A service-credit exposure model converts defined performance failures into the deductions and other consequences the offered contract would produce. Its work product links each input to the issued clause and shows eligible charges, measurement bands, failure counts, repeat rules, permitted overlap treatment, period limits, rolling headroom, settlement dates and remaining remedies. It preserves performance points even where a monetary limit applies. The model supports an acceptance decision; it does not establish legal enforceability, authorize a weaker service promise or prove that credits are the buyer’s only remedy.
The pricing workbook includes two percent of annual revenue for service credits. The contract instead calculates deductions on specified monthly charges, doubles points for certain consecutive failures and uses a rolling limit already partly consumed. The same outage may also trigger rectification duties or a separate termination threshold. The allowance cannot answer what the next invoice loses, how much cash the supplier needs or which obligation remains after the credit limit is reached.
Calculate a dated sequence from the issued mechanism before deciding whether its risk is acceptable. This guide reviews financial consequences of existing service-level terms, whereas a service-level answer first defines and authorizes the performance promise. Sources were checked on 6 September 2026. UK government models, US federal clauses and a cloud provider’s SLA illustrate different arrangements; none is a universal rule. All cases, rates and contract terms in the worked examples are fictional. Counsel owns interpretation and enforceability, and only the delegated approver can accept the commercial commitment.
Calculation record
Build the model from the issued mechanism, not an allowance
Collect the complete performance schedule together with the charging annex, relief events, service reporting, set-off, liability and termination terms. A definition in the main contract can change a formula in the schedule. Record document version, clause location and precedence beside the calculation. If a schedule says monthly but a definition says a quarter, the model needs an open interpretation point, not an analyst’s silent choice.
Use separate fields for the performance measure and the monetary consequence. A service may miss its target without earning financial points, or a subsidiary indicator may trigger corrective work rather than a credit. Identify the promised outcome, eligible population, observation window, thresholds and evidence source. Then identify the point award or direct deduction. Do not give every KPI the same financial rule merely because the reporting table presents them together.
The Model Services Contract for England and Wales connects performance provisions with charging and financial-limit provisions. The September 2025 schedules include point-to-charge calculations, while the core terms address remedy interactions. These are useful examples of why the review crosses documents. They are not instructions to import the model’s percentages, thresholds or commercial choices into a different procurement.
A usable calculation record holds enough information to rebuild one period without the original analyst. Include the source evidence reference, the exact input before display rounding, any approved exclusion, the resulting points, the applicable charge and each limit. Keep protected telemetry in its authorized location. The public-facing explanation should use synthetic examples, not incident logs containing customer data or confidential operating detail.
| Part | Required input | Unresolved example |
|---|---|---|
| Performance | Service, period, population, measure and exact threshold | Availability is rounded before its band is selected |
| Money | Eligible charge components, rate and tax treatment | Pass-through costs appear in two different bases |
| Interaction | Repeat, overlap and cap ordering rules | The same incident has two indicators but no agreed treatment |
| Decision | Scenario, evidence, owner, remedy review and approval | A supplier refund is assumed before eligibility is established |
Inputs and boundaries
Keep measurement, exclusions and the charge basis explicit
Availability needs a defined denominator. State the service calendar, time zone, eligible minutes and treatment of partial service, planned maintenance and customer-caused disruption. For response-time measures, define when the clock starts and stops, which tickets qualify and what pauses it. If the contract does not resolve a material question, calculate the plausible branches and seek an authorized clarification. Missing telemetry must not automatically become a successful period.
Test band edges using the unrounded value unless the contract directs otherwise. With a target of 99.9 percent, an exact 99.900000 percent and 99.899999 percent can fall into different bands even though both display as 99.9. A monthly average of daily percentages can also differ from the ratio of eligible minutes. Use the specified calculation, including weighting and zero-denominator treatment, and keep the displayed number separate from the value that selected the band.
Exclusions need evidence and an applicable clause. Planned maintenance outside the approved window is not necessarily permitted maintenance. A third-party outage is not necessarily buyer-caused relief. Record the affected interval, causal evidence, required notice and the person authorized to determine treatment. Retain an inclusion scenario when the exclusion is disputed, so the approver can see the amount at stake without treating the dispute as resolved.
For the monetary base, identify service fees before deductions, pass-through items, one-off implementation charges, indexation, partial months and volume adjustments. Check whether VAT or another tax is inside the specified calculation and obtain the required tax treatment separately. Applying a percentage to revenue after the same credit creates a circular formula unless that is what the contract expressly requires. When fees change, update both the credit base and any cap tied to a historical charge window.
