Reducing Scrap in the Machine Shop — A3 Practical Problem Solving Report

Author's Note · 2026

This is a corrected and expanded version of an A3 case I originally published in A3 Thinking (Productivity Press, 2008). The book was published in a smaller 6 × 9 inch format, which forced the A3 — a sheet designed to be read at 11 × 17 — to be shrunk to a tiny illustration. To make it legible at all, I had to collapse a proper multi-level cause-and-effect diagram into a single-level fishbone. That was a compromise, and a wrong one. A real A3 cause-and-effect analysis works in layers: 4M categories, then specific causes, then the deeper "why" beneath each one. The book version flattened all of that into a single rib of one-word labels.

This page restores the case at full scale, with the cause-and-effect structured the way it should have been from the beginning. The problem, the data, the team members, and the countermeasures are unchanged from the original.

Step 1

Background

Why scrap matters this year, and how the company is tracking against goal.

Corporate Goals · 2006

Increase global market share
Improve quality & service
Increase corporate profits

Manufacturing Goals · 2006

Reduce cost 5%
Reduce scrap 15%
Improve productivity 7%
Improve HSE Index 10%

Linkage: The corporate goal of improving quality and service flows down to a manufacturing target of cutting scrap by 15%. Scrap is one of the largest controllable cost categories in the machine shop, and a leading indicator of process stability.

Overall Scrap Rate · 3-Year Trend

2004

3.2%

2005

2.7%

2006 YTD

2.6%

            Black tick = 2006 goal of 2.3%
          

Not on track for 2006. Scrap has improved year over year — 3.2% → 2.7% → 2.6% — but 2006 YTD is still 0.3 points above the 2.3% goal. Year-over-year improvement has slowed; another general push will not close the gap. The team needs to find where the scrap is actually concentrated.

Step 2

Current Condition

Where the scrap actually lives — by department, then by process.

Step 2a · Scrap by Department

Machine Shop

target $500K

$700K

Welding Shop

at target

$200K

Assembly Shop

at target

$86K

Machine Shop is $200K over target. Welding and Assembly are both at their targets. The entire gap to the corporate scrap goal is concentrated in one department.

Step 2b · Machine Shop Scrap by Process

Milling

1.5%

vs 1.56%

▲

Turning

0.9%

vs 0.8%

●

Drilling

0.7%

vs 0.6%

●

Rough Grinding

3.7%

vs 2.0%

Final Grinding

8.7%

vs 4.0%

          Legend: ● 0–1% · ▲ 1–2% · ✕ 2%+
        

Specific problem: Final and Rough Grinding are both well over target. Milling, Turning, and Drilling are at or near goal. Final Grinding alone is more than double its target rate. The investigation focuses here.

Step 3

Goal

A specific, measurable target tied to the gap.

By December 2006

Rough Grinding

Reduce scrap rate from 3.7% to < 2.0%.

By December 2006

Final Grinding

Reduce scrap rate from 8.7% to < 2.0%.

Closing the dollar gap. Bringing rough grinding from 3.7% to under 2% and final grinding from 8.7% to under 2% is what closes the $200K gap to budget. Grinding is the dominant scrap source in the machine shop — milling, turning, and drilling are all at or near target — so cutting these two rates to under 2% pulls total machine-shop scrap from $700K back to the $500K target. The percentage goal and the dollar goal are the same goal, expressed two ways.

Note on goal-setting. The targets are tied directly to the gap identified in Current Condition — not pulled from corporate. The goal is what closes the gap, and is stated as a specific number, on a specific process, by a specific date.

Step 4

Root Cause Analysis

Cause and effect — properly structured this time.

How to read this section. The original book version collapsed everything below into a single-level fishbone — one rib per 4M category, with one-word labels. That is not how cause-and-effect analysis actually works. Below, each of the four M's is a card. Inside each card, the specific causes (the bones) sit one level down, and the deeper "why" sits one level below that. The two confirmed root causes are highlighted in red.

M1 · Man

Human

Operator and setup-related causes

Contamination

Foreign material on workpiece or chuck Why? — variable cleaning practice between shifts
Coolant film carryover Why? — no defined wipe-down standard at load

Grinding wheel set-up

Wheel mounted off-center Why? — flange seating not verified by gauge
Dressing tool position drift Why? — no zero-check before first part

Manual offsets

Operator-entered offset error Why? — adjustment based on judgment, not measurement
Stale offset carried part-to-part Why? — no cycle-end reset routine

M2 · Machine · Confirmed

Machine

Equipment condition and capability

Spindle head runout

Measured runout: 30 microns vs. 10 micron standard — 3× over spec
Worn spindle bearings; cap loosening Why? — no PM check existed for spindle-bearing condition

Clamp & locator

Clamp force consistency Checked: nothing abnormal
Locator wear pattern Checked: nothing abnormal

Grinding wheel

Wheel balance Checked: within balance window
Wheel grade / bond Confirmed correct spec

M3 · Material

Material

Incoming workpiece characteristics

Dimensions

Pre-grind diameter variation Why? — upstream turning capability margin
Stock allowance for grinding Measured: within spec

