.webp)
Once a container is discharged, accurately predicting its rail departure time has traditionally been nearly impossible. Uncertainty at this critical stage makes managing customer expectations and internal planning difficult. Many ocean carriers don't even attempt to predict rail departures, and those that do typically deliver highly inaccurate estimates.
With OpenTrack’s advanced predictive modeling, uncertainty is no longer inevitable. Building on the same innovative research that enabled us to deliver industry-leading rail ETA accuracy, we’ve significantly improved rail ETD predictions.
How much better is OpenTrack’s Rail ETD than the ocean carriers?
In this chart, you can see the “ETD inaccuracy” of OpenTrack versus ocean carriers as we count down from 10 days to 2 days before actual rail departure. Lower is better, and OpenTrack’s line is the one on the bottom.
As you can see, as the container gets closer to its rail departure, OpenTrack’s Rail ETD is significantly more accurate than carriers. At 3 days before departure, OpenTrack’s ETD is 7x more accurate (at the 95th percentile).
How did we do it?
One of most impactful features of OpenTrack Rail Visibility is our predictive rail ETA model that is now 50% more accurate than traditional rail visibility sources. We were able to accomplish this by leveraging rail carrier ETAs directly from the source, augmented with statistical models based on huge amounts of data we’ve observed over the years.
However, rail carriers generally don’t report ETAs until the rail leg begins. So when the container is on the water, or at the port of discharge, traditionally the only inland rail ETA available is the one reported by the ocean carrier. This is generally not very accurate.
In order to report a more accurate ETA at this stage, we leveraged our highly accurate ocean ETA model, and a new predictive model for rail dwell. This has increased the accuracy of our inland rail ETA, but it also provides a useful prediction in and of itself: The estimated departure on rail, or rail ETD.
Accurate prediction of departure time is complicated due to the high volatility of rail dwell time.
Observe the distribution of rail dwell times during a busy time at the Port of Los Angeles:
As you can see, containers are as likely to dwell two days as they are to dwell 15 days. That’s a high variance distribution!
Because of the high variance, simple averages or medians just don’t cut it. When more than half your containers will be different from the average, you need a more sophisticated solution to avoid constant ETD paranoia.
OpenTrack’s models adapt to the true shape of the data. By using a set of different machine learning models trained on historical dwell time patterns, current trends, and container-specific features, we’ve been able to issue a far more accurate rail ETD prediction.
This not only improves our rail ETA – it ultimately gives you the realistic accuracy you need for reliable operations and planning throughout each container’s unique journey.