Worked performance sequence
A repeated failure changes the next period, not the past one
The fictional Marlow tender bills GBP 120,000 each month: GBP 100,000 of eligible managed-service charges and GBP 20,000 of excluded pass-through charges. One service point deducts 0.5 percent of the eligible base, or GBP 500. Availability at least 99.9 percent earns no points; at least 99.5 but below 99.9 earns two; at least 99.0 but below 99.5 earns five; below 99.0 earns ten. These are invented terms, not recommended targets or a reproduction of a government schedule.
Marlow’s assumed repeat rule doubles only the availability points in the second and subsequent consecutive failed months. It does not compound the previous doubled number. A month meeting the target resets the sequence. A separate support measure awards one point per qualifying late response, capped at four support points monthly. The example’s issued overlap rule removes support misses attributable solely to the same outage already counted for availability. The support counts below are the remaining, independently eligible misses after that check.
The four availability inputs are stipulated exact values. September produces two availability points plus one support point. October produces five doubled to ten, plus four support points. November produces ten doubled to twenty, plus three. December meets the availability target and has two support points. The gross deductions are therefore GBP 1,500, GBP 7,000, GBP 11,500 and GBP 1,000. A separate eight-percent monthly cap on eligible charges limits November to GBP 8,000 before the rolling limit is considered.
Do not extend the example’s overlap rule to unrelated measures. Shared root cause alone does not prove that the contract permits one deduction to replace another. Preserve the causal links and the treatment of each affected indicator. Also review the issued repeat terms against the applicable procurement guidance: the Cabinet Office’s June 2026 risk-allocation guidance discourages escalating deduction mechanisms, while its 2025 model contains repeat-failure provisions. That difference is a reason to clarify and obtain advice, not permission to delete a tender clause unilaterally.
| Period | Exact availability | Availability plus support points | Gross credit | After monthly cap |
|---|---|---|---|---|
| September | 99.7% | 2 + 1 = 3 | GBP 1,500 | GBP 1,500 |
| October | 99.2% | (5 × 2) + 4 = 14 | GBP 7,000 | GBP 7,000 |
| November | 98.8% | (10 × 2) + 3 = 23 | GBP 11,500 | GBP 8,000 |
| December | 99.95% | 0 + 2 = 2 | GBP 1,000 | GBP 1,000 |
Cap headroom
Carry the rolling limit forward and remove expired amounts
Marlow’s fictional contract has a fixed GBP 30,000 limit on credits within the current and previous eleven service months, applied after the monthly cap. For this example, the cap amount does not change with revenue, and only the applied credit consumes it. The opening ledger contains GBP 18,000 of prior applied credits. None expires before September’s calculation, GBP 4,000 expires before October, none before November and GBP 3,000 before December. Those are stipulated dated ledger inputs, not extra credits charged during the four-month scenario.
For each month, subtract the expiring amount from the previous closing balance. Headroom equals the greater of zero and GBP 30,000 less that retained balance. The applied credit is the smaller of the monthly-limited credit and headroom. Add only the applied credit to the retained balance. Keep the original gross amount and the performance points in separate columns; a smaller payment does not make November’s service better.
November has only GBP 7,500 of headroom, so its GBP 8,000 monthly-limited credit becomes GBP 7,500. The four applied credits total GBP 17,000, compared with GBP 21,000 gross and GBP 17,500 after monthly limits alone. December’s GBP 3,000 expiry reopens headroom even though the limit was reached in November. An annual reset on 1 January would produce a different answer and must not be substituted for this rolling rule.
Actual contracts may calculate the limit from previous charges, aggregate other compensation into it, or treat accrual differently from invoicing. Follow the specified basis and time window. The UK model illustrates another important separation: a financial limit need not stop performance points from accruing. Retain the history used for escalation, reporting and termination tests even when the credit ledger has no headroom. Do not imply that a cap eliminates all financial or nonfinancial consequences.
| Period | Expired before calculation | Retained prior balance | Headroom | Applied credit | Closing balance |
|---|---|---|---|---|---|
| September | 0 | 18,000 | 12,000 | 1,500 | 19,500 |
| October | 4,000 | 15,500 | 14,500 | 7,000 | 22,500 |
| November | 0 | 22,500 | 7,500 | 7,500 | 30,000 |
| December | 3,000 | 27,000 | 3,000 | 1,000 | 28,000 |
Commercial effect
Show lost contribution without treating a provider credit as cash
Over the four Marlow months, billed charges before credits total GBP 480,000. Assume the delivery cost already priced is GBP 95,000 per month, including the pass-through cost and stated overhead allocation. Base contribution is GBP 100,000. The GBP 17,000 of applied credits reduces it to GBP 83,000. A further GBP 12,000 of corrective work in this scenario reduces it to GBP 71,000. Tax, financing, damages and exit costs are excluded from this teaching calculation and require their own review. This is a scenario contribution, not a forecast profit or maximum loss.
