Build a Carrier Delivery-Performance Scorecard

To build a carrier delivery-performance scorecard, compare each shipment's documented outcome with the service requirement that was actually accepted. Define the event, appointment version, time zone and eligible shipment group before calculating a rate. Show missing evidence and disputed responsibility separately. An arrival outside an agreed window establishes an appointment exception; it does not, by itself, establish who caused it.

This guide helps shipper and broker operations teams review completed work and choose a specific improvement. The output is a reproducible service record with an exception owner, rather than an unexplained carrier grade. The definitions and worksheet below are editorial recommendations that your team should adapt to its actual agreements.

Choose one service question before collecting numbers

Start with a question such as: “For the deliveries due at this facility during the review week, how often did the truck complete the agreed check-in event within the accepted window?” That is narrower and more useful than “Was the carrier reliable?”

Select one event for each measure:

• Appointment adherence: the agreed arrival or check-in event compared with its accepted window.
• Unloading completion: the agreed release or completion event compared with its own requirement.
• Document return: receipt of the required document compared with the agreed submission deadline.
• Exception communication: the required notice compared with the agreed trigger and timing.

Do not use one timestamp for all four. A proof-of-delivery document may establish that a handoff occurred without recording gate check-in. A location ping may place a vehicle near the facility without establishing completion of the receiving procedure.

For the selected measure, write what counts as evidence and what each timestamp means. Specify whether early arrival is acceptable, whether the window endpoints are included, and whether a grace period exists in the applicable terms. Do not invent a grace period after seeing which shipments would otherwise fail.

The facility-fit worksheet helps establish the receiving instructions before work begins. This scorecard starts with those accepted instructions and examines what happened afterward.

Preserve the appointment that was actually accepted

Save the original appointment reference, date, start and end time, time zone, and acceptance evidence. Add every material revision with its request time, authorizing parties and effective scope. Keep the prior version available.

A requested change is not an accepted replacement. If the receiver asks for 14:00 but the carrier's response is unknown, do not silently replace an accepted 10:00 appointment in the scorecard. Mark the change unresolved and have the appointment owner confirm which requirement applied.

Agree how legitimate revisions affect the measurement before reporting results. One useful approach is to measure adherence against the accepted operative appointment while separately reporting how many appointments changed, who requested the changes and when they were accepted. This keeps a revised service plan visible without disguising schedule instability.

If a late delivery is retrospectively relabeled with a later appointment, preserve the original exception and the later disposition. Do not rewrite history to make the score improve. A contractual waiver or commercial settlement belongs in a separate field reviewed by the authorized person.

Retain original local timestamps and their UTC offsets when available. A conversion can support comparison, but it should not erase the source time or conceal an unknown zone. For a timestamp without enough information to compare it reliably, leave the outcome unconfirmed until the record owner clarifies it.

Freeze the eligible group and show evidence coverage

Define the reporting period, facility or lane, service type and carrier entity. State whether the unit being counted is a shipment, delivery stop or delivery attempt. One shipment with three stops must not become one observation in one carrier's score and three in another's.

Build the eligible list from the accepted service commitments, not only from records with a successful delivery scan. Otherwise missing records and unfinished work can disappear before anyone calculates the percentage.

Record exclusions explicitly. A duplicate record is not another delivery. A canceled movement may belong in a separate cancellation measure, provided the cancellation rule and timing are stated. A delivery still outstanding at the review cutoff remains visible in the due-work group; explain its status instead of treating it as completed or removing it because it lacks a delivery record.

For a completed-delivery appointment review, use mutually exclusive outcome states:

1. Requirement met: the accepted requirement and sufficient event evidence establish adherence.
2. Requirement not met: sufficient evidence establishes a result outside the accepted requirement.
3. Unknown: missing or conflicting evidence prevents classification.

Keep a separate field for why the evidence is unknown. Missing event time, unresolved appointment version and conflicting source records need different follow-up. If your report is restricted to completed deliveries, also display the count of due deliveries still open so readers can see what the completed-work measure leaves out.

Publish the underlying counts alongside two clearly named calculations:

• Known-outcome adherence: requirements met divided by deliveries with a known outcome.
• Evidence coverage: deliveries with a known outcome divided by all eligible deliveries in that measure.

