Outcome-based pricing is viable when the payable result is precisely defined, the supplier can materially influence it, the required evidence is obtainable, and the approved business case survives the payment mechanism’s downside. The deliverable is an outcome-payment decision record containing the eligible population, baseline, adjustment rule, quality conditions, data responsibilities, payment formula, scenarios and authority to proceed. A measurable indicator alone is insufficient: two reviewers must be able to derive the same amount from the agreed evidence, and the supplier must be able to fund performance before that amount is received.
An invitation offers a modest service fee and a large bonus for reducing repeat enquiries. The bid team values every fall in enquiries as its own achievement and assumes the bonus will cover delivery cost. It has not checked whether the buyer is changing its policy at the same time, whether contacts through another channel are counted, or whether the payment is capped. A plausible benefit story becomes an unsupported revenue forecast.
Decide whether to accept this payment mechanism before promising the result. This task differs from writing a benefits claim: it determines what earns money, what can reduce that money and who bears measurement failure. All example organisations, figures, thresholds and commercial terms are fictional. Sources were checked on 5 September 2026; the NAO’s 2015 findings are identified as historical evidence, not current market statistics. The guidance supports analysis, not legal advice or authority to alter mandatory tender terms.
Commercial boundary
A completed service and a paid outcome are different events
Training sessions delivered, participants assessed and people remaining in employment are different measures. A supplier may control the sessions closely, influence learning and have much less control over later hiring. Similarly, resolving an enquiry correctly is different from reducing total demand on the buyer. Identify the exact result that changes consideration. Calling the whole contract performance-based does not tell you which uncertainty the supplier is being asked to finance.
In the applicable US federal services framework, FAR 37.601 calls for measurable standards and an assessment method, with incentives linked to those standards where used. It does not mean every performance-based service contract pays only for final outcomes. Read the actual solicitation and applicable agency rules. An availability target backed by service credits, an accepted-deliverable payment and a results bonus should not be modelled as interchangeable mechanisms.
The Cabinet Office’s June 2026 risk-allocation guidance connects pricing to the risks a party can manage and warns about transferring outcomes while prescribing the supplier’s inputs. Use that question on the actual bid: can the proposed team change the process, timing or intervention that drives the result? If the buyer fixes all those choices, a demand to guarantee the resulting improvement needs a specific treatment rather than a larger unexplained contingency.
Write the commercial boundary in one paragraph: required service, outcome-linked amount, measurement period, buyer responsibilities and unresolved conditions. Identify whether the issued model is mandatory or allows alternatives. A blended fee can be worth examining, but the bidder cannot replace a pure results model unilaterally. Any proposed change must follow the permitted procedure and reach the price schedule and contract response consistently.
| Payment basis | What needs proof | Question before pricing |
|---|---|---|
| Activity | Specified work was performed | Are the allowable inputs and limits clear? |
| Output | A defined deliverable meets requirements | Can acceptance be evidenced independently? |
| Service standard | Performance over the agreed population and clock | How do deductions interact with other remedies? |
| Outcome | The agreed customer result and settlement conditions | Who bears external influence and evidence uncertainty? |
Measurement protocol
Specify the population before quoting a price per result
For repeat enquiries, decide what counts as an initial eligible case, what makes a later contact a repeat and how contacts across phone, email and the service portal are linked. A reopened record must not vanish because it receives a new identifier. Specify the observation window from the initial resolution, the handling of cases opened near period end and the evidence that the first resolution was correct. Fewer recorded contacts can also mean that users could not reach the service.
Fix the baseline cohort and period before interpreting the change. Check demand mix, complexity, seasonality, opening hours, channel coverage and policy conditions. Store the counts behind a rate. If one period contains 20,000 cases and another contains a different population, a count comparison alone is insufficient. The measurement agreement must state how changes in mix or volume are treated and who has authority to approve an adjustment.
The baseline is not automatically the counterfactual, meaning what would have happened without the supplier’s intervention. HM Treasury’s QPIE guidance distinguishes observing a change from attributing it and explains why a simple before-and-after design can be weak. A comparison group also needs scrutiny: different trends, selection or exposure to the same intervention can invalidate the intended comparison. Specialist evaluation judgment cannot be replaced by naming a group “control”.
