CRO pricing models compared: retainer, project, performance
Three ways to buy conversion work: a monthly retainer, a fixed-scope project, or performance-based pricing. Retainers suit ongoing programmes and are the default for a reason. Projects suit a defined audit or a single build. Performance pricing sounds like the safest option for the buyer and is usually the worst deal available.
Why performance pricing usually fails
The pitch is irresistible. Pay only on results, and the agency shares your risk.
The problem is attribution. Conversion moves for a dozen reasons: seasonality, traffic mix, a competitor’s promotion, a product launch, a pricing change you made yourself. Isolating the agency’s contribution requires exactly the kind of controlled testing that a performance deal gives the agency an incentive to avoid, because a clean holdout group reduces the measured uplift they get paid on.
It also distorts the roadmap. An agency paid on measured uplift will prioritise tests that look good quickly. That means discount tests, urgency mechanics and scarcity messaging. In DRIP’s experiment data, scarcity tests were decisive 84.2% of the time, which is precisely why they are attractive to someone being paid on a short-window number, and precisely why they can damage a brand over a longer one.
And it fights the base rate. With average win rates near 12% across 127,000 experiments, an agency paid only on winners has to either take enormous risk or steer toward safe, small, quick results.
The three models side by side
| Retainer | Project | Performance | |
|---|---|---|---|
| Best for | Ongoing testing programmes | A defined audit or single build | Almost nothing, honestly |
| Cost predictability | High | High | Low and volatile |
| Incentive alignment | Process quality | Delivery of a deliverable | Short-term measured uplift |
| Roadmap control | Shared | Yours | Effectively theirs |
| Risk to buyer | Paying through a slow quarter | Scope ending before results land | Attribution disputes, brand-damaging tests |
| Typical commitment | 3 to 6 months minimum | 4 to 8 weeks | 6 to 12 months |
Read any pricing proposal for what it makes the agency want to do. That is a more reliable predictor of the next six months than the fee itself, because incentives survive contact with reality and good intentions frequently do not.
Where a project genuinely fits
A one-off audit is a legitimate purchase. You want a diagnosis, a ranked list of problems and five test briefs, and you intend to execute internally or decide afterwards whether to commit.
The risk is timing. Test results take weeks. The median test runs 42 days, and benchmarks collected by roast.page put the median nearer 23 days. A six-week project that includes “run tests” will end before the first one reads out, which means you paid for a launch rather than a result. If the project includes testing, the scope has to run long enough to read them or the deliverable is a hypothesis, not an answer.
What a retainer should actually cost you in commitment
Three months is the minimum honest engagement and six is more realistic, for a reason that is arithmetic rather than sales.
Month one is audit, research and roadmap, with the first test live inside two weeks. Month two produces the first readable results, most of which will be inconclusive. Month three is where compounding starts. A three-month engagement judged at the end of month three is being judged on one or two clean results, which is a small sample of a low-hit-rate activity.
Ask for a defined review point at month three rather than a shorter contract. It gives you the same protection without cutting the programme off before it can produce anything.
One clause worth adding to any of the three
Whatever model you choose, specify what happens to in-flight work at notice. Do running tests complete, and who reads them?
It sounds like housekeeping and it prevents a genuinely bad final month, where three tests are abandoned halfway and nobody learns anything from a quarter of spend. It also removes the incentive to stop launching anything in the closing weeks of a contract, which is otherwise the rational thing for an agency to do.
The question that separates the models
Are you buying a diagnosis or a capability?
A diagnosis is a project. A capability is a retainer. Performance pricing is neither, because it is buying an outcome you cannot cleanly measure from a party with an incentive to measure it generously.
You can see how Parah Group structures the retainer, including the month-three review point and what sits outside the scope.
