Interim containment

Screen out the bad parts now, before the cause is even known.

What it measures

D3 sits between “we have a problem” and “we know the cause,” and it has its own engineering content:

• Scope: every place suspect material can be: the line, WIP, finished goods, in transit, at the customer, and on sister lines running the same process. Containment scoped only to the line that complained is containment with holes.
• Stopping the source: the line producing the defect stops, and only the lines verified clean keep running. Production continues on the healthy lines while the affected one stays down until the investigation clears it, so the plant is not manufacturing more containment work while it looks for the cause.
• Action set: 100% screening with a check that actually detects the defect, quarantine of suspect lots, rework where rework is validated, and certified (“clean-point”) stock so good material is positively identified, not assumed.
• Effectiveness check: a containment screen has an escape rate like any inspection. Verify the screen catches seeded or known-bad parts before trusting it with the customer.
• Exit criteria: defined up front: containment ends when the permanent corrective action (D5-D6) is verified effective, not when everyone gets tired of the inspection cost.

How to read the output

Track containment like a process of its own: parts screened, fallout rate, escapes. The fallout trend is live evidence for the investigation, if quarantine-and-screen fallout suddenly drops to zero while the suspected cause is still untouched, the defect was intermittent upstream and the containment boundary may be drawn around the wrong window. And watch the calendar: containment cost is roughly linear in time, so a D3 that has run for months without a D5 behind it is the signature of an investigation that stalled, the bridge became the building.

A real use case

A customer flags intermittent low open-circuit voltage in delivered cells, a self-discharge signature that takes days to manifest. Containment lands within hours, scoped by genealogy: the implicated assembly line is stopped while the two lines testing clean keep shipping, the affected production window is quarantined across finished goods and in-transit stock, sister-line cells from the same electrode lots are pulled in, and a 100% OCV-retest screen with a tightened delta-V limit goes up at outgoing inspection. The screen is validated against the customer-returned cells first, it catches them, and a clean point is declared at the last lot that passes extended retest. Fallout clusters in two electrode lots, which quietly hands the D4 investigation its first real clue while the customer ships nothing else suspect.

Common mistakes

• Letting containment become the fix. If 100% inspection still runs a quarter later, the 8D stalled at D3 and the defect is now a permanent cost center.
• Screening with a check that does not detect the defect, unvalidated containment is theater, and the escapes continue with a paper trail saying they cannot.
• Scoping to one line. The same process on a sister line, the same material in another product, and stock already in transit are all part of D3.
• No exit criteria, containment without a defined end either ends too early (escapes) or never (cost).
• Discarding the containment data. Screening fallout by lot, shift, and date is free evidence for the root-cause work happening in parallel.