Skip to content

Systems and records

Green SLA, unmet need: recording supplier SLA business outcome data

By SourceX Editorial · Reviewed by Noah Loul ·

9 sections

Short answer

Supplier SLA business outcome data needs two verdicts for every episode: did the supplier meet the measured service level, and did the business get what it needed? Record what the SLA measured (clock, measurement point, population, exclusions), the need it stood in for, the record that exposes any gap and who noticed first. Grade an AI on both, because a met SLA is evidence about the contract, not the outcome.

Grader rubric

Green SLA, unmet need

This page covers green SLA, unmet need: recording supplier SLA business outcome data.Illustrative

Key takeaways

  • Keep two verdicts per episode, contractual and operational, because systems report only the first.
  • The gap is built in: guidance holds suppliers only to results they can influence, and FAR names contract requirements and agency needs separately.
  • Six design choices open the gap: measurement point, clock exclusions, averaging window, unit of count, response versus resolution and frozen targets.
  • Keep raw timestamps and status history; Request Tracker's own documentation shows a response clock that stays green while an order is overdue.
  • When the SLA was met, the useful remedy changes the measure, not the supplier's bill.

What does supplier SLA business outcome data have to capture?

Two verdicts per episode and the evidence for each. The first is contractual: did the supplier meet the service level as the agreement measures it? The second is operational: did the business get what it needed, where and when it needed it? Supplier SLA business outcome data keeps both, together with what the SLA actually measured, the need it was meant to protect, the record that shows any gap and who noticed it first.

One verdict is the default because one verdict is what systems produce. Supplier scorecards report attainment, service desks flag breaches and dashboards turn green above a threshold. An AI agent asked whether a supplier performed will answer from those records, and in the episodes where measure and need diverge it is likely to be wrong in the supplier's favor.

The sections below explain why the divergence is built into how service levels are written, map six places it opens, show how one ticketing system's documentation describes a clock that stays green, and set out an episode record, an attribution test, an illustrative month and grading rules.

Why can a met SLA and an unmet need both be true?

Because service levels are deliberately narrower than needs. A supplier can be held only to what it controls and what can be measured, while the need includes everything downstream of the handover, and procurement guidance says so in plain terms.

The UK government's guidance on risk allocation and pricing advises that key performance indicators be easily measurable and objective, with suppliers held accountable only for results they can influence. Its guidance on the Procurement Act 2023 says section 52(1) generally requires a contracting authority to set at least three KPIs for contracts above a value threshold, unless it considers that the supplier's performance could not appropriately be assessed by reference to KPIs. The Sourcing Playbook asks that KPIs align with the benefits the contract is meant to deliver, a reminder that alignment has to be designed rather than assumed.

US federal rules draw the same line from the other side. FAR 37.602 asks agencies to describe work in terms of required results rather than how it is done or the hours provided, and FAR 37.603 says performance standards set the level required to meet the contract requirements, then tells agencies to evaluate offeror-proposed standards to determine whether they meet agency needs. Requirements and needs get separate words because they can come apart.

When they do, money does not necessarily follow the outcome. GAO reported in December 2005 that the Defense Department's award fees were generally not tied to acquisition outcomes, and that the department had not compiled data, conducted analyses, or developed performance measures to evaluate the effectiveness of award and incentive fees.

Six places the gap opens between the measure and the need

SLA-need gaps can be grouped into six design choices, and each leaves a specific trace in the records. The matrix pairs every mechanism with what the SLA measured, what the business needed, the record that exposes the gap and the person who usually notices first:

Six places the gap opens between the measure and the need
Gap mechanismWhat the SLA measuredWhat the business neededRecord that exposes the gapWho usually notices first
Measurement pointShip confirmation within 24 business hoursThe part in the line-side store before the maintenance windowReceipt at the point of use against the need time on the work orderMaintenance planner
Clock exclusionsElapsed time outside paused statuses and business hoursElapsed wall-clock time until the problem stopped costing moneyStatus history with timestamps, recomputed without pausesThe requester who waited
Averaging windowMonthly attainment across all order linesEvery critical line on timeLine-level outcomes for items flagged criticalProduction or site manager
Unit of countLines or cases shipped completeComplete kits that let a job startKit completeness by work order, not by lineAssembly or field crew lead
Response versus resolutionTime to first replyTime until the product or service worksReopen events and follow-up tickets linked to the firstEnd user, then the service owner
Frozen targetA threshold set at contract signatureA need that tightened as demand or risk changedNeed dates and lead-time requirements by period beside the unchanged SLAPlanner or category manager at review

How an SLA clock stays green: one ticketing system's own example

Request Tracker, an open-source ticketing system used for IT and service work, documents the mechanics plainly in its service-level configuration. Its example for an order of goods sets a one-hour response deadline and a 24-hour resolve deadline, then warns that a week of always meeting the one-hour response deadline does not mean the ticket is not overdue, because the due date was frozen 24 hours after the order was created.

