I talk to customers across different industries every day, and one pattern keeps showing up. Multiple functional areas within a company, including Logistics, Finance, and Procurement, all want more data and, more importantly, the proper information to answer questions they are being asked about the business. They're watching the appropriate KPIs or comparing their KPIs to industry benchmarks or indices. But sometimes the KPIs themselves are telling a misleading story. And if you're making decisions based on the surface number without understanding what's underneath it, you're going to get caught off guard.
Let me give you a real example so we're not talking completely hypothetical here.
What the LMI Was Really Telling Us Last Year
The Logistics Managers Index, or the LMI, is a common industry index for measuring warehousing, inventory, and transportation, factoring in both capacity and pricing. Last year, the LMI was recording what looked like a pretty stable trend line. However, as is the case with many composite index values or KPIs, you can have competing variables that net sum to zero even though their underlying components or influencers are individually illustrating risk.
Inventory was being built up early in the year due to tariffs. Warehousing was getting constrained. And transportation capacity looked really good while prices weren't necessarily going up. With more inventory, transit times were elongated, thus permitting slower or more consolidated modes of transportation. So the composite index portrayed more stability than what the influencers were showing on their own merits.
My point is, what was actually happening is that companies were storing more inventory and taking longer to move it because they had it on hand. A slower transit time permits more load consolidation (holding orders to fill trucks) or different modes of transport, such as intermodal (rail/truck). It appeared there was extra capacity in the market. But that was a false sense of security. When that inventory was depleted, and more last-mile delivery was needed to move products with a shorter transit time requirement, the capacity everybody assumed was there simply wasn't.
We saw unexpected capacity constraints late in 2025 that were generally unplanned and are somewhat persisting now in early 2026. The uptick in demand was planned, but the capacity constraints weren't. And you could argue that an AI interpretation of that number would have been just as flawed. The math said one thing. The reality said something else. The point is we need to all be careful when interpreting data, especially indices and KPIs. We need to understand the underlying influencers that make up the KPI or index and determine how they are impacting the overall value.
Cost Per Unit Weight Is Another One to Watch Carefully
Cost per unit weight is another example of a very common KPI used in transportation logistics. The advantage of this KPI is that, regardless of shipping volume, it is a measurement of overall optimization of many underlying influencers (rate, carrier selection, service selection, distance, etc). Trending these KPI results in the measurement of efficiency as shipping volume increases or decreases. The objective of a well-run supply chain is to handle the scaling of capacity or demand without necessarily paying more per unit weight. In other words, I could produce more without increasing the cost per unit of weight I ship through the supply chain.
But that KPI has several variables that create the movement up, down, or flat. How much weight am I shipping per shipment? This is an important factor as pricing isn't linear (we can thank minimums for that). It's asymptotic on a curve due to either minimums and/or weight breakpoints. Fuel is also a factor along with accessorial costs (non-freight, non-fuel), etc. The point is if you don’t understand what influences your KPI, you might be deceived by simply looking at the surface and not by the individual contributing factors. So when you see cost per unit weight moving in a direction, the real question is what's driving it. And sometimes those drivers are telling you different things. You could be having a major problem in one area that is masked by a really positive aspect of another.
Is It a Blip or Is It a Trend?
This is another question I frequently ask when analyzing data. A “blip” is a transitory event that will eventually settle back into the previous state. A trend is something that's going to persist. And you've got to be careful, because if you overreact to a blip, you tend to miss the mark.
Think about what happened with the tariffs. Many planners overreacted a bit to the increased duty costs by overstocking inventory before the tariff was made effective. Maybe it played out in a way you expected, maybe it didn't. The key question is, did you have the data to perform the complex math? The math that shows the impact of inventory carrying costs vs. transportation costs to meet some level of forecasted product needs to be delivered using some standard transit time service level agreement. These calculations require accurate data to properly evaluate the cost tradeoffs of any forward buying decision.
The AI and Forecasting Reality Check
Let’s be honest, the technology we use to solve everyday supply chain logistics challenges has definitely improved. The underlying data remains challenging, but companies are realizing more and more the value of the data asset. Given the incompleteness of data and the quality of data, most AI models are not yet sophisticated enough to know what the “correct answer” is.
I've tried this myself. If I ask a simple question through an AI agent, something like "Why is my cost per unit weight increasing?", you're going to get a wrong answer almost guaranteed. It's not capable yet of doing multivariate math and problem-solving at that level. And it's also not aware when the answer and the data are wrong. The easy questions aren't the ones we're struggling with; it's the hard questions that remain elusive.
And here's the other thing. Because we're in a more volatile market, historical use cases don't always apply. These models learn from history. If history isn't a good judge of the future, you're going to get a wrong answer. So the real key is data quality and accuracy, yes, but more so, good human in the loop critical thinking remains the most practical way to manage the supply chain and leave AI for automating repeatable, simpler tasks.
Read the Tea Leaves, But Read Them Right
When you “read the tea leaves”, sometimes the tea leaves are confusing. The math gets more complicated, and the proper interpretation is more challenging than meets the eye. So you have to look at it not only from the surface level, but also investigate the influencers that contribute to the overall value. One should never take a composite KPI at face value, but instead, watch for divergence in the underlying factors. Build the discipline to separate blips from trends. And invest in your data asset, because that's the foundation for everything, even the most advanced AI models. Companies that have or are investing in the data and technology to understand it will no doubt be in a better place to navigate change and maintain a healthy level of resiliency, regardless of market conditions.
Steve Beda is EVP of Customer Solutions & Advisory at Trax. With extensive experience in supply chain automation and transportation logistics, Steve works closely with customers across industries to help them take full advantage of Trax's Total TSM/TSI Spend Intelligence services.
Want to dig deeper into your supply chain data and make smarter decisions? Contact Trax to learn how we can help.