Neither calculation turns an unknown into a pass. If coverage changes between periods, examine why before attributing a change in the adherence rate to better or worse service.

Worked example: 20 deliveries, two missing check-in records

Fictional teaching example. All shipments, records, confirmations and operational conclusions below are invented. They are not USDOTData customer results or an industry benchmark.

A shipper reviews 20 single-stop deliveries due at one facility during September 21–25, 2026. All 20 were completed by the September 28 review cutoff; there are no cancellations, duplicates, extra stops or deliveries still open. The same performing carrier and service type apply to all 20. Completed handoffs are documented, but two records lack the separate gate check-in event required by this measure.

For this example only, each accepted appointment requires completed gate check-in from 09:00 through 10:00 inclusive, local Eastern daylight time, UTC−04:00, on its scheduled date. There is no grace period. The appointment references and acceptance evidence are available for all 20; none was revised.

The event record divides the complete eligible set as follows:

• D01–D16: 16 documented check-ins fall within the applicable accepted window. Outcome: requirement met.
• D17: check-in is 10:18 on its scheduled day, 18 minutes after the window closes. Outcome: requirement not met.
• D18: check-in is 10:42 on its scheduled day, 42 minutes after the window closes. Outcome: requirement not met.
• D19–D20: each has a completion record but no usable gate check-in time. Outcome: unknown for appointment adherence.

The scorecard calculation is:

Eligible completed deliveries:       20
Requirement met:                     16
Requirement not met:                  2
Unknown appointment outcome:          2
Known outcomes:                  16 + 2 = 18

Known-outcome adherence:     16 / 18 = 88.9%
Evidence coverage:           18 / 20 = 90.0%
Confirmed on-time share
of the whole eligible group: 16 / 20 = 80.0%

The 80.0% figure is a confirmed share, not a final full-group adherence rate: the two unknowns have not been classified. If both eventually fail the requirement, final adherence will be 16/20, or 80.0%. If both meet it, it will be 18/20, or 90.0%. Thus 80.0%–90.0% is the possible full-group range given the existing records, assuming the 18 known classifications remain unchanged. It is not a statistical confidence interval or a forecast.

Do not publish “90% on time” by subtracting the two known late deliveries from 20. That calculation silently treats both missing records as successes. Do not publish “88.9%” alone either: it hides the fact that 10% of the eligible group's appointment outcomes remain unknown.

Keep responsibility separate from the outcome

In the fictional exception review, the parties document a dispatch-planning issue for D17 and an origin shipper release delay for D18. The accepted destination appointment did not change for either. Both remain late in the appointment-adherence measure, while their operational follow-up owners differ.

For D17, the carrier operations owner reviews the dispatch plan. For D18, the shipper operations owner reviews the origin release process. The receiving-record owner investigates D19 and D20; the scorecard reviewer owns updating the report when sufficient evidence arrives.

This does not establish legal liability or a right to deduct charges. It is an operational allocation supported by the invented evidence. Do not rename 16/18 “carrier-caused performance,” or remove D18 without describing an entirely separate adjusted measure and its eligibility rules.

The immediate result is three useful actions: review one dispatch issue, review one release issue, and recover two missing event records. A blended letter grade would obscure which team can improve which part of the service.

Use the scorecard to select an action, not announce a universal ranking

Keep the measure connected to the decision it supports. If the problem is missing check-in evidence, fix the recording and handoff process. If repeated exceptions concern one facility's release times, examine that facility before applying a broad carrier penalty. If responsibility is disputed, preserve the competing explanations and assign review instead of choosing the most convenient one.

Before comparing carriers or weeks, check that event definitions, service commitments, lane or facility mix, units, exclusions and evidence coverage are comparable. Twenty routine deliveries at one receiver and twenty time-critical deliveries at several constrained sites are different operating groups. A matching denominator alone does not make them equivalent.

Small groups are sensitive to individual events. In a fixed group of 20 deliveries, one classification changes the full-group percentage by five percentage points. This arithmetic is a reason to show counts and context, not a universal minimum sample rule. A serious individual service failure may require action immediately even when there are too few comparable observations to support a general ranking.