A contract may nevertheless choose a transparent proxy for settlement. Label it as the agreed payment calculation, not proof that each calculated unit was caused by the supplier. Record numerator, denominator, adjustment, rounding, floor and cap in their order of application. Have both parties calculate sample cases before signing. Include a boundary case just below a threshold and a case with incomplete evidence, not only the example that earns the maximum reward.
| Element | Definition required | Failure to resolve |
|---|---|---|
| Eligible population | Entry rule, exclusions and cross-channel identifiers | Do not quote against a movable denominator |
| Outcome | Observable event, durability window and quality condition | Do not equate administrative closure with success |
| Reference | Baseline and any agreed adjustment or comparison | Keep descriptive change separate from attribution |
| Evidence | Source, access, extraction, verification and cut-off | Price no unsupported assumption of complete data |
| Settlement | Formula, rounding, cap, dispute and correction route | Do not approve a calculation reviewers cannot reproduce |
Dependencies
Buyer-held data must be available for the payment it controls
List the buyer-controlled prerequisites: source records, referral decisions, policy updates, access to users, consent or other applicable lawful processing arrangements, and staff actions needed to implement the service. Record the responsible party, promised timing and evidence of performance for each. “The buyer will cooperate” does not establish the dataset needed to invoice. Ask for a representative, lawfully shareable sample or a documented rehearsal where direct access is inappropriate.
Test the evidence journey before committing revenue. Can the source owner identify the full eligible population? Can a verifier join repeat contacts without exposing unnecessary personal data? Are timestamps and correction histories available? Who resolves a mismatch between the supplier log and the buyer’s system? Cost extraction, verification, dispute handling and retention obligations explicitly. Access rights and confidentiality must be agreed; a measurement need is not permission to collect unrestricted customer data.
Missing data needs a pre-agreed state distinct from a demonstrated failure. If 1,000 cases lack a completed observation window, they cannot quietly disappear from the population or automatically count as successful. The protocol should determine whether settlement waits, a supported estimate is allowed, or a defined contractual treatment applies. Show the cash and revenue consequence of that rule. Do not invent a favourable default after seeing which treatment increases payment.
Changes in buyer policy or service mix also need a stated route. A new rule may reduce repeat enquiries across every provider, while a change in eligibility can make the remaining cases harder. Specify the evidence and authority for a prospective adjustment, with any retrospective correction permitted by the contract handled explicitly. An analytical recommendation to rebase does not itself change a public contract or grant a right to additional payment.
Worked case
An apparent 800-unit improvement becomes 400 payable units
Fictional Alderpoint considers a three-month service package with a £180,000 base fee for the required service, plus £100 per verified settlement unit, capped at £60,000 of variable payment. Base payment remains subject to delivery and the contract; it is not unconditional income. Expected delivery and verification cost is £210,000, including the case’s allocated overhead but excluding financing and tax. Management requires a 10% contribution on revenue before those separately reviewed exclusions. The case assumes cost is unchanged across the illustrated result levels within the same 20,000-case cohort.
The measured outcome is fewer eligible cases with a repeat contact within 14 days after resolution, subject to a separate correct-resolution and access check. The example assumes complete records, a fully matured observation window and passed quality conditions. The supplier cohort’s repeat rate moves from 12% to 8%. An agreed comparison cohort moves from 12% to 10%. The fictional settlement rule subtracts that two-percentage-point comparison improvement from the supplier’s four-point improvement. This is illustrative contract arithmetic, not evidence that the comparison provides a valid causal estimate.
Under that rule, the adjusted reference is 10%, and the payable reduction is 10% minus 8%, or two percentage points. Applied to 20,000 eligible cases, it produces 400 settlement units. They are aggregate calculation units, not 400 individually identifiable contacts proven to have been prevented. The formula floors negative improvement at zero, multiplies by the eligible population, rounds down to whole units once, multiplies by £100 and finally applies the £60,000 cap. No intermediate percentage rounding is allowed in the example.
The resulting bonus is £40,000 and total revenue is £220,000. Contribution is £10,000, or 4.55%, below the assumed 10% floor. Counting the raw four-point reduction would yield 800 units, but the cap limits that bonus to £60,000; total revenue would be £240,000 and contribution 12.5%. The choice of settlement definition reverses the apparent approval. Neither calculation justifies claiming that the supplier caused the entire observed change.
| Case | Payable units | Variable payment | Revenue | Contribution |
|---|---|---|---|---|
| No verified improvement | 0 | 0 | 180,000 | -30,000 (-16.67%) |
| Agreed adjusted result | 400 | 40,000 | 220,000 | 10,000 (4.55%) |
| Minimum whole units for 10% floor | 534 | 53,400 | 233,400 | 23,400 (10.03%) |
| Cap reached | 600 or more | 60,000 | 240,000 | 30,000 (12.5%) |
Downside and funding
Calculate how much verified performance the price needs
To reach a contribution fraction m with cost C, revenue must be at least C divided by (1 minus m). For Alderpoint, £210,000 divided by 0.9 is £233,333.33 recurring. Above the £180,000 base fee, the required variable amount is £53,333.33 recurring. At £100 per whole settlement unit, at least 534 units are needed. At 533 units, £23,300 contribution divided by £233,300 revenue is about 9.987%, still below 10%. Rounding the displayed percentage to 10.0% must not authorize the failing case.
The cap permits at most £240,000 of revenue, so the assumed floor is mathematically reachable at the stated cost. That establishes only possibility. It does not show that 534 adjusted units will be achieved or accepted. If cost rises, redo the cap test; do not increase the forecast bonus beyond the contractual maximum. Under the illustration, cost above £216,000 would make a 10% contribution impossible even at the cap.
Keep the zero-reward case visible. Full delivery at the stated cost with no verified improvement loses £30,000 before financing and tax. A failure to meet base service requirements could be worse, because base fees or other remedies may also be affected. Read deductions, repayment rights and liability provisions together. The £60,000 reward cap limits upside in this example; it says nothing about a universal ceiling on liability or service credits.
Outcome maturation can delay even a correctly earned amount. Cases resolved on the last service day still require their 14-day observation window, followed by verification, invoicing and the applicable payment process. Test that dated cash schedule separately, including disputed evidence and a delayed base payment. Do not count a results bonus as working capital before it is collectible. The milestone-payment funding review provides the separate calculation needed to confirm that delivery can continue.
Service protections
The paid indicator must not reward denying the service
Alderpoint could appear to reduce repeats by closing difficult requests early, hiding a contact channel or directing users to another department. None proves better resolution. Review the metric alongside access, correctness, complaints, reopens and the experience of harder cases. These protections must be measurable and part of the agreed treatment, not a promise to exercise good judgment after the bonus has been calculated.
The NAO’s 2015 review of payment by results warned that badly designed incentives can favour people who are easier to help. DWP’s later evidence review also discusses selection and neglect of harder groups, while reporting mixed evidence on broader benefits. Those findings do not establish that a particular bidder will manipulate performance. They justify testing whether the proposed formula makes undesirable behaviour financially attractive.
Use an inclusion reconciliation: eligible intake, service delivered, unresolved cases, observations still maturing, excluded records with reasons, and records used for settlement. Compare segments rather than only the overall average. A better headline rate caused by serving fewer complex cases should trigger investigation. If the contract uses different payment weights for different groups, verify group assignment and preserve the service obligation to every eligible group.
Test corrections before they happen. Suppose an apparently successful case is later found incorrectly resolved within the agreed observation window. State how it changes the numerator, quality condition and settlement. For findings outside that window, apply the actual correction and remedy provisions rather than inventing a perpetual clawback. A reviewer needs both the measurement cut-off and the continuing service duties to understand what has finally been settled.
Decision record
Approve the measurement bargain with the commercial offer
Assemble the outcome-payment decision record from the issued schedules, measurement protocol, sample calculation, cost basis, funding review and unresolved dependencies. State the mechanism as offered, not just the preferred alternative. Show the maximum revenue, required verified units, low-result loss and evidence-delay treatment. Preserve the distinction between a forecast, a contractual settlement proxy and a demonstrated causal effect.
The measurement owner must confirm data feasibility and reproducibility. Delivery must confirm influence over the result and the service protections. Finance must approve downside, funding and excluded costs. Contract reviewers determine the permissible terms and change route. The authorized commercial owner then decides whether to proceed. A model operator or bid writer cannot make an external result controllable by changing the wording of the promise.
If the baseline or data cannot support the mechanism, a permitted discovery phase, pilot, capped variable component or output-based alternative may be considered. Each needs a defined scope and authority; none is an automatic right. A pilot should test measurement and delivery assumptions before the price becomes binding, not postpone all difficult commercial choices until after award. If the tender requires an unsupported guarantee with no permitted treatment, return a clear stop or escalation decision.
Put the approved payment definition, necessary dependencies and verification method in the required buyer-facing schedules. Keep private margin floors and negotiating analysis internal unless disclosure is required. Reopen approval when the eligible population, buyer policy, data access, observation window, price formula or service design changes. The finished bid should offer a result that can be measured and responsibly priced, with a specific decision where either condition is missing.
What good looks like
Useful outcomes from outcome-based pricing viability in a tender
- The payable outcome is distinguishable from activity, delivery and a broad buyer ambition.
- Both parties can reproduce the calculation using an agreed data and measurement protocol.
- External influences and buyer responsibilities have an explicit treatment.
- The commercial decision includes zero-reward, adjusted-result and capped-upside cases.
- Quality protections prevent apparent success from rewarding worse service.
- The approved offer states its payment basis without exposing private commercial limits.