The same file shows how clocks are shaped on purpose. Starts, the point a deadline counts from, defaults to the first business minute after creation, so a request made outside business hours begins counting the next business day. IgnoreOnStatuses drops a deadline while a ticket sits in a status such as stalled, and ExcludeTimeOnIgnoredStatuses removes the time spent in those statuses from the count. Each option is sensible for a support team, and each means an SLA report can say met while the requester's wait keeps growing.

For a dataset the lesson is to keep raw timestamps and status history, never only the met flag. With the history, an episode's wall-clock wait can be recomputed beside the contractual clock; without it, the gap cannot be seen after the fact.

The SLA gap episode record, field by field

Build one record per episode, joined from the supplier's measurement and the buyer's operational systems, so that someone who was not there can reproduce both verdicts.

Recompute before trusting either verdict. Rebuild attainment from the raw event trail with the SLA term's own rules, business hours and exclusions included, and check it against the supplier's reported flag; then compute elapsed time from the same events against the need time. Where the two calculations differ, the difference is the exclusion or the measurement point doing its work, and it belongs in the record as a number, not a footnote. An episode whose need statement is missing is labeled need unknown, never guessed in either direction.

Who noticed is evidence too. The lag between the need time and the first recorded notice shows how blind the measurement system is to the need: when the first notice is always a planner's downtime entry or a customer complaint, and the scorecard never flags anything, the SLA is instrumented for the contract and not for the business. Record the role, channel and timestamp, so the lag can be computed rather than recalled.

The eight fields and their usual sources:

The SLA gap episode record, field by field
FieldWhat it holdsTypical sourceWhy it is needed
SLA term in forceMetric definition, clock start and stop events, exclusions, threshold and version dateContract schedule or service descriptionA met verdict means nothing without the definition that produced it
SLA measurementMeasured value, window, who computed it and the met or missed flagSupplier scorecard or service desk reportThe contractual verdict as reported
Raw event trailTimestamps for release, status changes, shipment, receipt and resolutionERP, ticketing and carrier eventsLets a reviewer recompute both clocks
Need statementNeed time, place, quantity or service level, and whether it was written down before the eventWork order, purchase order line or requestSeparates a supplier miss from an unstated requirement
ConsequenceWhat happened downstream, such as downtime, an expedite, a substitution or customer impact, in hours or unitsDowntime log, expedite records, complaintsThe operational verdict and its size
Who noticedRole, channel and time of first notice, and the lag after the need timeDowntime log, email, escalation ticketShows whether a measurement system or a person caught it
Attribution labelSupplier performance, unstated requirement, measurement design or outside bothReviewer's decision with evidence linksStops a model from blaming the supplier by default
Decision takenCredit claimed or not, SLA redesign, second source, risk acceptedContract change log, meeting minutesThe outcome a model should learn to recommend

Was it the supplier's miss or a requirement nobody wrote down?

Test the requirement before blaming the supplier. The strongest alternative explanation for a met-SLA, missed-need episode is that the need was never communicated: the work order said Saturday, the purchase order said standard terms, and the supplier did exactly what the contract asked.

Three questions settle attribution. Was the need recorded before the event, in a document the supplier received? Could perfect attainment of the SLA as defined have met the need, or would it still have missed? Did something outside both parties intervene, such as a carrier failure the contract excludes? A need recorded in time that the SLA's definition could never have met points at measurement design, not at the supplier's performance.

The cheapest response is often the wrong one. A service credit tied to the SLA is not due when the SLA was met, and leaning on the supplier informally changes nothing in the contract. Rushing a replacement from an unapproved source to cover the gap can breach a quality or sourcing constraint. The remedies that work change the measure: a critical-item clause measured at the point of use, a wall-clock limit beside the business-hours clock, or a kit-completeness metric.

Illustrative month: 98.5 percent on time and a part 82.5 hours late

Illustrative: the supplier, plant and counts are invented, and the clock arithmetic is in the writer's script. A maintenance, repair and operations supplier commits to ship 98 percent of order lines within 24 business hours of release, counting Monday to Friday from 08:00 to 17:00. In March it ships 394 of 400 lines inside the window, 98.5 percent, so the SLA is met.

One of the 394 is a spindle bearing released on a Thursday at 15:00. The business-hours clock stands at 23 hours when the supplier confirms shipment the following Tuesday at 11:00, so the line counts as met. The plant needed the bearing in its line-side store for a maintenance window starting Saturday at 06:00. It arrived on Tuesday at 16:30, 121.5 wall-clock hours after release and 82.5 hours after the need time, and the press line ran at reduced speed until the bearing was fitted.

Who noticed shows which system works. The maintenance planner logged waiting on part in the downtime log at 06:40 on Saturday, procurement heard on Monday, and the supplier's scorecard on April 1 showed March in green. Only a join between the downtime log's reason code and the purchase order line puts the two verdicts side by side.