Keep quantity and condition questions in their own measures. The shipment-quantity reconciliation guide explains why matching cartons, pallets and products requires defined units. An on-time arrival does not establish correct quantity, undamaged goods or invoice accuracy.

A service scorecard is not an FMCSA safety assessment

FMCSA describes the SAFER Company Snapshot as containing identification, fleet and cargo information, inspection and out-of-service summaries, crashes and a safety rating when one exists. These are different records from your customer's appointments and delivery-event evidence.

FMCSA's Safety Measurement System overview explains that SMS uses inspection, crash and investigation information to prioritize safety interventions. Your on-time percentage is not an SMS result, an official safety rating, or proof of current authority or insurance.

Keep service review alongside the separate carrier-verification evidence record. A strong delivery history does not complete today's identity and shipment checks; a documented service exception does not itself establish unsafe operation.

Copy this service-performance record

Use one report definition and repeat the event record for every eligible delivery. Reference controlled source files rather than placing private shipment documents in a public report.

REPORT DEFINITION
Question / measure / report version:
Review period / cutoff / time zones:
Carrier entity / performing carrier identifier:
Facility or lane / service type:
Counting unit: shipment / stop / attempt
Eligibility rule / eligible list reference:
Exclusions / reasons / due work still open:
Event definition / permitted evidence:
Window endpoints / early rule / grace rule:

DELIVERY EVENT
Shipment / stop / scheduled date:
Original appointment / acceptance evidence:
Revision / request time / acceptance evidence:
Operative requirement / unresolved change:
Actual event / timestamp / zone / source:
Outcome: met / not met / unknown
Unknown reason / conflicting evidence:
Exception cause / evidence / dispute state:
Operational action owner / next review:

REPORT AND ACTION
Met / not met / unknown / eligible counts:
Known-outcome rate / denominator:
Evidence coverage / denominator:
Open-work count / scope of any adjusted measure:
Action / owner / due date / closure evidence:
Reviewer / report time / retained prior version:

Before sharing the scorecard, check that every eligible delivery has one outcome; the counts add to the stated denominator; appointment revisions remain traceable; unknowns have owners; and responsibility has not been inferred solely from a late timestamp. Read the report title and percentage label together: they should tell the reader exactly which event and group were measured.

Frequently asked questions

Which deliveries belong in the denominator?

Use the written eligibility rule for the measure: period, carrier, service, facility or lane, and counting unit. Reconcile against accepted commitments so missing events do not disappear. If the measure covers completed work only, also disclose due deliveries still open and keep cancellation and duplicate rules explicit.

What if the appointment changed?

Preserve the original and revised requirements with acceptance evidence. Apply the predeclared revision policy and show schedule changes separately. A request without acceptance is unresolved; a retrospective disposition should not erase the original event or exception.

Who owns a missing timestamp?

Assign evidence recovery to the owner of the relevant event record, such as the receiving operation for its gate check-in record. The scorecard reviewer owns classification and the report revision. Missing evidence does not automatically establish that the carrier arrived late or failed to provide it.

Can a shipper-caused delay count against a carrier?

It can remain an appointment exception in an unadjusted service-outcome measure, while the cause and action owner are recorded separately. Do not describe that measure as carrier fault. A separate responsibility-adjusted view needs explicit evidence, rules and denominators, and it should remain traceable to the original results.

Is delivery performance an FMCSA safety score?

No. This worksheet compares agreed service events with shipment evidence. It does not calculate an official safety rating or SMS result, and it cannot replace current carrier identity, authority, insurance or safety-record checks.

How many observations support a conclusion?

There is no universal number supplied by this worksheet. Show counts, comparable scope and evidence coverage, and distinguish a decision about a specific exception from a general claim about a carrier. Do not call a short run of favorable observations proof of future reliability or a statistically established improvement.

Start with one measure and one unresolved event

Choose a completed delivery group, define its accepted event and reconcile the eligible list. Then resolve the first missing record or disputed requirement before adding more metrics. The useful outcome is a traceable action that improves the next shipment, supported by an honest account of what the current evidence can establish.

Sources reviewed October 7, 2026. John Hauler is USDOTData's editorial persona. USDOTData is independent of FMCSA. The scorecard is an editorial workflow, the worked data are fictional, and operational classifications do not determine contractual liability.