Evaluate · contracts and service credits

SLA penalty calculation formula, set as rules an agent applies

An SLA penalty calculation formula multiplies the shortfall below a floor by a rate and a fee, then stops at a cap. This guide works one month to the dollar, shows how the same month prices four ways, and how Policy Servicing Analytics applies the contract's own arithmetic.

Updated · 13 min read

The usual SLA penalty calculation formula for an outsourcing contract is: credit = the monthly fee × the credit rate × the number of points the measured result falls below the floor, capped at a stated share of the fee. In policy servicing the result is usually accuracy or turnaround, the rate is a percentage of the monthly fee per point, and the cap is a percentage of the fee or of a year's charges. The arithmetic is simple; the disputes are not, because they are about which number goes in: the vendor's sampled figure or the insurer's own measure. Botminds' Policy Servicing Analytics measures every finished transaction, then computes each contract's credit exactly as its schedule states and drafts the notice for the vendor manager, so the formula is applied the same way every month and nothing is sent until a person signs.

The worked example uses a fictional sample insurer, Ashcombe Mutual, and its fictional outsourcer, Harrowfield Solutions, under a master services agreement with a fee of $186,400 a month. Contract wording on public templates comes from the UK government's Model Services Contract and the US Federal Acquisition Regulation, linked where quoted.

From the contract schedule to a setting to October's credit, in Policy Servicing Analytics

TermWhat the schedule saysThe settingOctober to date
Accuracy floor98.0% a monthFloor 98.0%, monthlyMeasured 95.7%
Who measures, on whatThe insurer, on completed workEvery finished transaction, checked against its request4,212 transactions, 29 Sep to 12 Oct
What counts as inaccurateA transaction the insurer must redo within 10 business daysRework window 10 business days; customer-requested changes excluded181 transactions redone
Shortfall unitFull points below the floorWhole points only2.3 points, so 2 full points
Credit rate4% of the monthly fee per full point4% a point8%
BaseThe monthly fee$186,400$186,400
Cap12% of the monthly feeCap 12%Not reached
Dispute window5 business days per charged reworkVendor dispute window 5 business days; the quality analyst decidesOpen
CreditComputed on the measured monthDrafted, sent only after sign-off$14,912 if October closes here
Fictional sample contract between Ashcombe Mutual and Harrowfield Solutions (master services agreement of 14 January 2025, schedules C-4.1 and C-4.2), as set up in Policy Servicing Analytics on the sample data.

The SLA penalty calculation formula, term by term

Every service credit formula answers the same five questions: what is measured, where the floor is, how the shortfall is counted, what each unit of shortfall costs, and where the total stops. Write them as one line and the formula for a floor-and-rate schedule is: credit = min(cap, units below floor × rate per unit) × fee for the period.

The terms carry more judgement than the line suggests. The measure decides whose view of the month counts. The unit decides whether 2.3 points below the floor is two points or 2.3. The rate decides how much a point hurts, and the base decides what it hurts: the monthly fee, a share of it allocated to that service level, or the charges for the service that failed. The cap decides when the credit stops growing and other remedies take over. A contract that leaves any of the five vague will be argued over every month in which the vendor misses.

Public-sector templates make the same terms explicit. The UK Cabinet Office's Model Services Contract converts service level failures into "Service Points", and its charging schedule turns points into money "on the basis of one point equating to a [Insert percentage]% deduction in the Service Charges", applied to "the total Services Charges payable for the relevant Service Period (prior to deduction of applicable Service Credits)" (UK Cabinet Office, The Model Services Contract, Schedules, version 2.2A, Schedule 11 Part C). The base is the charge before credits, so a credit is never computed on a fee already reduced by another credit.

What is a service level credit, and is it a penalty?

A service level credit is a reduction in the price of a service that did not meet its stated level, usually deducted from the next invoice. Contracts are written to treat it as a price adjustment rather than a punishment: the UK model states that "Service Credits are a reduction of the Service Charges payable in respect of the relevant Services to reflect the reduced value of the Services actually received", and that they are "shown as a deduction from the amount due from the Authority to the Supplier in the invoice for the Service Period immediately succeeding the Service Period to which they relate."

