Skip to content
    Supply Chain Technology

    How to measure transit time variability across your lanes

    Oct 2, 20266 min read

    Schedule reliability fell below 50% this summer. How to measure transit time variability by lane and turn a planned transit time into a number your plan can trust.

    A planner quotes a customer October 14. The number behind that date is a 32-day transit time that has been sitting in the ERP since someone set the lane up. The last twenty sailings on that lane actually ran between 29 and 46 days. Nobody lied to the customer. The date was never a forecast in the first place: it was a number typed once and inherited ever since.

    That gap got noticeably wider this year. Sea-Intelligence put global schedule reliability at 49.9% in August 2026, the lowest reading since September 2022, and the average delay for vessels that did arrive late climbed to 6.81 days, the worst since March 2022. Congestion across Asian ports is expected to take months to work through. When roughly half of all vessel arrivals miss the published schedule, a single planned transit time stops being a plan and starts being a coin flip with a date attached.

    The fix is not a better guess. It is measuring what your lanes actually do, and planning to the spread instead of the average.

    Your planned transit time is doing more work than you think

    That one number rarely sits in one place. It is the lead time in the ERP that drives when purchase orders get released. It is an input to the safety stock calculation. It is the basis of the delivery date your customer service team commits to. It sets the week production expects material on site, and it tells the inbound warehouse when to roster labor.

    Every one of those inherits the same error, silently. A transit time that is four days optimistic does not produce one late shipment. It produces late orders, a safety stock number that was sized for a narrower world, a promise date that was wrong when it was made, and a receiving dock that staffed up for a Tuesday.

    This is a different problem from ETA accuracy on shipments already in flight. An ETA tells you where one container will land. The planned transit time is the number you commit with before the shipment exists. Both can be wrong independently, and teams usually only fix the first one.

    Measure the spread, not the average

    Most transit times are stored as a single figure, which forces a distribution into one cell. That hides the only thing you need to know: how wide is it.

    Pull the last 90 days of completed shipments on a lane and look at the percentiles rather than the mean. The median (P50) tells you the typical run. P85 and P95 tell you what a bad run looks like. On a lane where P50 is 32 days and P95 is 46, the planned 32-day number is not conservative or aggressive, it is simply the point at which half of your shipments are already late by construction.

    Which percentile you plan to is a business decision, not a data decision, and it should follow the consequence of being late. A lane feeding a production line that stops when material is missing deserves P95. A replenishment lane into a warehouse holding eight weeks of cover can sit near P50 and absorb the variance in stock. Making that choice explicitly, lane by lane, is most of the work.

    Then watch the spread itself over time. A lane whose median holds steady while its P95 stretches out is degrading, and a mean-only report will show you nothing until a customer calls.

    Cut the number by lane, not by carrier

    A carrier-level on-time percentage is close to useless for planning, because it averages across services that behave nothing alike. The unit that matters is the lane: origin port, destination port, service string, mode, and where relevant the equipment type.

    August made that visible at the alliance level. Gemini recorded schedule reliability of 77.8% while Premier Alliance came in at 43.2%, with MSC at 70.1% and Ocean Alliance at 48.3%. Those are not small differences in service quality, they are different planning assumptions. If your lane table carries one ocean transit time for Asia to North Europe, it is averaging a 77% service and a 43% service into a figure that describes neither.

    Cutting that finely runs into a data problem before it runs into an analysis problem. Each carrier reports milestones on its own schedule, with its own event names and its own idea of what "departed" means, and forwarder updates arrive in a different shape again. Normalizing those into one comparable event history is exactly what a command center like Orkestra is built to do, through pre-built connectors into carriers and your ERP, TMS, and WMS, so a lane distribution is something you can query rather than something you rebuild by hand each quarter. The underlying data model matters more than the dashboard here, and analytics built on normalized events is what makes the percentile view possible at all.

    Agree where the clock starts and stops

    Most disagreements about transit time are really disagreements about boundaries. Does the clock start at cargo ready, at gate-in, or at vessel departure? Does it stop at discharge, at gate-out, at customs release, or at delivery to the door?

    Port-to-port is the easiest number to source and the least useful one to plan with, because it excludes the segments where the variance actually lives: terminal dwell, customs clearance, and drayage. A lane can hold a stable port-to-port time for a year while its door-to-door time drifts by a week, and only the second number is the one your customer experiences.

    Pick door-to-door as the planning measure, keep the leg-level splits underneath it, and write the definition down. The splits are what tell you where to act when a lane degrades, and they stop the quarterly argument about whose number is right.

    One caveat on the raw data: a blank sailing or port omission does not produce a late shipment so much as a different shipment, rolled to another vessel. Flag those rather than letting them sit in the distribution as ordinary delays, or a handful of rolled containers will stretch your P95 and send you adding safety stock against an event that calls for a different response entirely.

    Re-baseline on a cadence, and let the number drift

    A transit time measured once is just a newer stale number. Set a rolling window, 90 days works for most ocean lanes, weight recent shipments more heavily, and recalculate on a schedule rather than when something goes wrong.

    Then set a trigger. If a lane's P85 moves more than a defined threshold against its own trailing baseline, that is a signal to re-plan: adjust the order release date, revisit the cover, tell customer service before they promise the old date again. Detecting that drift as it happens, rather than reconstructing it afterwards, is where alerting earns its place, and it is the same mechanism Bridgestone used when it combined data normalization with production planning and customizable alerts to take $11.6K a week out of 22 shipments.

    Re-cutting the distribution matters most right after a route changes for good, which is the first move in re-baselining a lane after a permanent reroute. It also makes the lane comparable to its peers, which is the starting point for benchmarking performance across regions without each region quietly measuring something different. The spread is also the first column in a lane risk score that ranks which lanes deserve attention before anything is late.

    What to do Monday

    Take your five highest-value lanes. For each, pull the last 90 days of completed door-to-door shipments, calculate P50, P85 and P95, and put those next to the transit time currently sitting in your planning system. The gap between the two columns is the size of the problem, and on most lanes this year it will be larger than anyone expects.

    Then decide, per lane, which percentile you are willing to plan to, and change the number. Reliability below 50% is not something any shipper can fix. Planning as though it were 90% is.

    If you want that view across carriers without assembling it by hand every quarter, book a demo and we will show you what your lane distributions look like on normalized data.

    Trusted by leading supply chain teams
    MatalcoOIA GlobalKingsdownDBWMatalcoOIA GlobalKingsdownDBW
    DLAIllyLevitt SafetySquareInstoreDLAIllyLevitt SafetySquareInstore
    TAKE THE NEXT STEP

    Action breaks down when operations are disconnected.

    Orkestra gives teams one place to see what matters, coordinate the response, and move operations forward.