Add settlement dates before using the result for cash planning. The example assumes the buyer deducts each applied credit from the next monthly invoice. November’s GBP 7,500 therefore affects the December invoice, while December’s GBP 1,000 reaches the following January invoice. Corrective work may be paid earlier. Record invoicing, due dates and forecast receipts separately, and include credits still due after service exit. A contribution calculation alone does not prove that the supplier can fund the payment gap.
Upstream recovery needs its own eligibility and timing record. A provider may measure a narrower component, calculate on a smaller fee base, require a timely claim and offer only a credit against future use. The public Amazon Compute SLA illustrates why a downstream service promise cannot be assumed to mirror an upstream credit. Its definition and claim conditions must be checked for the actual architecture and account. A proposed recovery must remain conditional until the right and amount are established.
Run a no-recovery case even when procurement expects reimbursement. If the supplier later receives an applicable credit, show when it can reduce an actual future payment; do not subtract it from buyer deductions as if it were immediate cash. The Marlow contribution above assumes no recovery. Nor should a low expected credit cost justify deliberately missing a service target: the performance duty, customer impact, corrective work and other remedies remain part of the decision.
Legal and operational limits
Read the exceptions to any exclusive-remedy statement
An exclusive financial remedy for a defined KPI failure is narrower than the sole remedy for everything that happened. Identify the covered event, beneficiaries, exclusions and exceptions. Check severe or repeated failure, deliberate misconduct, data loss, inaccurate reporting, third-party loss and termination provisions where present. Counsel must assess classification, enforceability and applicable law; calling a deduction a service credit does not settle whether another jurisdiction treats it as a price adjustment, penalty or other remedy.
The England and Wales Model Services Contract contains exceptions to its exclusive-financial-remedy wording and separate provisions for serious performance consequences. That is direct evidence against assuming that the phrase always closes the analysis. The US FAR clause 52.246-4, when applicable, illustrates another mechanism: nonconforming services can lead to reperformance, price reduction or further action under its conditions. Do not rename all such consequences credits or apply a credit cap to them without a basis.
Prevent double counting in the model without inventing a legal right to offset. A contractual replacement payment might displace the ordinary credit for the same period, while an independent repair cost remains. Record gross alternative branches until the contract resolves whether amounts stack, replace or are credited against another remedy. The current UK risk-allocation guidance asks buyers to consider the combined payment mechanism, but that guidance does not itself alter an issued agreement.
Keep nonfinancial triggers visible beside the money. A missed reporting deadline, a corrective-plan failure or a specified accumulated point total can matter independently of the credit amount. Name the owner who tracks each trigger and the required response date. A capped deduction is not permission to stop measuring, omit an adverse report or abandon corrective action. Where a legal conclusion remains unsettled, preserve the hold and the question rather than presenting an unsupported maximum exposure.
Decision and handover
Approve the model, the operating response and the remaining risk
Use scenarios that could change the decision: a normal month, an exact band boundary, a recurring failure, a shared incident, a severe period, an exclusion rejected by the buyer and a cap nearly consumed. They need not have invented probabilities. If credible performance data supports an expected-value calculation, show its population and limitations separately. A short history from another service is not enough to assign a precise probability to this tender’s worst case.
The Marlow record supports a specific conclusion: the four-month scenario produces GBP 17,000 of credits and GBP 71,000 contribution after stipulated corrective work, with no assumed provider recovery. It does not approve all possible losses. The commercial authority must decide whether that scenario and the remaining exposure are acceptable; delivery must own measurement and recovery work. Any unresolved clause affecting a material branch remains a condition on release.
Bind approval to the performance schedule, charging version, service design and price. Reopen the model when targets, eligible fees, supplier architecture, measurement periods or remedy terms change. Preserve who owns source telemetry, exclusions, monthly reports, credit notes, supplier claims and rolling-ledger reconciliation. A signed commercial approval does not repair an operating process that cannot produce the required evidence on time.
An agent can extract authorized rules, produce calculation branches and check the disclosed arithmetic. It must retain missing data and disputed interpretation as such. It may not reduce a promised service level, approve an exclusion, contact the buyer, submit a provider claim, issue a credit note or accept contract risk without explicit authority. Its useful output is a versioned exposure record with reproducible numbers and a clear decision boundary.
| Question | Evidence from this model | Required decision |
|---|---|---|
| Are monetary calculations reproducible? | Four periods, gross credits GBP 21,000 and applied credits GBP 17,000 | Finance verifies charge bases, cap ordering and dated opening ledger |
| Can the service operate the rules? | Measurement and response duties continue after a cap is reached | Delivery assigns reporting, exclusion evidence and corrective-work owners |
| Is the scenario commercially acceptable? | GBP 71,000 contribution after credits and stated corrective work | Delegated authority reviews scenario and separately unquantified consequences |
| Is the offer ready to accept the terms? | Recovery assumed zero; remedy interpretation must be settled | Resolve material legal questions and bind approval to the exact offer |
What good looks like
Useful outcomes from model tender service credit exposure
- Each deduction can be reproduced from a clause, a measured input and an eligible charge basis.
