Port of Cardiff — Period 2 2026-27 (26 April – 23 May) report
Port of Cardiff received 29 monitored inbound services in Period 2 2026-27 — a small monitored sample. 93% of 28 completed arrivals were within 15 minutes of schedule (Wilson 90% interval 81%–98%), against a national freight benchmark of 90% over 12,042 completed services. The cancellation rate was 3% across 29 scheduled paths. PRELIMINARY at T+2; FINAL re-publication at T+35 after Network Rail's batch correction.
- Edition
- v1.0 (industry review edition)
- Last updated
- 28 May 2026
- Next update
- Period 3 2026-27 — late June 2026
Headline facts and figures
Services monitored
29
A small monitored sample of 29 services; the wide interval reflects the sample size; across the three Port of Cardiff TIPLOCs (CARDFTS, CARDGBR, CARDRP); Period 2 2026-27 (26 Apr – 23 May 2026); NROD-derived arrivals data; baseline 35 services in Period 1 2026-27; queried gauge_intelligence_v2 at 2026-05-28
Within 15 minutes
93%
28 completed services arriving within 15 minutes of schedule (Wilson 90% interval 81%–98%; a small monitored sample of 28 services — the wide interval reflects the sample size); baseline 97% [87%–99%] over 31 completed services in Period 1 2026-27 — ranges overlap; national freight benchmark 90% over 12,042 completed services in Period 2 2026-27; queried gauge_intelligence_v2 at 2026-05-28
Cancellation rate
3%
1 cancellation across 29 scheduled paths; down from 11% (4 of 35) in Period 1 2026-27 — improvement on the period, though small sample throughout; queried gauge_intelligence_v2 at 2026-05-28
Busiest operator
DB Cargo UK
28 of 28 completed arrivals (100%); single-operator steel and scrap flow; aggregated by operator from TRUST toc_id; Period 2 2026-27; flat on 100% share in Period 1 2026-27; queried gauge_intelligence_v2 at 2026-05-28
Context
ORR publishes aggregates only; this page publishes destination-attributed port-level data. Material to yard planning, terminal allocation, and port-side performance reviews.
What you can do with it
- Track arrivals reliability at a port, attributed by destination TIPLOC.
- Compare operator-by-operator reliability on services inbound to a single port.
- Cite the methodology in port operations and corridor reviews.
Who can use it
Port operations leaders · Terminal allocators · Rail journalists
Across Period 2 2026-27, roughly 0 inbound trains a day arrived at Port of Cardiff more than 15 minutes late, about 1 in 23 of 24 completed arrivals over the 28-day period, 1 late arrivals in total. This is a small monitored sample of 29 scheduled services on a single-operator steel and scrap flow; the 90% Wilson interval (81%–98%) is wide and the movement from 97% in Period 1 sits within overlapping ranges. All 24 completed arrivals were operated by DB Cargo UK inbound along the South West & Wales to Midlands corridor. Figures are drawn from Network Rail's NROD/TRUST feed. Cancellations are reported separately below.
Operators at Cardiff-inbound
DB Cargo UK ran the most cardiff-inbound services in Period 2 2026-27 (24 of 24 completed arrivals). The full per-operator reliability comparison is deferred while the v1.0 industry review edition corrections inbox remains open.
Publishing a per-operator on-time comparison for the busiest freight operators at a single destination before the launch-cohort corrections cycle closes would treat provisional attribution as settled. The table returns at v2.0 general availability. The league-table methodology documents the comparison framework; the release cadence page sets out the four exit criteria for v2.0.
Source: Gauge Intelligence NROD-derived journey data, Period 2 2026-27. Single-destination headline figures (completed arrivals, aggregate within-15-minute rate, cancellations) continue to appear in the panels above; the per-operator breakdown is the panel deferred here.
Outbound destinations — services leaving Port of Cardiff
No outbound services from Port of Cardiff reached a registered inland terminal in Period 2 2026-27.
What the 15-minute threshold hides
Half of Period 2 2026-27 services arrived early — the median was 10 min ahead of timetable. The long tail the A2F binary misses: 5% of services ran +11 min or more late, and 1% arrived over +25 min late. The 15-minute A2F threshold sits at the 96th percentile — 4% of services exceeded it.
Median arrival delay is 4 minutes later than Period 1 2026-27 (median was -14 min then, -10 min now).
| Percentile | Arrival delay |
|---|---|
| 50th percentile (median) — A2F threshold sits at the 96th percentile | -10 min |
| 75th percentile — A2F threshold sits at the 96th percentile | -1 min |
| 85th percentile — A2F threshold sits at the 96th percentile | +1 min |
| 90th percentile — A2F threshold sits at the 96th percentile | +3 min |
| 95th percentile — A2F threshold sits at the 96th percentile | +11 min |
| 99th percentile | +25 min |
24 services with confirmed terminal arrival in Period 2 2026-27. A2F is binary at 15 minutes — this distribution surfaces what the headline figure does not show. National 99th percentile across all freight operators: 95 min.
