Track & Trace pilot evaluation: coverage, latency and delivery estimates

Agree the scope and observation boundary

CEP Insight proposes a pilot for a named origin–destination route, carrier product and operating period. State the shipment-selection rule, participating systems, authorized data, expected milestones and the review cutoff before examining results. Keep service changes and transport directions identifiable.

Define which milestones should already have occurred by the cutoff and why. Do not count future milestones as missing; do not remove an overdue expected milestone simply because its record is absent. Keep completed shipments, still-open shipments and returns visible. A small or selected sample supports conclusions only about its stated scope.

Separate source definitions from the proposed measures

Selected provisions of GS1 EPCIS 2.0 distinguish eventTime from repository recordTime; neither is automatically the time a reader’s application first received the event. FedEx documents limits on estimated-delivery-window availability. DHL Unified documentation says optional fields may be absent when undefined.

These published definitions explain why timing and availability need separate review. They do not prescribe the metrics below or establish carrier performance. All measures in the table are CEP Insight proposals. Agree acceptable thresholds with the parties for the intended decision; this article supplies no universal pass mark.

Keep denominators and unavailable records visible

For event coverage, compare valid, attributable records with the defined expected milestones already due. Count a duplicate once and retain the original status and any documented correction. Report unknown or conflicting mappings separately. For latency, use a stated pair of timestamps: first receipt by the reviewing application minus the event time, with time zones aligned.

Report how many records have usable timestamp pairs. A missing event has unknown latency, not zero; a negative difference requires clock or interpretation review. If percentiles are useful, state the sample count and calculation method. A frequent polling schedule can affect observed receipt delay and is not itself a measure of physical transit speed.

Evaluate estimates at a fixed prediction point

Choose the prediction version before observing delivery, such as the latest estimate available at an agreed point before dispatch for delivery. Freeze its issue time, predicted date or window and time zone. Define the matching actual-delivery event and tolerance. Do not substitute a later corrected estimate when scoring the earlier prediction.

Illustrative arithmetic only: among 50 completed eligible shipments, 20 have the required frozen estimate and comparable delivery evidence; 18 are within the agreed window. Estimate availability is 20/50 = 40%, while the hit rate among evaluable estimates is 18/20 = 90%. Report both and explain the other 30 records. These invented counts show a denominator effect; they are not customer results or a provider benchmark.

Separate estimate availability from evaluability: a supplied estimate can still lack comparable delivery evidence.

Review exceptions and record an acceptance decision

Separate time to acknowledge an exception, time to the first useful action and time to close it. State the starting event and what counts as acknowledgement, useful action and closure. Record cases still open at the cutoff instead of treating them as resolved or silently removing them. A closed case can mean a documented disposition, not necessarily successful delivery.

Our proposed acceptance record states sample and exclusions, metric definitions, counts, unresolved gaps, agreed thresholds, responsible reviewer and decision date. A conditional decision should identify remaining evidence and the next review. Retain unknown statuses, duplicate and corrected messages, retries and changed identifiers as test cases. Provider documentation or sandbox success alone does not establish production coverage.

Sources and limits of this evaluation framework

Sources were reviewed on 12 October 2026: selected timing provisions of archived GS1 EPCIS 2.0, selected FedEx estimated-window provisions and selected DHL Unified access, timing and optional-field guidance. This is not a full standards review or an assessment of every service.

No live shipment dataset, private API or supplier trial was examined. The formulas and illustrative counts are CEP Insight’s editorial method, not findings from the cited organizations. Comparisons need the same route, service boundary, observation period and sample rules, with differences disclosed. Use the existing discussion and source-correction channels to supply verifiable evidence.