US federal contracting uses the same logic. Under the standard fixed-price services inspection clause, when defects "cannot be corrected by reperformance", the government may "reduce the contract price to reflect the reduced value of the services performed" (Federal Acquisition Regulation 52.246-4(e)). "SLA penalty" is what people search for; "service credit" is what the contracts say, and the framing as reduced value is why a credit should be priced on what the buyer actually received.

SLA penalty calculation example: one month, worked to the dollar

Here is the sample contract's credit for October, computed in the order a worksheet should follow. Each step uses only the contract's terms and the measured month, so anyone with the schedule and the list of transactions can reproduce it.

SLA penalty calculation worksheet for one month of outsourced endorsements and billing

StepCalculationResult
1. Transactions finished in the periodEndorsements 2,486 + billing and refunds 1,7264,212
2. Transactions the insurer redid within 10 business daysEndorsements 147 + billing 34, customer-requested changes excluded181
3. Measured accuracy1 − 181 ÷ 4,21295.7%
4. Shortfall below the floor98.0% − 95.7%2.3 points
5. Full points (the contract's unit)Whole points in 2.32
6. Credit rate2 × 4% of the monthly fee8%
7. Check the cap8% against a cap of 12%Under the cap
8. Credit8% × $186,400$14,912
9. The same steps on the vendor's own figure98.0% − 99.2% is above the floor$0
Fictional sample month, 29 September to 12 October, for Harrowfield Solutions in Policy Servicing Analytics. Steps 1 to 3 are measured by the insurer; steps 4 to 8 apply the schedule. The figure is provisional until the month closes and disputes are decided.

Step 2 is where the money is. Each of the 181 transactions is linked to the request that started it and to the later change that corrected it, and four later changes in one morning's run were left out because the customer had asked for them. A vendor that disputes a rework within five business days gets a ruling from the insurer's quality analyst, and the credit moves with the ruling. How those links are made, and why the team that fixes errors should not be charged with them, is set out in the rework rate formula for policy services.

Whose measure goes into the formula?

The contract should say, and it should name the buyer. The US rule for performance-based service contracts requires "Measurable performance standards (i.e., in terms of quality, timeliness, quantity, etc.) and the method of assessing contractor performance against performance standards", and requires any incentives to "correspond to the performance standards set forth in the contract" (Federal Acquisition Regulation 37.601(b)). The method of assessment is part of the deal, not a detail for later.

The same October under the vendor's measure and under the insurer's

Vendor's own sampleInsurer's measure, every transaction
What is checked

Vendor's own sample: 220 endorsements across the month, chosen by the vendor.

Insurer's measure, every transaction: All 4,212 transactions finished from 29 Sep to 12 Oct, each against its request.

What counts as an error

Vendor's own sample: What the vendor's checker finds on review.

Insurer's measure, every transaction: Any field that differs from the request, and any transaction the insurer had to redo within 10 business days.

Errors found

Vendor's own sample: 2

Insurer's measure, every transaction: 181 transactions redone, causing 433 more items of in-house work.

Accuracy

Vendor's own sample: 99.2% reported

Insurer's measure, every transaction: 95.7% measured

Credit under Schedule C-4.2

Vendor's own sample: $0

Insurer's measure, every transaction: $14,912

Fictional sample month for Harrowfield Solutions in Policy Servicing Analytics. The contract says the insurer measures, on completed work, and that a transaction the insurer must redo within 10 business days is not accurate.

The difference is not a rounding question. A sample of 220 endorsements that finds 2 errors cannot come from a population where 5.9% of endorsements were wrong except by very bad luck, so the two measures are counting different things: the vendor's checker looks at a transaction on its own, the insurer's measure looks at what happened to it afterwards. Why a vendor's sample and a check of every transaction diverge, and what a sample of that size can prove, is the subject of 100% inspection vs sampling.

The contract term, the measured month and the credit worked out step by step

Vendors and contracts screen for Harrowfield Solutions, Pune, master services agreement 14 January 2025, fee $186,400 a month. Contract terms with promise and measured October to date: endorsements turnaround 2 business days, 81.4% on time, 418 will miss by Thursday; billing and refunds turnaround 3 business days, 96.8% on time; accuracy floor 98.0% a month, 95.7% measured and 99.2% reported; how accuracy is measured, by Ashcombe on completed work, every transaction with rework within 10 business days; correction of errors 1 business day from send-back, 11 sent and 9 corrected on time. Service credit panel: measured 95.7%, full points below 98.0% 2, credit at 4% a point 8%, owed if October closes here $14,912. A drafted accuracy notice to the vendor's delivery lead waits for approval.
Policy Servicing Analytics on the fictional sample contract. The note under the credit reads: against the vendor's own figure (99.2%) nothing would be owed.

