Where does the scrap go? Marking production loss with data instead of instinct
In a production workshop, scrap is the thing everyone knows about and no one can put a number on. At the end of the shift someone says “we gave away a bit more today.” At the end of the month the gap between material consumed and goods shipped appears — but it stays one lump total. Which job, which station and which reason produced the loss dissolves inside that total.
The first step in reducing loss is not reducing it. The first step is marking it where it happens. And before measurement comes definition.
First, agree on what counts as scrap.
In the same workshop the word can cover several different things. There is unavoidable process loss, such as cutting allowance. There are the first pieces that come out during a setup. There is output that fails and gets reworked, and output written off entirely. Put them all on one pile and the resulting number tells nobody anything.
The definition should be short and written down. A few headings are enough: process allowance, setup loss, rework, scrap. Each heading gets a one-sentence meaning, and that sentence answers “in which case does it get written here.” The unit is part of the definition too: pieces, kilograms or metres? If two units are in use for the same item, two different numbers come out and each looks correct on its own.
The definition is written together with the people doing the job. A definition written at a desk does not hold on the floor. Every point where an operator hesitates — “which one is this now?” — is a point where the definition is incomplete.
How wide the definition needs to be depends on the question being asked. “How much do we lose in total” and “on which job do we lose it” call for different records; the second ties every entry to an item and a station. Writing that question down before measurement starts is what keeps the data usable later.
A measurement point sits where both sides of the number are known.
Loss becomes visible as the difference between two counting points. So the measurement point goes where what went in and what came out can both be known. Material intake at the head of the line, the stations where the product changes shape, quality control, and the final count before shipping are the usual candidates.
They do not all have to be set up at once. The place where material changes shape, and where a decision follows, is enough to start. Once the question “does this loss occur before or after that point” has an answer, where the second point belongs becomes obvious by itself.
The unit counted at a measurement point is fixed too. If what goes in is kept in kilograms and what comes out in pieces, the difference cannot be computed until the conversion is a written agreement. The moment of counting is defined as well: when material enters the station, or when the operation finishes? That detail is the most common reason two shifts report the same job differently.
The record is kept at the moment and at the place the loss occurs.
The problem with a paper sheet is not the paper; it is the delay. A sheet filled in at the end of the shift is a remembered record. Small, frequent losses do not survive memory — only the one big event that annoyed someone that day gets written down.
So the record moves onto a tablet at the station where the loss happens. The screen asks four things: how much, which item or work order, which station, which reason. The reason is picked from a short list; leave a free-text box and the list loses its meaning over time. Whether the entry takes more than a few seconds is what actually decides whether it gets made at all.
A machine connection is not required. Operator entry, barcode scanning and records from existing systems are enough for a first measurement. Sensors come up only where the measurement plan genuinely calls for them.
In the first weeks, the first thing you see is not the loss.
The earliest finding after measurement starts is usually the quality of the record itself. Some shifts have no entries at all. On some days most reasons land under “other.” That is not a failure, it is a finding: either the list is incomplete or the definition does not match what happens on the floor. The list gets corrected from that feedback.
Once the record settles, the second finding is distribution. Loss does not spread evenly; a few reasons, a few stations or a few products stand out. That ranking is not a conclusion — it is a pointer to where to look.
The third finding is consistency. The recorded scrap total is placed next to the gap between material consumed and goods shipped. If the two converge, the measurement is catching reality. If a wide gap remains, either there is a loss point nobody records, or an item the definition leaves out.
Some things stay invisible in the first weeks. The cause of a loss — tool wear, a change of raw material, a habit in setup — does not fall out of a single measurement; that needs a series built under one unchanged definition. Converting loss into money is not rushed either: an amount computed before unit cost is defined pulls the discussion away from measurement.
Scrap measurement is not an instrument of blame. If a loss an operator records is used against that operator, recording stops within weeks. Measurement survives only when it is clear that the number describes the job, not the person.
At Miletus this work starts on the engineering side: the loss is defined and the measurement points are chosen first, and software follows. On the production line, the first build is usually exactly that — one screen that keeps downtime and scrap records on the line itself. Once measurement is in place, “where does the scrap go” is answered by the record instead of the argument.