FAQ

Triumph Rate Intelligence is a dynamic pricing engine for freight brokers that uses AI and machine learning to provide predictions of real-time truckload buy rates customized to an individual freight brokerage's buying power.

A buy rate is the cost of a truck in a given lane at the time of the prediction.

The Verified Buy Rate is the rate that the model predicts will be the best option for a customer when they enter a lane and transport type into our UI. Different brokerages will usually have different predicted rates on the same lane due to their difference in buying power as derived from their historical data. This rate is primarily based on the customer’s historical data, but can also be influenced by external data, such as other brokerages that contribute data to the network or external macroeconomic factors.

The Verified Buy Rate is AI generated using a data training set to train a unique, tailored model composed of and heavily dependent upon the data and buying power of a specific brokerage in conjunction with market conditions and Triumph network data.

The predicted fuel cost included in the Verified Buy Rate is based on data for the lane from your brokerage and the Triumph Network. Because Triumph calculates rates based on historical data for the entire linehaul, the amount allotted for fuel in the Verified Buy Rate will differ from the market average listed under Fuel, which is there for the sake of comparison and for offering a cost breakdown to shippers.

When the machine learning model predicts a Verified Buy Rate, it also predicts maximum and minimum possible rates, with the Verified Buy Rate somewhere in the middle. The Verified Buy Rate is the model’s prediction for the most likely rate on the load. The Start Rate is the prediction for the lowest possible successful rate.

The algorithm deduces the Start Rate by evaluating the potential variability of the load price around the Verified Buy Rate. It rounds the lowest number predicted as a possibility down to the nearest $50 to get the Start Rate. Where the Confidence Level is lower, you’ll see a lower Start Rate. The Start Rate helps guide price discovery where the Confidence Level is on the low side. When the machine learning model predicts a Verified Buy Rate, it also predicts maximum and minimum possible rates, with the Verified Buy Rate somewhere in the middle. The Verified Buy Rate is the model’s prediction for the most likely rate on the load. The Start Rate is the prediction for the lowest possible successful rate.

The Triumph Market Rate represents the expected rate in the market. The Market Rate has no bias towards any one customer’s data. It can be lower or higher than the target rate, depending on your brokerage’s buying power vs the general buying power of the market.

The Verified Buy Rate is the rate that Rates Intelligence predicts will succeed for your particular brokerage. It reflects both market conditions and your brokerage’s specific buying power. The Market Rate is the predicted market rate. It’s based on data from the Triumph Network and is not biased toward any one brokerage.

The Market Rate is based on data from brokerages in the Triumph network. When a brokerage requests a rate, Rates Intelligence draws on data from over 100 customers with $18 billion in annual truckload revenue to provide this rate and accompanying insights into pricing trends in the same market area.

The Verified Sell Rate is the quote the machine learning model predicts as most likely to win the customer. The rate is based on the brokerage's previous quotes for the lane and customer. The brokerage can also set a specific markup for a particular customer, and the model will use that instead. If no historical data is available and no markup rules apply, the markup will default to 15%.

The Confidence Level is a measure of probable accuracy. It usually falls between 50 and 100. This number does not necessarily represent a more or less accurate prediction. The higher the Confidence score, the easier it should be to book at the predicted rate.

The Confidence Level is primarily determined by three things:

The Better Rate tells you which rate between the Market Rate and Verified Buy Rate has a higher Confidence Level indicating which rate might be quicker to cover and easier to find capacity for based on current market conditions.

Similar Lanes share similar features, such as a common destination. A Rates Intelligence machine learning model recognizes similarities and connections between lane features, which helps it to make accurate predictions, even for new lanes. This is especially useful when making predictions for lanes with little or no historical data.

Some Rates Intelligence widgets include a toggle at the top right with the options Lane, 3d ZIP, and Mkt. This lets you view data for the wider area surrounding the lane in the prediction. A green dot next to one of these options means that data is available.

The Top Carriers listed for the shipment are carriers from the brokerage's carrier network who have moved the load for the brokerage over the past 60 days. The Top Carriers section includes details about volume, price point, how recent the shipment was, and contact information. It's there to help you choose the best carrier and connect with them easily.

Negotiation Tips help you negotiate with customers by offering a quick summary of conditions that might affect rates, such as current events, load history, market dynamics, and holidays.

Machine learning is the process behind systems that use inferences and statistical models to find connections in data. Patterns in the data allow the model to draw conclusions and make predictions. Our models are used to predict prices of future loads based on patterns found in historical load data.

Machine learning training is the process of building and calibrating a model's logic based on historical data to help it answer questions correctly. In most cases, the machine “learns” by testing examples and receiving feedback on how close each result is to being correct. It continues to adjust the “weight” it assigns to each factor in determining a result, and over time, the model gets better and better at predicting outcomes, while following new trends appearing in more recent data.

Before making a new record of data available to test and training sets, we run the record through a filtration process that identifies whether the given record has surpassed certain thresholds. These thresholds are designed to protect the model’s integrity from data that might skew predictions.

Triumph machine learning models depend heavily on historical load data to assist in rate generation. While there is alternative market data that helps paint a forecast, the model performs and corrects best when analyzing historical load data, which gives it a foundation for how rates should be compiled and helps the model improve.

The machine learning models consider numerous market conditions and trends, such as seasonal change and load characteristics. These are used daily to train the AI, which makes adjustments to the prediction algorithms across every lane Triumph supports. Predictions can be noticeably lower/higher than the historical average at times, due to the nature of the lane and where the data may be pointing next.