Ocean Freight Is More Likely to Arrive Late: Score Carriers by Delay Distribution, Not Average Transit

An ocean service advertised at 16 days can still be the wrong choice if its containers routinely arrive in 22 days and its worst shipments take 35. The published transit and the average actual transit describe the center of the service. Inventory failures, customer penalties, emergency airfreight, and domestic spot-market exposure usually live in the tail.

That distinction matters now. On some Asia–North America lanes, late arrival has become more likely than on-time delivery. Procurement teams should stop ranking carriers by nominal transit and average performance alone. The better question is: what does the complete delay distribution look like, and can the business absorb its downside?

Carrier choice can outweigh lane choice

FreightWaves reports that execution and reliability—not aggregate capacity—are the defining ocean risks on exposed Asia–North America services. Its cited September outlook found that carrier selection can be a stronger predictor of a good outcome than lane selection because performance varies widely among carriers serving the same corridor. It also identifies a container arriving 30 days late as a particularly severe downstream cost driver.

The capacity headline can be misleading. Supply Chain Dive reports that scheduled capacity on Asia–U.S. East Coast trades grew 46% in the first half of 2026 versus the same period in 2019, while blank sailings grew 215%. On the West Coast, scheduled capacity rose 16% while blank sailings increased 62%.

More physical vessel capacity therefore does not guarantee more usable capacity. When a sailing disappears, its cargo competes for space on later departures. Rollovers and bunching then push volatility into terminals, drayage, rail, and truckload operations.

Replace one average with five measures

Averages compress reliable shipments and extreme misses into one number. A carrier with consistent three-day delays may share the same average as one that is usually on time but occasionally three weeks late. Those services create very different inventory and customer risks.

Build the scorecard at carrier, service, origin, destination, and equipment level. A carrier-wide score is too broad; performance on Shanghai–Los Angeles does not prove performance on Ningbo–Savannah. At minimum, calculate these five measures over a consistent rolling period:

• Median end-to-end transit: the midpoint of actual booking-to-availability time. This shows the normal outcome without letting a few severe failures dominate.
• 90th-percentile transit: the time within which 90% of shipments complete. This exposes the delay buffer needed for high-confidence planning.
• Rollover rate: the share of confirmed containers moved to a later vessel, measured at origin and at transshipment ports.
• Blank-sailing exposure: the share of booked or forecast volume affected by canceled departures, including how early the carrier communicated the cancellation.
• Recovery performance: the time from a missed milestone to rebooking, revised estimated arrival, and final delivery. Disruption is costly; slow, opaque recovery makes it worse.

Use end-to-end milestones rather than vessel arrival alone. Booking confirmation, empty release, gate-in, loaded departure, transshipment, discharge, customs release, container availability, and final delivery reveal where variability enters the move. A vessel can reach port close to schedule while the container still misses its promise because discharge or availability takes longer.

Make percentile math operational

For each lane and carrier, order actual transit times from fastest to slowest. The median is the middle observation. The 90th percentile is the point below which 90% of observations fall. Report both transit and delay against the contractual or planned promise.

Sample size and recency matter. Display the shipment count beside every result, and avoid making allocation changes from a handful of moves. Use a rolling 90- or 180-day window, but weight the most recent four to eight weeks when alliances, rotations, ports, or operating conditions change. Separate planned exceptions—such as shipper-requested holds—from carrier-controlled misses.

Do not blend every delay into one composite score too early. Decision-makers need to see whether the risk comes from unreliable departure, slow transit, transshipment failure, or destination recovery. A weighted index is useful for ranking, but the component measures make the ranking actionable.

Match risk to the shipment, not a universal winner

The carrier with the lowest 90th-percentile delay is not automatically best for every load. Apply service risk to the economics of the cargo.

For low-margin, stable-demand goods with ample inventory, a slower or more variable service may be acceptable if savings exceed expected disruption cost. For a product launch, seasonal item, production component, or order carrying a late-delivery penalty, the right comparison is rate plus risk-adjusted cost.

That cost should include inventory buffer, stockout exposure, demurrage and detention, recovery labor, premium drayage or truckload, customer penalties, and emergency mode upgrades. The 90th percentile should drive the buffer for service-critical freight; the median can support routine planning. If the gap between them widens, the service is becoming less predictable even when its average is unchanged.

Use allocation tiers rather than declaring one carrier the permanent winner. Assign critical cargo to the most predictable service, flexible volume to the best risk-adjusted price, and a controlled share to alternates so their real performance remains measurable. Establish triggers that shift allocation when rollover rate, blank-sailing exposure, or 90th-percentile transit breaches a threshold for two reporting periods.

Feed booking actuals back into procurement

A scorecard built once for an annual bid will decay quickly. Ocean networks change through blank sailings, vessel redeployment, alliance adjustments, port congestion, and seasonal demand. Every completed booking should update the evidence used for the next tender.

CXTMS can join bookings, milestones, planned transit, actual transit, exceptions, costs, and customer outcomes in one lane-level record. Procurement can compare a quoted service with observed median and tail performance, while operations can see which failures created downstream spot buys or missed promises.

The goal is not to predict every late container. It is to stop treating variability as an invisible surprise. When teams buy a delay distribution instead of a brochure transit time, carrier selection becomes an inventory and margin decision—not merely a rate comparison.

Request a CXTMS demo to build carrier scorecards from booking-level actuals and route each shipment according to the reliability its business promise requires.