Operating model
How to run the work
- 01
Identify the payable result
Read the price and performance schedules together. Separate the required service, the measured result and the part of consideration linked to that result.
- 02
Fix the measurement agreement
Define population, baseline period, observation window, units, exclusions, comparison method, data sources and calculation order. Resolve who can change or approve them.
- 03
Test influence and evidence
Map supplier actions and buyer dependencies to the outcome. Check lawful access, completeness, verification effort and the treatment of missing or disputed records.
- 04
Calculate the payment cases
Apply the actual base fee, variable rate, thresholds, rounding, caps and deductions. Reconcile costs and identify the outcome level needed for approval.
- 05
Challenge the incentives
Test whether the formula rewards selection, premature closure, channel displacement or reduced access. Require enforceable quality and coverage conditions.
- 06
Approve or change through a permitted route
Have finance, delivery, measurement and contract owners approve the same position. If evidence or funding is inadequate, seek an allowed clarification or redesign before committing.
Evaluation
Questions that change the decision
- Is payment attached to a deliverable, a service standard or a change in the buyer’s condition?
- What part of that change is counted under the agreed settlement rule?
- Can the supplier influence the result while meeting all service obligations?
- Will the necessary buyer-controlled data be available and usable in time?
- What is earned if no variable outcome is verified?
- Does the maximum reward make the approved cost and contribution achievable?
Failure modes
Where teams lose control
A before-and-after change is priced as a proven supplier effect.
The baseline or eligible population changes after the price is committed.
Hard cases are excluded to improve the paid indicator.
Missing evidence is silently treated as either success or failure.
The reward cap is mistaken for a cap on all contractual liability.
An expected bonus is used to cover costs before it can be measured or paid.
Measurement
Measure the finished job
Measure the completed workflow, including review effort and exceptions. Output volume on its own is not evidence of a better process.
- Revenue exposed to outcome verification
- Minimum verified settlement units needed for the contribution floor
- Maximum attainable payment under the stated cap
- Measurement and verification cost included in the estimate
- Records unresolved at the measurement cut-off
- Time between service delivery and final outcome settlement
Questions
Common questions
Is outcome-based pricing the same as milestone billing?
No. A milestone often pays for an identifiable stage of work, while an outcome payment depends on an agreed result experienced by the buyer or service user. Either can have acceptance and evidence conditions. Determine the actual payment trigger rather than relying on the label.
Does a lower error or repeat-contact rate prove supplier impact?
Not on its own. Other changes can influence the rate. Examine the baseline, population, comparison and alternative explanations. A contract may define an agreed adjustment for settlement without establishing a scientifically validated causal estimate. Keep those claims separate.
Can missing records be counted as successful outcomes?
Only a supported, explicitly agreed rule can determine settlement treatment; missingness itself does not prove success. Preserve unresolved records and apply the stated verification, waiting or estimation procedure. Do not remove difficult cases from the denominator to improve payment.
What does a blended payment model change?
It separates a base service amount from a result-linked amount and can change the supplier’s exposure. The base fee still has its own conditions. Calculate downside and funding rather than assuming the blend covers costs. A bidder can propose it only where the procedure allows.
How can a cap make the commercial model impossible?
The cap sets maximum reward. If base fee plus that maximum cannot meet the required revenue at the approved cost, no better result can repair the economics. Alderpoint needs at least 534 settlement units for its illustrative 10% contribution floor; the cap is reached at 600 units.
Who should own the measurement data?
Identify the actual source owner and agree lawful access, extraction, verification, correction and retention responsibilities. Ownership alone does not establish that the bidder will receive complete usable evidence in time. Confirm the process with the buyer and measurement reviewer before relying on the associated revenue.
When should a supplier decline a results-linked offer?
Escalate or stop when the result cannot be defined, evidence cannot be obtained, the supplier lacks material influence without an agreed risk treatment, or downside and funding exceed approval. Seek a permitted clarification or redesign where available; do not promise that unresolved conditions will be fixed after award.
Sources
Primary references
- Risk allocation and pricing approaches, updated June 2026 Cabinet Office
- Quality in policy impact evaluation: outcome and attribution HM Treasury
- FAR 37.601: standards and assessment in performance-based services Acquisition.gov
- Outcome-based payment schemes: historical findings from 2015 National Audit Office
- Effective contracting of employment and health services: evidence review Department for Work and Pensions
About the author
Tony Kim
Tony writes about applied AI, dependable product engineering and the systems that turn complex response work into controlled delivery.
Zelius
Managed tender intelligence and bid execution for teams that want the commercial outcome.
Suppliers, founders and commercial teams pursuing public or private opportunities. Start with the workflow, constraints and evidence you already have.