- Repeated failures and related failures are treated according to explicit rules.
- The rolling ledger shows unused headroom and amounts leaving the window.
- Contribution loss, invoice timing and conditional supplier recovery remain separate.
- The approval records unresolved interpretation and operational responsibilities.
Operating model
How to run the work
- 01
Freeze the contract mechanism
Collect performance, charging, relief, reporting, liability and termination provisions with their versions and priority. Record unresolved cross-references rather than filling them with usual practice.
- 02
Define the measurement inputs
Identify the service boundary, population, period, clock, source evidence, thresholds, exclusions and rounding rule. Assign owners to missing or disputed data.
- 03
Calculate gross deductions
Apply the relevant bands, points, repeat treatment and approved overlap rule to each dated scenario. Convert points using the correct charge base before monetary limits.
- 04
Run the cap ledger
Apply limits in their contractual order. Carry eligible prior deductions, remove amounts that expire and preserve performance history independently of capped money.
- 05
Reconcile commercial effects
Show contribution, corrective costs and invoice dates. Keep conditional upstream credits separate from buyer deductions and map additional remedies for specialist review.
- 06
Approve and hand over
Resolve material ambiguities through the permitted route, obtain operating and commercial authority, bind the model to the final offer and assign monitoring and settlement owners.
Evaluation
Questions that change the decision
- Which charges actually enter the percentage calculation?
- Does a second failed period increase points, money or both?
- Which related failure can be excluded and on what evidence?
- How much headroom remains on this date?
- What survives the monetary cap or an exclusive-remedy clause?
- Who approves the residual exposure and operates the reporting process?
Failure modes
Where teams lose control
Annual average performance hides a month that crosses a severe band.
A credit percentage is applied to total revenue instead of eligible service charges.
A monthly cap is mistaken for a rolling annual limit.
One root cause is used to remove deductions without a contractual overlap rule.
A future provider credit is treated as cash already recovered.
A monetary cap is used to erase performance points or termination exposure.
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.
- Unresolved calculation rules
- Gross and capped deductions by service period
- Rolling headroom after expiries
- Contribution after credits and corrective work
- Buyer deduction date versus supplier recovery date
- Open remedy questions and release conditions
Questions
Common questions
Is the service-credit cap the maximum possible loss?
Only for the liabilities and period it actually covers. Check excluded categories, other compensation, corrective costs and exceptions to any exclusive-remedy wording. Performance points and termination triggers may continue independently. Obtain legal review before making a maximum-exposure claim.
Can one outage produce several deductions?
Read the rule for related indicators and remedies. A common cause alone does not authorize removing a deduction, while a specific replacement or overlap provision may prevent stacking. Keep alternative calculations open until the relevant interpretation is resolved.
Does repeat failure mean the charge doubles every month?
Not necessarily. The rule may multiply the current band once, compound a previous amount or use another mechanism. Identify the repeated measure, period, reset and multiplier. In the fictional Marlow case, only the current availability points double once; no compounding occurs.
Why keep points after the monetary cap is reached?
The contract may use performance history for escalation, rectification or termination independently of financial deductions. Preserve the measured failure and gross points even when the amount applied is capped. Do not infer that the target has been met because no further credit is payable.
Can provider credits offset the buyer’s deduction?
They require a separate review of entitlement, measurement, fee base, claim evidence, deadline and settlement form. A future-use credit is not immediate cash. Model no recovery and show any established recovery in the period when it can actually reduce an outflow.
Should the model assign a probability to every scenario?
Only where evidence supports that estimate. Scenario analysis can expose thresholds and funding needs without pretending to forecast their frequency. Label assumptions, retain severe and no-recovery branches, and let the authorized reviewer decide how the resulting exposure affects the bid.
Sources
Primary references
- Model Services Contract core terms, England and Wales: performance and financial limits Cabinet Office
- Model Services Contract schedules: performance points and charges, v2.2A Cabinet Office
- Risk Allocation and Pricing Approaches guidance, updated June 2026 Cabinet Office
- FAR 52.246-4: remedies for nonconforming fixed-price services US Federal Acquisition Regulation
- Amazon Compute SLA: credit eligibility, fee basis and claim procedure Amazon Web Services
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.