GUIDE · SCRAP
How to build a scrap Pareto
you can trust.
Published · ProdCrest
Short answer. Before you rank scrap, put every record in one unit and one code list. Convert kilograms to pieces with part weights that carry a confirmation date, flag the ones that are out of date, and state the unit and the period on the chart. Then read the result with its uncertainty: when two bars are closer together than the conversion error, they are a tie, not a ranking.
Why the same scrap gives different charts
Scrap is often recorded in two ways at once: weighed in kilograms at the regrind or scrap bin, and counted in pieces on the shift sheet. A Pareto built on kilograms favours heavy parts; one built on pieces favours small, light parts. Neither is wrong, but they answer different questions, and mixing them in one chart answers none.
Two more things move the bars: reason codes (the same defect under two names splits one bar into two) and part weights (an old weight in the master turns the same kilograms into a different number of pieces). The reason-code guide covers the first; this one covers units and weights.
A synthetic example: the top bar depends on the unit
The ProdCrest demo uses synthetic data: an invented moulding plant whose ERP records scrap in kilograms with a defect code, 682 rows over four weeks. Every figure below was produced by that dataset; none is a customer result.
Ranked by weight, short shot is the largest defect. Converted to pieces with the product master’s part weights, flash moves to the top:
| Defect | kg (share) | Pieces (share) |
|---|---|---|
| Short shot (SS) | 534.6 (34.1%) | 53,058 (32.8%) |
| Flash (FL) | 425.5 (27.1%) | 53,470 (33.1%) |
| Sink mark (SK) | 211.3 (13.5%) | 21,934 (13.6%) |
| Black specks (BL) | 150.8 (9.6%) | 13,306 (8.2%) |
| Dimensional (DM) | 125.6 (8.0%) | 9,991 (6.2%) |
| Contamination (CT) | 121.8 (7.8%) | 9,912 (6.1%) |
The reason is the mix of parts behind each code. Most flash pieces in the synthetic data are on the Clip, the lightest part, so a flashed part weighs about 8 g on average against about 10 g for a short shot. Counting each part once lifts flash; weighing lifts short shot. The plant has to decide which question it is asking: where the most material is lost (kg) or where the most parts are lost (pieces).
Old part weights move the bars too
In the same synthetic plant, two products had master weights last confirmed more than 12 months earlier: Cap 20mm (6.2 g in the master) and Clip (2.4 g). Both parts had since become heavier, at 6.7 g and 2.6 g. For the Clip, 202.6 kg of scrap converts to 84,397 pieces with the old weight and 77,905 with the current one: the old weight overstates its scrap by about 8%.
Redone with current weights, flash and short shot end up 44 pieces apart (50,391 and 50,347). The gap between the top two bars is smaller than the error from the stale weights, so the honest reading is “flash and short shot are tied, together about two thirds of scrapped pieces”, not “flash is the top defect”.
The method
1. Choose the unit for the question
Pieces for “which defect loses the most parts”, kilograms for “where is material lost”. Put the unit in the chart title.
2. Convert with dated weights
Keep part weight in the product master with the date it was last confirmed. Flag every conversion that uses a weight older than an agreed limit (the demo uses 12 months) and treat those counts as estimates. Re-weighing a few parts is cheap; arguing over a chart built on an old weight is not.
3. One code list
Use the defect codes from one source and map any free text to them before counting. Keep the rows that do not map visible, as their own bar.
4. State the denominator for scrap rates
The demo uses scrap ÷ (good + scrap), in pieces, per machine and product. Whatever definition you choose, write it on the report, so a rate from one month can be compared with the next.
5. Read close bars as ties
If two bars differ by less than your known conversion or counting error, report them together. Decide where to look first using what you know about each defect, not a gap of a few hundred pieces.
Short version: “The scrap chart had two names”, a 30-second fictional factory story on YouTube.
Limits
A Pareto says where scrap is concentrated, not why it happens, and neither ranking says what the scrap costs: that depends on material, the process step where the part was lost and what can be reground. The cause is found on the floor, with the people who run the process.
See the scrap view in the interactive demo, or the unit check in the list of checks for production records. ProdCrest helps define scrap codes and units as part of fractional quality management. Describe your situation.