Maximizing the revenue of every cargo hold

Every flight has a deadline for earning cargo revenue

Once a flight takes off, any unsold cargo capacity loses its revenue potential. But selling that space starts weeks earlier, while airlines are still estimating how much they can carry and how much demand they can expect.

For cargo traveling in the belly of passenger aircraft, available capacity depends on passenger loads and operational constraints. Demand varies by route, shipment size and cargo type. Airlines need to set prices while these conditions are still changing. Price too low, and they risk giving away revenue. Price too high, and space may go unsold.

Our client provides software across the air cargo ecosystem, supporting airlines and the companies that handle and move their shipments. They wanted to extend their platform's role in day-to-day operations by helping airline partners make these pricing decisions with better information.

Tryolabs worked with their team to build the forecasting and optimization components for an AI revenue management system, with a planning horizon of up to approximately 20 days before departure.

The data needed to reflect how cargo actually moves

Historical bookings held valuable information, but their structure made it difficult to use directly for pricing individual flights.

A shipment could travel across several flight legs under a single total price. It could also carry multiple special handling classifications, each affecting its rate. Before modeling demand, we needed to estimate how much of that price belonged to each leg and represent the different cargo categories and weight brackets.

We used market pricing data to allocate shipment revenue across flight legs and separate the pricing factors the system would need to evaluate.

There was another gap: booking records showed the cargo that flew. The model also needed examples of combinations with no bookings. We added those cases to the training data to reduce the risk of forecasting demand where none existed.

This work gave the models a foundation that reflected the commercial decisions airlines needed to make.

Pricing recommendations up to 20 days before departure

A matrix of recommendations for different combinations of time to departure and remaining capacity.

Connecting forecasts to a pricing decision

The system brings together capacity forecasting, demand forecasting and a constrained price optimizer.

The capacity model estimates how much cargo a flight can carry, accounting for passenger loads and operational limits. The demand model estimates expected shipment volumes by flight, cargo type and weight bracket, using historical patterns and market pricing signals. It also allows the optimizer to evaluate how predicted demand changes at different prices.

The optimizer tests alternative base rates and weight-bracket adjustments against those forecasts. It searches for prices that maximize expected revenue while respecting available capacity and the airline's business rules.

Airlines also needed recommendations for changing booking conditions. We built a pricing matrix covering different combinations of time until departure and remaining capacity. This gives them a set of recommendations to consult as bookings progress, without running the full optimization again for every scenario. Airline teams retain control over the decisions.

Testing the revenue opportunity

In next-day testing, the forecasting model used for capacity planning achieved a mean absolute percentage error under 20%. The framework also supports forecasts extending to approximately 20 days, with separate evaluation of performance over that longer horizon.

Historical backtesting showed how the optimizer could identify pricing configurations with higher modeled revenue than the historical baseline. These results come from historical simulations. Results from live operations are still being measured.

The analysis also highlighted a limit to discounting: lower prices do not always create enough additional demand to improve revenue. The pricing engine helps evaluate that trade-off for each flight.

For the software provider, the work established the forecasting and optimization foundation for a new revenue management capability. It connects the data already passing through its platform with recommendations that help airline partners decide how to price the cargo space they have left to sell.