are stronger than before-and-after comparisons. Microsoft's research on online experimentation explains how randomized tests help isolate the effect of a change. When randomization is not practical, the parties need a weaker but explicit method and should reduce the size of the performance fee accordingly.

Avoiding metric distortion

A narrow metric can encourage behavior that damages the larger business. Optimizing click-through rate can reward sensational copy. Optimizing immediate revenue can produce excessive discounting, increase returns, or weaken long-term trust.

The commercial scorecard should pair the paid result with guardrails. For ecommerce email, those may include margin, unsubscribes, complaints, deliverability, refunds, repeat purchase, and brand review. The Green Book's evaluation guidance reinforces the need to define objectives, alternatives, costs, benefits, risks, and evaluation rather than treating one output as the entire case.

Evidence from payment-by-results programs

Public-sector payment-by-results programs show why contract design matters. The UK government's evidence review of employment and health service contracting discusses incentives, provider behavior, data, outcomes, and risks such as concentrating effort on easier cases.

The OECD review of paying for results similarly shows that outcome funding is not one model. Design choices affect behavior, measurement burden, access, and the distribution of risk.

These sources do not evaluate marketing agencies directly. They provide transferable contract and evaluation principles.

Publication boundary

Episode 103 raises outcome pricing as a response to AI-driven production efficiency. The public article should not claim that outcome pricing is always superior to hourly, project, retainer, or value-based pricing.

The strongest recommendation is conditional. Use outcome-linked fees when the result is measurable, the agency has meaningful control, the data is shared and auditable, the baseline is defensible, and the downside does not reward harmful behavior. Otherwise use a hybrid or a clearly scoped project model.