Why Manufacturing Companies Lose Visibility Across Departments

Almost every manufacturing business can describe the moment it happened. At some size, the owner stopped being able to answer questions about the plant from memory, and started having to ask.

It is rarely attributed to anything structural. The usual explanations are that people have stopped communicating properly, or that a particular manager is not keeping on top of things. Both are almost always wrong, and both lead to expensive non-solutions.

Reason 1 — The company outgrew the range of a conversation

A small plant coordinates by conversation, and it works extremely well. Everyone is within earshot, the number of things in progress is small enough to hold in your head, and the person who knows is two metres away.

Conversation scales badly. It does not degrade gradually either — it works, and then at some threshold of people, shifts and concurrent jobs it stops working, and nothing has visibly changed to explain why.

Reason 2 — Data is produced for reports, not by the work

This is the deepest structural cause. In most plants, the operational record is created afterwards: output written on a register and entered the next morning, material issued and booked later, a daily report compiled from several sources.

Data produced this way has three properties that make visibility impossible. It is late, so it describes a situation that has already changed. It is second-hand, so it reflects what the compiler believed. And it is rounded, because nobody reconstructs exact figures from memory.

If the information only exists because somebody compiled it, then visibility is limited by how often somebody is willing to compile it. That is the ceiling, and no amount of reporting effort lifts it.

Reason 3 — Each department counts differently

Sales counts an order when it is confirmed. Production counts it when it is scheduled. Dispatch counts it when it ships. Every one of those numbers is correct, and they contradict each other.

The effect is corrosive: when three true numbers disagree, the response is to distrust all of them and go back to asking people. Shared definitions are unglamorous and they are the precondition for everything else.

Reason 4 — The handoffs have no record

Work moves between departments by message, meeting and memory. Those handoffs leave no trace, so nothing about them can be measured.

This is why so many plants can tell you their output but cannot tell you why an order was late. The delay did not happen inside a department, where there are records. It happened between two of them, where there are none.

Why adding reports makes it worse

The instinctive response to lost visibility is to ask for more reporting. This reliably makes the situation worse, for three reasons.

1. It consumes the time of the people who should be running the operation, so the operation gets less attention rather than more.
2. It adds another compilation step, which means another opportunity to differ from the source.
3. It creates a comfortable feeling of visibility that is actually a delayed and averaged view — which is more dangerous than knowing you cannot see.

What actually restores visibility

Only one thing: making the work produce the record. Not as an additional step, but as the same step.

• Output recorded on the line as production happens, not compiled afterwards.
• Material issued against a job, so consumption is a by-product of issuing.
• Approvals actioned in the system, so pending work has an owner and a clock.
• Handoffs as transactions rather than messages, so the gaps between departments become measurable.
• One written definition per metric, so the numbers stop contradicting each other.

Once those five are true, a dashboard is almost trivial to build — and until they are true, a dashboard is a liability. That is the whole of it.

A diagnostic you can run this week

Ask two departments the same question on the same morning. How many orders are currently open? How much of item X is available? How many jobs are in progress right now?

If the answers differ, you have measured the problem. If nobody can answer without going away to check, you have measured it more precisely.