By Luis Lopez, AI transportation consultant, CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast
More shippers are being asked to report the emissions from their freight. The request may come from a large customer, an investor, a lender, or an internal sustainability team. Software tools, many of them now using AI, promise to produce those numbers from the data you already have. Some of what they produce is useful. Some of it is a rough estimate presented with more precision than it deserves.
This article explains how freight emissions are typically estimated, what inputs the software needs, and where the data falls short, so you can judge the numbers you are given and improve them over time.
The frameworks behind the numbers
Most freight emissions reporting is built on a few widely used methodologies. Two names come up constantly.
- The GHG Protocol Scope 3 standard. This is the general framework companies use to account for emissions in their value chain. For a shipper, freight paid for by the company generally falls under the category for upstream transportation and distribution, often referred to as Category 4. Freight paid for by customers may fall under a separate downstream category.
- The GLEC Framework. This is a logistics-specific methodology for calculating and reporting emissions across modes, including road, rail, ocean, air, and logistics sites. It is designed to make freight emissions calculations more consistent across shippers, carriers, and service providers.
When evaluating a tool, ask which methodology it follows and whether it documents its emission factors and assumptions. A number without a stated method is hard to defend when someone asks how it was calculated.
Spend-based vs activity-based estimates
There are two broad ways to estimate freight emissions, and the difference matters more than any software feature.
Spend-based
A spend-based estimate multiplies what you paid for transportation by an average emissions factor per dollar. It is quick and needs only accounting data. It is also crude. If rates rise, your reported emissions rise even if you moved the same freight the same way. If you negotiate a better rate, your emissions appear to fall. Spend-based numbers are acceptable as a starting point, but they cannot show whether operational changes are working.
Activity-based
An activity-based estimate uses what actually moved: the weight of the freight, the distance it traveled, and the mode. The basic logic is tonnage multiplied by distance, giving ton-miles or tonne-kilometers, then multiplied by an emission factor for that mode and vehicle type. More advanced versions use actual fuel consumption reported by carriers, which is the most direct measure when it is available.
Activity-based estimates reward real changes. Shifting a lane to rail, consolidating shipments, or improving trailer fill all show up in the result. For context on those levers, see this comparison of intermodal vs truckload and this explainer on freight consolidation.
The inputs software needs
For an activity-based calculation, a tool needs a few things for every shipment:
- Origin and destination, precise enough to calculate a realistic route distance, not just a straight line between ZIP codes.
- Shipment weight, ideally actual rather than estimated.
- Mode, such as truckload, LTL, intermodal, ocean, or air.
- Equipment details, where available: trailer type, temperature control, and vehicle class.
- Leg detail for multi-leg moves, such as drayage to a rail ramp, the rail leg, and final delivery.
- Carrier fuel or emissions data, if your carriers can provide it.
What AI actually adds
AI does not change the underlying math. What it does well is the data preparation that makes the math possible:
- Extracting data from documents. Pulling weights, locations, and modes from bills of lading, invoices, and rate confirmations when they are not in a structured system.
- Filling gaps. Estimating missing weights from product and order data, or inferring mode from carrier type and invoice descriptions, with the estimate clearly flagged.
- Routing distances. Calculating likely road distances rather than straight-line distances.
- Classifying spend. Separating freight charges from accessorials, duties, and other costs that should not be counted as transportation activity.
- Spotting outliers. Flagging shipments with impossible weights, duplicate records, or distances that do not fit the lane.
These are real improvements, but each one involves an assumption. A good tool tells you which records were measured and which were estimated.
Where the data falls short
As an AI transportation consultant, the first thing I look at in any emissions report is the share of shipments calculated from real activity data versus defaults. Common problems include:
Missing or estimated weights
LTL and parcel records often have weight, but truckload invoices frequently do not. If the tool assumes a standard load weight for every truck, the result reflects the assumption, not the freight.
Empty miles and backhauls
A shipper sees its own loaded leg. It usually does not see the empty miles the carrier drove to reach the pickup or after delivery. Methodologies handle this with average load factors, which may not match your actual lanes.
LTL and multi-stop allocation
When your freight shares a trailer with other shippers, its share of the truck’s emissions has to be allocated. Allocation by weight and distance is a reasonable approach, but it is still an approximation of how the network actually ran.
Carrier data availability
Many carriers, especially smaller fleets, cannot provide fuel or emissions data by shipment. Most reports therefore rely on default factors for most of their volume.
Inconsistent records across systems
If shipments live in several systems with different location formats and carrier names, the report inherits those inconsistencies. This guide on what AI cannot do with poor freight data covers the broader problem.
How to improve the numbers over time
- Start with a spend-based baseline if that is all you have, but label it clearly.
- Move your largest lanes to activity-based calculation first, since they drive most of the total.
- Capture actual weights and modes at the shipment level in your transportation system.
- Ask key carriers what emissions or fuel data they can share, and in what format.
- Document every assumption and default factor so the report can be reproduced and audited.
- Track the percentage of shipments calculated from real data as its own metric.
Emissions reporting is a data quality project before it is a software project. Clean, connected shipment records make every tool more accurate.
For more conversations on freight and AI, subscribe to the Freight Guru Podcast.
About the author: Luis Lopez is a Miami-based AI transportation consultant and logistics entrepreneur, the CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast.