When you operate a freight visibility platform that processes millions of shipments, you accumulate a dataset that most individual logistics teams will never see. Not because the data does not exist - it does, in carrier systems and TMS databases across the industry - but because it is fragmented, inconsistent, and rarely analyzed at scale. We looked at two million domestic shipments tracked through our platform over 14 months to understand where transit time variability actually comes from. Some of what we found confirmed conventional wisdom. Some of it did not.
The Basic Numbers
Across the full dataset, the average deviation from planned transit time was 0.8 days for truckload shipments and 1.4 days for LTL. Those numbers sound manageable until you look at the distribution. The average includes a large population of shipments that ran exactly on time and a long tail of shipments that ran significantly late. The median deviation for truckload was 0.3 days. But the 90th percentile deviation was 2.1 days, and the 95th percentile was 3.4 days. One in twenty shipments was running more than three days off plan.
For supply chain planning purposes, the tail matters more than the average. If your safety stock model is calibrated to the average transit time deviation, you are underprotected against the 10 to 15% of shipments that will cause real problems. The planning assumption most companies use - a single transit time number per lane, sometimes with a buffer day added - does not reflect the actual distribution of outcomes in the data.
Day of Week Is a Bigger Factor Than Most Realize
One of the clearest patterns in the data was day-of-week variability. Shipments tendered on Monday or Tuesday consistently outperformed shipments tendered Thursday through Saturday, with Friday the worst-performing tender day across almost all lanes and carriers in our dataset.
Friday tenders ran an average of 0.7 days later than Monday tenders on identical lanes with the same carrier. That gap sounds small but it compounds. A shipper tendering 100 loads per week, with 25% of those going out on Thursdays and Fridays, is adding meaningful planned transit variance to a quarter of their freight without realizing it is a controllable variable.
The mechanism is intuitive: drivers prefer not to be on the road over the weekend, and carriers manage capacity to reflect that preference. Friday pickups often sit until Monday. Transit time expectations should reflect this reality, but most TMS lane parameters do not distinguish by day of week. The planned transit is the same Monday as Friday, which means Friday loads are systematically late against a benchmark that was never realistic for that tender day.
Lane Distance and Carrier Type Interactions
Short lanes under 300 miles showed the highest absolute variability in transit times, which surprised us given that the planning assumptions for short haul tend to be confident (next day, same day, two days). The variance on short lanes stems from a different cause than long lanes: short haul loads are more likely to be handled by regional carriers with fewer trucks, be used as fill loads alongside other freight, and experience consolidation delays at terminals.
Long-haul lanes over 1,500 miles showed higher average deviation but lower percentage variance relative to planned transit time. A load that takes 4.5 days when you planned 4 is a 12% miss. A load that takes 1.5 days when you planned 1 is a 50% miss even though it only added half a day. Short haul variability is often underweighted in planning models because the absolute delays are small, but the operational impact of a same-day delivery that slips to next day is often greater than a coast-to-coast load that arrives two days late.
Weather Events Are Overestimated as a Variance Driver
This one runs counter to the most common narrative. When logistics managers explain why a shipment was late, weather comes up frequently. When we categorized delay causes in our dataset, weather and physical disruptions accounted for only 11% of significant transit time deviations. That is material, but it is significantly smaller than carrier capacity issues (29%), operational execution failures at origin (23%), and route consolidation or equipment changes (18%).
Weather disruptions are memorable and highly visible when they occur. But they are seasonal, geographically concentrated, and often foreseeable enough to allow re-routing or pre-positioning. The more diffuse and harder-to-manage causes - carrier execution variability, driver availability, equipment availability at origin - are contributing far more to aggregate transit time variance than weather, but they get less attention because each individual failure is small and not memorable.
Carrier Consistency Varies More Than Carrier Performance
The most counterintuitive finding: two carriers can have identical on-time delivery rates but very different transit time variance profiles. Carrier A might be on time 85% of the time, with the other 15% typically running one to two days late. Carrier B might also be on time 85% of the time, but with a more erratic miss pattern that includes occasional five and six day delays.
For most planning purposes, Carrier B is worse to work with than Carrier A even though their OTD scores are identical. A one-day-late shipment is usually manageable. A five-day-late shipment creates a supply chain event. Standard carrier scorecards that report only on-time percentage miss this distinction entirely. Transit time variability - specifically the variance and 95th percentile of deviation - is a metric worth tracking separately from on-time rate, and the data shows it differentiates carrier quality in meaningful ways.
Two million shipments do not answer every question, but they answer the right ones more honestly than intuition and anecdote. The patterns in aggregate freight data are consistent enough to be actionable, and the companies that use them as planning inputs make better decisions than the ones still building safety stock on average transit time assumptions.