Across the month the verdicts disagree on 8 of 400 lines, 2 percent: 3 lines met the SLA and missed the need, and 5 missed the SLA but arrived before anyone needed them, lateness that called for no expedite. A dataset should reproduce this split rather than only the dramatic cell:

Illustrative month: 98.5 percent on time and a part 82.5 hours late
SLA verdictNeed metNeed missedTotal lines
SLA met3913394
SLA missed516
Total3964400
Share of all lines99.0%1.0%100%
Two timelines for the illustrative bearing episode from Thursday 15:00 to Tuesday 16:30: the SLA clock counts only business hours and stops at 23 of 24 hours when the part ships on Tuesday at 11:00, marked met, while the need clock passes the Saturday 06:00 need time and ends with arrival 82.5 hours late.
Same episode, two clocks. Illustrative timings from the worked month on this page.

Grading an agent on two verdicts

Grade the contractual and operational verdicts separately, then grade the remedy against both. A model that reads only the scorecard will pass every met-SLA episode and miss the reason the dataset exists.

At a glance

Grading an agent on two verdicts

Grade on what was knowable at decision time. If the agent is asked mid-episode whether to escalate, judge it on the events recorded up to that moment, not on the consequence that came later; if it is asked afterward to assess the supplier, the full record applies. Mixing the two rewards hindsight. Five rules for the after-the-fact assessment:

  • Contract verdict: correct when the model recomputes attainment from the SLA term in force, including business hours and exclusions, and matches the reported flag or explains the difference.
  • Need verdict: correct only when the model cites the need statement and the consequence record; inferring that the need was met from the SLA flag scores zero.
  • Attribution: the label follows the three questions, and an unstated requirement is attributed to the buyer, not the supplier.
  • Remedy: no credit claim when the SLA was met; for a measurement-design gap the right recommendation changes the measure, such as a point-of-use clause for critical items.
  • Balance: the set includes all four cells, harmless SLA misses among them, so a model cannot score well by treating every SLA verdict as wrong.

Where SLA gap episodes come from, and what this design leaves out

Episodes sit where the buyer's operational records meet the supplier's measurement: scorecards and service reports on one side, downtime logs, work orders, expedite requests, complaints and contract change logs on the other. The join key is usually a purchase order line, ticket number or work order, and an episode without one cannot be rebuilt reliably.

SourceX holds no inventory of these records; it looks for them among U.S. businesses after a buyer explains what it needs. A useful request names the service types, the SLA definitions involved, how many episodes of each verdict combination are wanted, the language, the format, the timing and whether the license covers training, evaluation or both.

This design does not measure how often SLAs and needs diverge in any industry. The illustrative month is invented, the Request Tracker example comes from IT service management, and the FAR, UK and GAO sources describe public-sector contracting. Private supply contracts vary widely, and counsel should read the specific SLA before an episode is treated as a breach or as compliance.

Frequently asked questions

What is the difference between an SLA and a business outcome?

An SLA is a measured commitment with a defined clock, measurement point, population and exclusions, chosen so the supplier can control and prove it. A business outcome is what the buyer actually needed, such as a part on the line before a shift or a service working again. The two overlap but are not the same, so records should carry both.

Why do suppliers meet SLAs while operations still suffer?

Usually because the SLA measures something narrower than the need: shipment instead of arrival at the point of use, business hours instead of elapsed time, monthly averages instead of critical items, or first response instead of resolution. UK guidance, for example, asks that suppliers be held only to results they can influence, which builds the gap in.

Can a buyer claim a service credit when the SLA was met?

Generally not under the SLA itself, because the credit follows the contract's measure. The useful response to a met-SLA, missed-need episode is to fix the measure, for example with a point-of-use clause for critical items, or to put the need in documents the supplier receives. Counsel should review the specific contract.

How do you measure a supplier SLA at the point of use?

Define the stop event as receipt or availability where the work happens, such as the line-side store or the job site, and timestamp it in the buyer's system. Keep the supplier's ship confirmation too, so both clocks can be compared and the transit and handling share of any delay is visible.

Can an AI agent judge supplier performance from SLA reports alone?

It can judge contract compliance, but not whether the business need was met. For that it needs the need statement, the consequence record and the raw timestamps. An evaluation set should include episodes where the two verdicts disagree in both directions.

Sources

  • Request Tracker's service-level documentation: an order-of-goods example where meeting the response deadline does not mean the ticket is not overdue, and clock options Starts, IgnoreOnStatuses and ExcludeTimeOnIgnoredStatuses. Source
  • FAR 37.602 asks agencies to describe work in terms of required results rather than how it is done or the hours provided. Source
  • FAR 37.603: performance standards set the level required to meet contract requirements, and agencies evaluate proposed standards against agency needs. Source
  • UK Procurement Act 2023 KPI guidance: section 52(1) generally requires at least three KPIs unless performance could not appropriately be assessed by KPIs. Source
  • The Sourcing Playbook asks that KPIs align with the intended benefits to be realised during contract delivery. Source
  • UK risk allocation guidance: KPIs should be easily measurable and objective, with suppliers held accountable only for results they can influence. Source
  • GAO (December 2005) found Defense Department award fees generally not tied to acquisition outcomes and no data or measures to evaluate award and incentive fees. Source

Tell us what data you need

Share scope, volume, format, timing and licensing needs. Datasets are sourced on request.

See if your company qualifies