Last reviewed: June 19, 2026
Reducing scrap rate or improving OEE by 20 to 40 percent on the targeted loss category is achievable in 90 days of consistent execution. Most programs fail not because the timeline is wrong but because the data collection phase is treated as progress rather than as preparation.
Key takeaways:
In plants we have worked with, consistent execution of a targeted scrap or OEE improvement sequence produces a 20 to 40 percent reduction in the targeted defect category or loss type within 90 days.
The first 30 days are measurement, not improvement. Until you know where the loss is coming from, any action is a guess.
Most improvement programs stall at the data collection phase: the number looks better for two weeks and tracking fades when the supervisor has a busy month.
A fix is not closed until it has held for 30 days of verified first-pass yield data. Verbal agreement that the fix worked is not evidence.
THE REALISTIC TIMELINE FOR SCRAP REDUCTION
A 90-day timeline is the right unit for a targeted scrap or OEE improvement. Shorter timelines do not produce enough data to distinguish a real fix from random variation. Longer timelines lose the accountability that makes teams execute.
The 90 days break into three phases of roughly equal length. The first phase is measurement and identification. The second is targeted intervention. The third is verification and standardization. The phases are sequential: phase two is not useful until phase one is complete, and phase three is meaningless without a documented countermeasure from phase two.
| PHASE | DAYS | WHAT YOU ARE DOING | WHAT YOU SHOULD BE SEEING | RED FLAGS IF NOT |
|---|---|---|---|---|
| Measure | 1 to 30 | Logging all scrap and downtime events with reason codes; building a 30-day Pareto | Data collected consistently by shift; top defect category identified before day 30 | Supervisors reconstructing data at end of shift; more than 20% of events coded as "other" |
| Intervene | 31 to 60 | Running 5-Why on the two or three largest events in the top Pareto category; implementing countermeasure at the process level | Targeted defect category trending down; rework and scrap both tracked | 5-Why stopped at "operator error"; countermeasure was a verbal reminder, not a process change |
| Verify | 61 to 90 | Tracking first-pass yield daily on the targeted category; holding 30 consecutive days at or above target before declaring the fix closed | Improvement has held for three to four weeks; countermeasure embedded in standard work | Number improved for two weeks and returned to baseline; fix not documented |
WHAT THE FIRST 30 DAYS LOOK LIKE
The first 30 days are measurement, not improvement. This is the phase most plants rush through, and the rushing is why the improvement does not stick.
The work in days 1 to 30 is collecting reason-coded scrap and downtime data at the work center. Not at the end of the shift. Not reconstructed from memory on Friday. At the machine, at the time the event occurs, by the operator or supervisor who was present.
By day 20 you should be able to name the top defect category in your Pareto. If you cannot, it is usually because "other" or "miscellaneous" is absorbing a large share of events. Pull the "other" events and recategorize them. The information is in the data; it is just not labeled yet.
The Pareto from this phase tells you where to work. Everything in phase two follows from it. If the Pareto is based on dirty or reconstructed data, the phase two countermeasure will target the wrong problem and the 90-day result will be disappointing.
In plants we have worked with, 20 to 35 percent of scrap occurs in the first 15 minutes of a production run. If your Pareto shows setup and changeover errors in the top two categories, the intervention in phase two is a first-off inspection procedure, not a machine repair or a new tool.
WHAT 30 TO 90 DAYS LOOKS LIKE
Phase two begins when you have a clear top category from the Pareto. Pick the two or three largest individual events in that category and run a 5-Why on each one. Do this in front of the machine, with the operator who ran the job, within one week of when the event happened. Details fade. Memory is not a reliable substitute for presence.
The 5-Why will surface a system failure in most cases: a missing setup parameter, an inspection step that was skipped, a tooling condition that should have been flagged the prior shift but was not. The countermeasure goes into the system, not into a verbal reminder to the operator.
A process-level countermeasure changes the setup sheet, the inspection form, or the maintenance schedule. A verbal reminder changes nothing after two weeks. In plants we have worked with, countermeasures documented in standard work sustain. Countermeasures communicated at a shift meeting are gone within 30 days.
By day 60, the targeted defect category should be trending down. Not necessarily at target yet, but moving. If it is not moving at all by day 60, the countermeasure either addressed the wrong root cause or it was implemented incorrectly. Go back to the 5-Why and look for what was missed.
Phase three is verification. From day 61 to 90, track first-pass yield daily on the targeted category and look for 30 consecutive days of data at or above target before declaring the fix closed. The number may dip in the first week of phase three as the improvement settles. A single good week is not a closed fix.
WHY MOST PROGRAMS STALL AFTER THE FIRST WIN
The pattern repeats across operations of every size: a team runs a focused improvement on one defect category, gets a visible result in the first six to eight weeks, presents the result in a monthly meeting, and then tracking fades. Three months later, the scrap rate is back near its original level.
In plants we have assessed, the programs that stall do so for one of three reasons.
The first is that the fix was never embedded in standard work. The countermeasure lived in someone's memory or in a presentation from the monthly review. When that person's attention moved on, the countermeasure moved with it.
The second is that tracking stopped when the result was achieved. The team declared victory and removed the measurement. Without ongoing tracking, there is no signal when the process drifts back.
The third is that the supervisor was replaced or reassigned, and the incoming supervisor had no handoff documentation describing the fix, why it was made, and how to tell if it is still working.
HOW TO KNOW THE FIX ACTUALLY HELD
A fix is closed when three things are true: the targeted metric has held at or above target for 30 consecutive days; the countermeasure is documented in a standard work document currently in use; and a different operator or supervisor who was not part of the original investigation can describe the fix and the reason for it from that document alone.
That last test is harder than it sounds. If sustaining the fix requires the presence of the person who ran the 5-Why, it will not survive a personnel change or a six-month gap in attention.
Verify with first-pass yield data, not with scrap data alone. A scrap rate improvement that masks a rework increase is not a real improvement: the parts are still failing inspection on the first pass, they are simply being caught and repaired before they reach the final inspector rather than being scrapped. First-pass yield captures what scrap rate misses, and the full tracking sequence is covered in the manufacturing scrap reduction guide.
The 90-day sequence works when the phases are distinct, the data is real, the countermeasures are documented, and the verification period is taken seriously. Compressing the timeline or skipping the documentation does not save time. It produces a result that will not be there at month six.