Full points, fractions or bands: one month priced four ways

The unit of shortfall changes the credit as much as the rate does. The same measured month, 95.7% against a 98.0% floor on a $186,400 fee, produces four different credits under four common structures. None is wrong; each has to be written down so that nobody chooses after the month is known.

The same month under four credit structures: from $4,287 to $22,368

Straight line: shortfall in points as a % of the fee (2.3%)
$4,287
Bands: 3 service points at 2% each (illustrative)
$11,184
Full points × 4% (the sample contract)
$14,912
Fractional points × 4% (2.3 × 4% = 9.2%)
$17,149
Bands, second month running (points doubled, at the 12% cap)
$22,368
Our arithmetic on the fictional sample month (95.7% measured, 98.0% floor, $186,400 fee, cap 12%). The bands are an illustration in the shape of the UK Model Services Contract: 1 point for 97.0–97.9%, 2 for 96.0–96.9%, 3 for 95.0–95.9%, 4 below 95%, at 2% of the fee a point. Only the full-point line is the sample contract's own rule.

Whole-point schedules are easy to explain and leave a cliff at each point: at 97.00% the sample vendor owes 4% of the fee, at 97.01% it owes nothing. Fractional schedules remove the cliff and add an argument about decimal places, which is why the rounding rule belongs in the clause. Banded schedules, like the UK model's minor, serious and severe failures, trade precision for predictability and are often paired with a multiplier for repeat failures.

For a second, smaller vendor the unit decides whether anything is owed at all. Selwyn measured 97.2% against the same 98.0% floor: 0.8 points short, which is zero full points, so its full-point schedule at 3% a point produces no credit, while a fractional schedule would produce 2.4% of its fee. That is a commercial choice the contract made, and the measure should show the shortfall either way.

Caps, repeat failures and when credits stop being the only remedy

A cap limits the vendor's exposure in a bad month, and a well-written contract says what happens when the cap is reached. The UK model sets the cap, after the first year, at a percentage of "the Service Charges paid and/or due to be paid to the Supplier under this Contract in the period of 12 months immediately preceding the Service Period in respect of which Service Credits are accrued". It then treats credits that "meet or exceed the Service Credit Cap" as a "Critical Performance Failure", and credits reaching "75% of the Service Credit Cap" as an event that lets the buyer step in with a remedial adviser.

Repeat failures are the other lever. Under the same model, a failure on the same indicator in two consecutive periods is a "Repeat KPI Failure", and its points are calculated as "SP = P x 2". On the sample month, a vendor in the severe band for a second month running would reach the 12% cap.

The American clause sets the fallback in plain terms. If the contractor fails to reperform promptly, the government may "by contract or otherwise, perform the services and charge to the Contractor any cost incurred by the Government that is directly related to the performance of such service" or "terminate the contract for default" (Federal Acquisition Regulation 52.246-4(f)). In servicing, the first of those already happens every day without a contract term: the in-house team re-keys the vendor's wrong transaction. A clause that recognises it lets the insurer charge for it.

Service level credits clause: what to write down before the first month

Most service credit disputes come from a term the clause did not settle. The checklist below lists the terms that decide the arithmetic, in the order they enter the formula.

Terms a service level credits clause should settle

  1. The measure, and who takes it

    Name the buyer as the party that measures, on completed work, over every transaction or a stated sample.

  2. What counts as an error

    Field-level differences from the request, and any transaction the buyer must redo within a stated window.

  3. What does not count

    Changes the customer asked for after the first transaction; items the buyer itself delayed.

  4. The floor and the period

    For example 98.0% a calendar month, per service line or across the contract.

  5. The unit and the rounding

    Full points, fractional points to one decimal, or bands, written with an example.

  6. The rate and the base

    A percentage of which fee, before or after other credits.

  7. The cap and what happens at it

    A share of the monthly fee or of the last 12 months' charges, and the remedy that applies once it is reached.

  8. Repeat failures

    A multiplier or a step up when the same service level fails in consecutive periods.

  9. Disputes

    A window for the vendor to dispute each charged error, and the person on the buyer's side who decides.

