What freight network optimization truly means

There's a question that comes up on almost every shipper call when we start talking about SemiCab and how we see transportation as a network optimization problem. Almost all of them tell us they run a network optimization exercise regularly and then describe the conventional supply chain problem of where their facilities should be. Where to open new distribution centres and manufacturing plants, which ones to close. It's the same story every time. A consulting team. A few months of data gathering. A model of the distribution footprint. A final presentation recommending they close two distribution centres and open one somewhere central. Usually there's a slide with a map covered in circles.

1: The shape question

That is valuable work, but I think of it as network design, because what it really determines is the shape of the network. Where inventory lives, how many facilities, how far from the customer. You answer it rarely, because the answers are expensive to act on, you don't relocate a distribution centre on a whim.

2: The flow question

When we talk about freight as a network optimization problem, we're asking something else: given the current shape of the network, how should freight flow through it? Which truck should carry which load, in what order, to make the entire operation run as efficiently as possible. This isn't a one-time decision. It's a daily, thousands-of-times decision made by planners and dispatchers who might not even call it optimization while they're doing it.

It's also the harder problem. The two key inputs - demand and capacity, shipments and trucks are continuously moving across time and geography, and the variability is intrinsic, not a defect to be engineered out. That's part of why the industry has put off tackling it. It's exactly what we built SemiCab to solve.

3: Where the math went

The industry gave its best mathematics to the rarer question. Static network design got the modelling, the scenario engines, the PhDs. The daily flow decisions - the ones that repeat, that compound, that actually consume the transportation budget - run on spreadsheets, experience, and whoever's been doing the lane for six years.

4: What it looks like in practice

Four things, in my experience.

• One view of the whole. Demand and capacity visible in one place, not scattered across lanes and email threads. You cannot optimize what you can only see one shipment at a time.
• Loads planned together, not one by one. A dispatch decision that looks right in isolation is often wrong for the network.
• Re-planning as routine, not exception. A breakdown at 2pm, a delayed inbound, a demand spike - these aren't disruptions to the plan. They're inputs to the next one. If your response to change is firefighting, you don't have a plan.
• Measurement at network level. Rate per mile tells you how well you bought. Utilization and empty miles tell you how well you ran. Most teams track the first weekly and the second never.

5: Math and judgment

None of this replaces planners. It gives their everyday decisions the same mathematical rigor that design studies have always had. The math does the calculation; people make the call. That combination is what optimization should have meant all along.

The way forward

So, an honest question: how much rigor do your daily freight decisions get, compared to the study you ran three years ago?

‍