Hardness

Heat-treat hardness variation Measured: within spec
Lot-to-lot consistency No abnormal lot trend

Surface finish (incoming)

Pre-grind surface roughness Measured: within spec

M4 · Method · Confirmed

Method

Process settings and routines

Grinding conditions

Feed, speed, depth-of-cut Checked: at standard
Wheel-speed-to-work-speed ratio Checked: at standard

Coolant concentration

Hydraulic oil contamination in coolant tank Why? — damaged hose was bleeding hydraulic oil into the coolant
Coolant concentration drifting off spec Why? — no PM check existed for coolant-line condition or concentration

Wheel dressing

Dressing frequency Checked: at standard
Dresser tool wear Checked: nothing abnormal

Two confirmed root causes. Maintenance verified excess runout of 30 microns on the spindle head versus a 10-micron standard — driven by worn bearings and a loosening cap. Separately, hydraulic oil was confirmed in the coolant system, traced to a damaged hose bleeding into the coolant tank. Everything else in the investigation came back within spec or as expected.

Step 5

Countermeasures

Twelve suspected causes were identified and considered. Two were confirmed as the primary causes.

In total, twelve potential suspected issues were identified and considered. Each was assigned to an owner, given a due date, and physically checked or measured. Of the twelve, two were found to be the primary causes — item 4 (worn spindle head bearings) and item 11 (coolant concentration off, traced to a damaged hose). The remaining ten came back clean. Verifying all twelve, rather than jumping at the most likely candidate, is what made the conclusion defensible.

#	Suspected Cause	Action Item	Responsible	Due	Finding
1	Dirt and contamination	Daily 5S & PM tasks	Tony (T/L)	11/2	No issues
2	Grinding wheel set-up	Set-up check	Tony (T/L)	11/4	Checked OK
3	Manual offset function	Check offset function	Tony (T/L)	11/4	Checked OK
4	Worn spindle head bearings	Installed anti-loose bearing cap	Ed (Maint)	11/5	Worn bearings / cap loosening
5	Clamp & locator damage	Clamp & locator check	Ed (Maint)	11/5	Nothing abnormal
6	Grinding wheel balance	Wheel check	Tony (T/L)	11/5	Nothing abnormal
7	Incoming part dimensions	Measure dimensions	Janet (QC)	11/9	Within spec
8	Material hardness	Measure hardness	Janet (QC)	11/9	Within spec
9	Surface finish (incoming)	Surface finish check	Janet (QC)	11/9	Within spec
10	Grinding conditions	Conditions check	Mary (Eng)	11/13	Nothing abnormal
11	Coolant concentration off	Replaced and rerouted hose	John (Maint)	11/12	Damaged hose
12	Wheel dressing	Check conditions	Mary (Eng)	11/13	Nothing abnormal

Why investigate all twelve. A common shortcut is to chase only the most likely cause. Toyota-style problem solving treats the cause-and-effect as a hypothesis list to verify, not a guess to act on. Items 4 and 11 came back as confirmed root causes; the other ten came back clean. That negative information has value too — it rules out competing explanations and prevents future re-litigation.

Step 6

Check Results

Did the countermeasures actually move the metric?

Defect % over Time · YTD Average through 11/13

Final Grinder

Rough Grinder

Final Grinding

8.7% → ~1.0%

Step change on 11/5 when spindle bearings were replaced and the anti-loose cap installed; second step down on 11/13 when the damaged coolant hose was replaced.

Rough Grinding

3.7% → ~1.0%

Same step pattern as Final Grinding. Both processes shared the spindle and coolant systems, so both responded.

Target Comparison

Both below 2.0%

The 2006 goal of < 2% was met on both processes within the two-week verification window. Sustaining the gain is now a standardization problem, not a problem-solving problem.

Step 7

Follow Up & Standardize

Locking in the gain and sharing what was learned.

#	Investigation Item	Responsible	Due	Status
1	Establish coolant check PM	Ops & Maintenance	11/15	Complete
2	Establish bearing check PM	Ops & Maintenance	11/15	Complete
3	Communicate findings to similar plants	Tom (Eng Mgr)	11/22	In progress
4	Discuss bearing issue with OEM	Tom (Eng Mgr)	11/29	Pending

A note on countermeasure type. The follow-up actions here are a mix of administrative countermeasures (establishing PM checks for coolant condition and bearing condition) and detection countermeasures (catching the abnormality earlier the next time it occurs). Neither is a true prevention countermeasure. The bearing failure mode in particular points back to the machine designer — the spindle-bearing arrangement allowed a cap to loosen and runout to drift to 30 microns before anyone could detect it. A genuine prevention countermeasure would require going back to the OEM to change the design, which is a realistic and common outcome of investigations like this one. Item 4 in the table above — "Discuss bearing issue with OEM" — is the start of that conversation. The distinction between administrative, detection, and prevention countermeasures is one I wish I had drawn in the original book; the corrected A3 is a good place to make it now.