  10. Timing

    Deducted from the next invoice, as the UK model provides, or settled at a quarterly review.

Our checklist. The sample contract in Policy Servicing Analytics settles each term: insurer measures every transaction, 10-business-day rework window, 98.0% monthly floor, full points, 4% of the monthly fee, cap 12%, 5-business-day dispute window.

Why the credit belongs in settings an agent applies, not in a spreadsheet

A service credit is arithmetic the contract has already defined, so it should be computed the same way every month by something that cannot be talked into a different rounding rule. In Policy Servicing Analytics each schedule's floor, rate, unit, cap and windows are settings; changes are saved as a draft and approved by the head of policy services, and every change is kept on the audit trail.

Floors, credit rates, caps and windows held as settings, changed only by approval

Settings screen: promises in business days for endorsements 2, certificates 1, cancellations and reinstatements 1, billing and refunds 3 with refunds also against the 30-day state limit, and mortgagee changes 2; rework window 10 business days, exclude customer-requested changes on, count follow-on items on, vendor dispute window 5 business days decided by Imran Siddiqui; floors and credits: Harrowfield accuracy floor 98.0% monthly under Schedule C-4.2, Harrowfield credit 4% of monthly fee per full point below with cap 12%, Selwyn floor 98.0% under SOW 3, Selwyn credit 3% per full point with cap 9%, in-house target 98.5% for Columbus and Dayton; forecast horizon 3 business days, alert when misses pass 100, queues read every 15 minutes; rebalancing approved by Grace Oyelaran; notices and settings approved by Dana Whitcombe. A button saves changes as a draft.
Policy Servicing Analytics on the fictional sample insurer. The two outsourcers' schedules differ (4% a point with a 12% cap, 3% a point with a 9% cap), and each is applied as written.

The Report writer computes the credit from the measured month and drafts the accuracy notice in the vendor manager's voice, with the list of charged transactions attached, each with its request and both versions. On the sample, the notice also asks the vendor for a supervision plan for its new starters, whose 612 endorsements were 82.2% accurate against 98.0% for its tenured staff. Nothing reaches the vendor until the head of policy services approves it.

What the solution does not do is negotiate. It does not decide whether to waive a credit for a vendor that is fixing a staffing problem, or whether to invoke the cap remedies; those are commercial decisions for the vendor manager and the head of services. Its job is to make sure the number in front of them is the contract's number, on the insurer's measure. For how the measured month compares with running the same work in-house, see policy servicing outsourcing services, costed, or start from Policy Servicing Analytics.

Questions people ask

What is the difference between an SLA penalty and a service credit?

In practice they describe the same money: an amount the vendor gives back when a service level is missed. Contracts call it a credit and present it as a reduction in price for reduced value, as both the UK Model Services Contract and the US federal inspection clause do. Writing it that way, rather than as a punishment, keeps the amount tied to what the buyer actually lost.

How is a service level credit allocation percentage set?

In contracts that use one, a share of the monthly fee is put at risk and the allocation percentage spreads that share across the service levels, so a contract might put 10% at risk and allocate 40% of it to accuracy and 30% each to two turnaround measures. The allocation should follow the cost of failure: on the sample, one wrong endorsement caused a 27-hour claim hold and 74 minutes of in-house work, so accuracy deserves a large share.

Are service credits the only remedy for poor service?

Usually not. The UK model treats credits that reach the cap as a critical performance failure with further remedies, and lets the buyer appoint a remedial adviser once credits pass 75% of the cap. The US fixed-price services clause lets the government reperform the work at the contractor's cost or terminate for default if the contractor does not correct it.

Can a vendor dispute a service credit?

It should be able to, item by item, within a fixed window. In the sample contract the vendor has five business days to dispute any rework charged to it, and the insurer's quality analyst confirms or removes it; one disputed effective-date correction was waiting for that ruling on the sample Monday.

How do you calculate service credits for several service levels in one month?

Compute each service level's credit separately under its own floor and rate, add them, then apply the cap to the total, unless the contract caps each one. Keep the base as the charge before any credit, as the UK model does, so one credit never shrinks another.

Policy Servicing Analytics.
See it on your own files.

An agentic solution from Botminds. All service work, in-house and outsourced, on one screen: what is waiting, what will miss its promise, and how accurate the finished work is. Book a demo, or explore how it works.