Math & Tribal Knowledge: friends or foes?

Every operation runs on two different brains, and they almost never talk to each other.

The first brain is math. You set the year on it. The quarter, the month, the route plan, the capacity model, the cost-to-serve targets. Someone builds a model, the numbers reconcile, leadership signs off, and the plan goes live. For a brief window, the operation and the math agree on what should happen.

Then the day starts, and the math disappears.

By 6 a.m. the plan meets the real world. A driver calls out. A store runs short. Volume spikes in one zone and craters in another. Weather closes a lane. And the people running the floor do what good operators have always done: they make the call. They reroute, they reassign, they hold an order, they split a load. They do it fast, and they do it on instinct built over years of doing the job.

That instinct is real. Tribal knowledge is not a bug. A dispatcher who knows that a particular route always runs hot on Fridays is carrying information your model never captured. The problem is not that operators override the math. The problem is that when they do, the math is gone. They cannot see what the override costs.

You have seen this everywhere

This problem shows up the same way in every kind of operation.

Take the dispatcher at a food distributor. A routing tool generates a route. It is mathematically sound. It balances stops, respects time windows, minimizes drive time. Then the dispatcher looks at it, decides it is wrong, and changes it. Maybe they are right. Maybe that driver does hate that part of town and will run slow and unhappy all day. Maybe they are wrong and just protecting a habit. Either way, nobody in that moment knows the financial difference between the two routes.

Take the branch manager at an equipment rental company. Two jobs, one crew, and a call to make: send the truck to drop off a unit on a new rental, or send it to pick up an asset sitting idle in the field. The drop-off feels urgent because a customer is waiting. So the pickup gets pushed to tomorrow. But that idle asset is dead capital, off-rent, earning nothing and unavailable to re-rent to the next customer. The branch manager is weighing a loud customer against a quiet one, when the question that actually matters is which choice protects more revenue. That math is nowhere on the screen.

Take the wholesaler with their own fleet. An order comes in and the instinct is automatic: use our truck. We own it, the driver is on payroll, it feels free. So the owned truck takes the run instead of a third-party provider who, on that specific lane and load, is cheaper and faster. The truck was never free. It carries a driver, fuel, and the opportunity cost of the more profitable run it could have been doing instead. But “use our truck” is a reflex, and the real cost-to-serve of that reflex is invisible at the moment of the call.

Three operations, three operators, one pattern. An experience-based decision made against a math-based reality nobody could see. Multiply that by every shift and every call. The plan you reconciled in the model quietly erodes through a thousand small decisions, and you only find out at the end of the month when the numbers come in soft and nobody can tell you exactly why.

This is the real cost of the two-brain problem. Not that operators use judgment. That they use it blind.

Both inputs, in the moment

This is the part Nash changes.

When an operator goes to make a change, Nash surfaces the math the moment they reach for it. Not as a report the next morning. Right there, in the decision. Pull that stop off the optimized route and here is what it does to drive time and margin on the order. Push the field pickup to tomorrow and here is the off-rent cost of that asset sitting idle another day. Send the owned truck instead of the third-party provider and here is the true cost-to-serve once you count the driver, the fuel, and the run that truck could have been doing. Here is the EBITDA impact across the shift if the choice becomes the pattern.

The operator still decides. That is the whole point. Their experience is an input the model never had, and it stays an input. But now it sits next to the financial reality of the choice, in real time, before the decision is locked. The dispatcher who knows that route runs hot on Fridays can weigh that against what the override actually costs, and make a call that holds up both on the floor and in the model.

That is the resolution. Not math winning. Not experience winning. Both in the same decision, at the same moment, with nothing hidden.

Equilibrium in motion

We believe the best possible outcome in an operation is rarely the pure model answer and rarely the pure gut call. It is the point where both are accounted for, given the constraints in front of you right now. And because the conditions never stop moving, that point keeps moving too. The operation that can find it, again and again, through every decision, is the one that will win.

Math sets the plan. Experience runs the floor. For decades those two have lived in separate rooms. Putting them in the same decision, in real time, is what operational equilibrium actually looks like.

If you want the broader thesis, it lives in The Logistics Singularity and the launch post The Autonomic Era of Logistics.

See EBITDA-based agentic routing live.

Nash on the floor at Samsara Beyond 2026. Wednesday, June 24 · Las Vegas.

View session details →