Baseline (Period 1 2026-27): median -14 min, p99 +130 min. Period-on-period median shift +4 min.
ORR's Freight rail usage and performance release publishes by Network Rail region and route, not by operator. This page publishes by operator and corridor.
Daily on-time arrivals across Period 2 2026-27
Daily on-time arrivals are down 2.0 percentage points on Period 1 2026-27 (98% then, 96% now).
Range 67%–100% across 28 days. Weekend mean 100% vs weekday 95% (+4.8 percentage points). Mean 96%. (15 of 28 days had recorded services.)
Period 2 2026-27 daily A2F averaged 100% across 15 observations. One or more observations fell outside the expected range (13 day(s) beyond the expected range; 13 run(s) of two-of-three days off-mean; 8 run(s) of four-of-five days off-mean), indicating an unusual event rather than ordinary noise.
Mean punctuality in Period 1 2026-27 was 98%. The current period is down 2.0 percentage points on Period 1 2026-27.
Source: Gauge Intelligence NROD-derived journey data, Period 2 2026-27. A service is counted on time when terminal arrival falls within 15 minutes of schedule. Days with no scheduled services (typically Sundays) render as gaps — a missing bar means no measurement, not a value at zero. Red bars flag days that triggered a statistical signal — see anomaly-detection methodology.
ORR's Freight rail usage and performance release publishes by Network Rail region and route, not by operator. This page publishes by operator and corridor.
Cancellation rate
In Period 2 2026-27, 10% of scheduled services were cancelled (3 of 29) — 6.3pp above the national mean of 4% across all freight operators; FOC traction, crew, and terminal decisions accounted for the largest share.
Cancellations are down 4.00 percentage points on Period 1 2026-27 (14% then, 10% now). The largest shifts in the responsible-party split are reported in the table below.
| Scheduled or activated paths | 29 |
| Cancelled (post-activation + pre-activation) | 3 |
| Cancellation rate | 10% 4%–23% |
| National cancellation rate (all freight) | 4% |
Attribution
| Responsible party | Cancellations | Share of cancelled |
|---|---|---|
| Network Rail infrastructure and possessions | 0 | 0% |
| FOC traction, crew, and terminal decisions | 3 | 100% |
| Third party | 0 | 0% |
| Unattributed | 0 | 0% |
Cancellation rate in Period 1 2026-27 was 14% across 35 scheduled paths. The current period is down 4.00 percentage points on Period 1 2026-27.
Source: NROD/TRUST activation feed and BPLAN scheduled paths, Period 2 2026-27. Cancellation rate = (post-activation + pre-activation cancellations) ÷ scheduled or activated paths. Pre-activation cancellations are TRUST CANCEL messages on schedules that never activated; post-activation cancellations have an activation and a CANCEL event. The bracketed range beside the headline rate is the 90% confidence range on the cancellation proportion given the sample size. Cause attribution uses the responsible-party mapping of the recorded cancellation reason code; see delay-attribution methodology.
What this page does not measure
Pre-activation cancellations fall outside the dataset: TRUST requires an activation event before it records a journey, so services cancelled before activation produce no TRUST record. VSTPs are excluded from the path-utilisation denominator because no Working Timetable row exists against which to measure them. Journeys without a confirmed terminal arrival are excluded from the delay distribution. The 15-minute A2F threshold is a regulatory artefact, not a natural cut-off — the "What the 15-minute threshold hides" panel surfaces the distribution behind that line.
Commercial licence
Four derivations are released to commercial-licence subscribers rather than the public archive. Each one reconciles TRUST source-data against Network Rail's Historic Delay Attribution record independently, not republished from public statistics.
- Schedule 8 net position
- Why late — delay attribution
- Schedule 4 possession exposure
- Path utilisation — corridor breakdown
Available under commercial licence. Contact [email protected] for subscription terms.
Methodology & sources
All figures derive from Network Rail's NROD TRUST feed, processed via the Gauge Intelligence ingest pipeline. Period definitions, A2F (Arrival to Fifteen), the Wilson 90% interval, daily anomaly rules, cross-view reconciliation, and delay attribution are documented at:
Data window — period boundaries and the data-as-of cutoff
League tables — cross-view reconciliation across the four analytical views
Anomaly detection — daily A2F SPC rules
Delay attribution — DAPR cause-code breakdown
Forecasting — prediction intervals and possession-exposure estimates
Revisions
No revisions to date.
A revision restates a published figure because the upstream data changed, typically a Network Rail delay-attribution refresh. The minor tier is corrected in the next release; cells affected carry an (r) flag. The intermediate tier triggers immediate amendment with an (r) flag. The wholesale tier is flagged at the top of the page and notified directly to operators and rail journalists subscribed to revision alerts.
Corrections
No corrections to date.
A correction fixes an error in the published output: an arithmetic slip, a typo, a mis-rendered chart. Revisions are different; they restate a figure because the upstream data changed. The distinction matters because corrections never silently rewrite a published number.