What Is Predictive Rail ETA?

Predictive Rail ETA is Terminal49's ML-powered inland arrival estimate for containers moving by rail to inland destinations in the US and Canada. It combines two complementary capabilities:

• T49 Inland Destination ETAs: Terminal49's machine learning model generates a predicted arrival time for every container as soon as it departs the port on rail — even when neither the shipping line nor the rail carrier has provided an ETA yet.

• Intelligent ETA source selection: As the container moves, multiple ETA sources arrive — our T49 prediction, rail carrier updates, and shipping line estimates. Rather than using the most recent estimate, our ML-based selection model evaluates all available ETAs and promotes the one most likely to be accurate. It re-evaluates each time a new ETA is received from the shipping line or rail carrier.

The result: a mean absolute error of 1.0 day at port departure (down from 1.72 days — a 42% reduction, or ~18 fewer hours of uncertainty), with coverage expanding from ~91% to ~99% of US and Canada rail shipments.

How Does It Work?

The prediction model is trained on historical transit data and uses the following signals to estimate arrival time at the inland destination:

• Corridor historical performance — median transit times from port of discharge to inland destination, further broken down by shipping line, terminal, and rail carriers involved

• Rail distance — routing distance between the port and inland destination

• Port dwell time — time elapsed between container discharge at the port and rail departure

• Departure timing — day of week and month of rail departure, capturing weekly and seasonal transit patterns

• Container equipment — type (e.g. dry, reefer), length (20ft / 40ft), and height (standard / high-cube)

The T49 predicted ETA is calculated the moment the container departs on rail from the port of discharge.

How source selection works: The source selection model uses the same features as the prediction model, plus one additional signal: how long the container has been in transit since rail departure. Using this combined set of inputs, the model identifies which ETA source — shipping line, rail carrier, or T49's own prediction — has historically been most accurate for that specific context, and promotes it as the recommended estimate.

Which Carriers Are Supported?

Predictive Rail ETA is available for containers moving on all six Class 1 North American rail carriers:

• BNSF Railway

• Union Pacific (UP)

• CSX Transportation

• Norfolk Southern (NS)

• Canadian National (CN)

• Canadian Pacific Kansas City (CPKC)

No Letter of Authorization (LOA) is required to access this data. Please refer to our rail carrier data matrix for details on available information from supported rail carriers.

Which Plans Include Predictive Rail ETA?

Predictive Rail ETA is available on the Essential plan and Enterprise. It is included automatically — no additional add-on or configuration is needed. If you are on an Essential, or Enterprise plan, this feature is active in your account today.

If you are on the Lite plan and want access, you can upgrade to Essential. Contact your account manager or visit the pricing page to learn more.

Where to Find Predictive Rail ETA in the Dashboard

1. Navigate to the Containers dashboard, the Shipments dashboard, or open an individual container or shipment

2. Look for the Inland ETA field — this displays the best available arrival estimate as selected by the ML model (from the T49 prediction, shipping line, or rail carrier)

3. A pill next to the Inland ETA indicates which source is currently selected: T49 for Terminal49's own prediction, SSL for the shipping line, or RAIL for the rail carrier

4. Hover over the field to see a tooltip showing all available ETAs — T49 prediction, shipping line, and/or rail carrier — so you can compare sources at a glance

Accessing Predictive Rail ETA via the API

The Inland ETA is available as a standard property on the container object in the Terminal49 API — no changes to the API contract are required. You receive a single inland ETA value, which now reflects the best available estimate as selected by the ML model (T49 prediction, shipping line, or rail carrier), rather than simply the most recent ETA received from a carrier.

See the Terminal49 API documentation for the full container object schema and webhook event reference.

How Do ETA Sources Compare?

At port departure, three ETA sources are potentially available: shipping line, rail carrier, and Terminal49's prediction. Here's how they compare:

Combined, the shipping line and rail carrier cover ~91% of containers at port departure — leaving ~9% with no ETA at all. T49 Predictive Rail ETA closes that gap to ~99%.

We recommend using Predictive Rail ETA for all operational planning — it provides the widest coverage and matches or exceeds rail carrier accuracy at port departure, while continuously improving as the container moves toward its destination.

See our blog post for more detailed information.

Frequently Asked Questions

Is Predictive Rail ETA available for export (outbound) containers?
Currently, Predictive Rail ETA is available for import containers moving on Class 1 North American rail carriers.

How far in advance is the ETA available?
The T49 predicted ETA becomes available as soon as the container departs on rail from the port of discharge (Rail Departed milestone).

What if my container is on a route that connects two Class 1 carriers?
Our model accounts for interchange points between carriers. The predicted ETA covers the full journey to the inland destination, including multi-carrier segments.