Global schedule reliability came in at 56.4% for July, down 6.1 percentage points on June and 8.8 points on July last year. It is the lowest figure of 2026 so far. What makes it worse is the second number: ships that arrived late arrived 6.06 days late on average, up more than half a day in a month.
Both moving together is the bit that matters. Reliability falling while average delay holds steady means more ships are missing the window but only just. Reliability falling while delay lengthens means the misses are getting more severe. That is the second kind of month.
Two failures, one statistic
Reliability is a binary: a vessel either arrives inside its published window or it does not. It tells you the probability of a miss and nothing about the consequence.
Average delay is the consequence. At 6.06 days, a missed window on a monthly-cycle programme is not a rescheduling exercise, it is a missed cycle. Warehouse slots, inland trucking, retail set dates — all of them are built on a lead time that assumed the ship was roughly when it said it would be.
Run the two together and July looks like this: 44 bookings in every 100 landed outside the window, and each of those landed the better part of a week late. Anyone carrying a five-day buffer was covered for about half of their late arrivals.
The spread between carriers is now the real story
Maersk came in at 73.7%, the only carrier above 70. Hapag-Lloyd took second at 69.3%, alone in the sixties. Five carriers sat between 50% and 60%. Wan Hai finished last at 29.8%.
Forty-four points separate the top from the bottom. That is not a quality gradient, it is two different products being sold under the same name and, frequently, at very similar prices.
MSC lost 11.9 points month on month, the steepest fall of the thirteen. And not one carrier in the group improved on June — which tells you this was systemic rather than operator-specific. The North Asian congestion that has ships waiting ten days to berth at Shanghai and Ningbo does not care whose funnel is on the ship.
That systemic point cuts both ways, though, and here is where I would push back on the obvious reading. If every carrier is exposed to the same congestion and one of them still delivers 73.7%, the congestion is not the explanation for the bottom of the table. It is the test that separates a network with recovery buffers built into it from one that runs everything hot.
What to do with a number like this
Stop treating carrier reliability as a tie-breaker between otherwise equal quotes and start treating it as a cost line.
Price the difference. On a lane where a miss costs you a week of inventory, moving from a 50% carrier to a 70% carrier removes one missed cycle in five. Work out what that cycle is worth to your customer and compare it against the rate premium — quite often the premium is smaller.
Buffer against the delay figure rather than the reliability figure. Six days is the average for late arrivals right now. If your planning buffer is three, you are covered for less than half of the misses you will actually get, and the ones that break through are the ones your customer hears about.
Check the lane, not the carrier. Published reliability is a global average across a carrier's whole network. A carrier at 55% overall can be well above that on one trade and far below on another. The carrier pages break performance down by service, and the live sailing schedules show which strings are currently holding their windows rather than which ones did last quarter.
One caveat worth stating: these are July figures published in late August, and July was the peak of the typhoon disruption in North Asia. August will very likely read worse before it reads better. Do not rebuild your carrier scorecard on a single month at the bottom of a cycle.
What would change my mind on the structural read is Maersk's number. If the leader slips into the sixties while the tail stays where it is, this stops being about network design and starts being about a market nobody can operate reliably.
