Walk into a mid-sized distributor at eleven in the morning and ask a simple question: which of today’s deliveries is going to fail? Almost nobody can answer it. Not because the business is badly run — often it is run by extremely good people — but because answering it requires assembling five facts that live in five different places, and no system is accountable for assembling them.
The information is not missing
This is the part that gets misdiagnosed. Leaders assume they have a visibility problem, so they buy reporting. But the shortage was known at 09:12, by the picker. The damaged pallet was known at 07:40, by the receiver. The broken-down vehicle was known at 10:05, by the driver and about nine people in a message group.
None of that information was missing. It simply never travelled to the person whose commitment it threatened, in time for them to do anything about it. A dashboard does not fix that, because a dashboard waits to be looked at, and it shows you conditions rather than consequences.
Your ERP knows what was agreed. Your floor knows what is happening. The gap between those two sentences is where operations actually fail.
Why the existing categories do not close it
The systems in a distributor’s stack are each competent inside their own boundary, and each indifferent to what happens outside it.
An ERP is a system of record. It is authoritative about what was agreed — the order, the price, the terms — and it is deliberately not designed to know that a pallet arrived crushed. A WMS runs a warehouse extremely well and stops caring at the door. A TMS runs transport and starts caring at the door. A BI tool tells you last month was worse than the month before, which is true, unhelpful, and too late.
Between all of them sits the actual work: noticing that a receipt discrepancy this morning has quietly broken a delivery promise on Thursday, deciding what to do about it, getting a named person to do it, and being able to show afterwards what happened and why. In most operations, that work is done by people, in their heads, over the phone, and it does not survive them going on leave.
The hidden cost is not the failure. It is the chasing.
Ask an operations manager what they did yesterday and the honest answer is usually some version of: found out what was going on, worked out what it affected, and got hold of the right person. Three tasks, repeated all day, generating no output that any system retains.
That is a real, measurable cost — we treat manual coordination hours as a primary metric for exactly this reason — but it is also a structural cost. It means the operation cannot grow without adding coordinators, cannot open a second site without cloning its most experienced people, and cannot explain its own performance at the end of the month.
What the missing layer has to do
The gap is not closed by a better version of any existing category. It is closed by something that sits deliberately in the middle and takes responsibility for five things, continuously:
- Assemble the fact. From a form, a spreadsheet import, an ERP event or a scan on the warehouse floor — treated identically once it arrives.
- Resolve the consequence. Connect that fact to the orders, customers, commitments and sites downstream of it.
- Raise the right work. Not an alert. A task or an exception with a cause, an owner, a severity, a deadline and a next action.
- Apply the rules. Your policy, your approvals, your separation of duties — enforced rather than remembered.
- Keep the evidence. What happened, what it affected, who decided, on what basis, and what the outcome was.
Why nobody has built it
Because it is an unglamorous place to sit. A coordination layer has to be genuinely useful to a company that will not replace its ERP, will not integrate for six months, and has a warehouse with unreliable signal in the back corner. It has to be honest about authority, so it never quietly becomes a second source of truth. And it has to be trustworthy enough that a supervisor will act on what it tells them at four in the afternoon on a bad day.
That set of constraints rules out most of the obvious approaches. It rules out rip-and-replace, because the risk is unbearable. It rules out integration-first, because it excludes the operations that need help most. And it rules out autonomous AI agents making unsupervised decisions about physical goods, because when that goes wrong, it goes wrong in a truck.
Coordination is the last part of operations that is still done entirely by people, in real time, from memory — and the first part that breaks when the operation grows.
What we are building
Lumanw is our attempt at that layer. It connects your systems of record to your physical operation, identifies what is blocked, shows what it affects, coordinates the next action inside your policy, and keeps the evidence. It supplies native execution only where you genuinely have a gap, and it defers to your existing systems everywhere else.
We are early, and we would rather say so than manufacture proof. What we can offer instead is a published standard for what has to be true before a release touches an operation, a scorecard we agree with you before we start, and a first deployment narrow enough that you can judge us on it.