Prevent recurrence

Lock it in so the same failure can't come back.

What it measures

Corrective action (D5-D6) fixed this occurrence; D7 attacks the conditions that let it happen, at three levels:

• Error-proofing the step: poka-yoke. Interlocks, fixtures that only accept correct orientation, recipes that refuse out-of-window setpoints. The hierarchy matters: physical impossibility beats alarm beats procedure beats training.
• Systemic updates: control plan, PFMEA occurrence/detection scores, work instructions, maintenance schedules, and incoming-material specs all revised to reflect what the investigation learned. If the documents don’t change, the organization hasn’t.
• Horizontal deployment: the same weakness usually exists on sister lines, sister plants, and similar products. A read-across review answers “where else could exactly this happen?” and applies the countermeasure there before it fires.

How to read the output

Judge a D7 by what would happen if everyone who worked the 8D left tomorrow. If the protection survives, because it lives in an interlock, a fixture, a control plan an auditor checks, D7 is real. If it lives in “the team knows now,” it is D5 with good intentions. The other read is breadth: count the lines and products the lesson was deployed to versus the ones that share the failure mode. A recurrence on a sister line six months later is not a new problem; it is this 8D’s D7 scored as a miss.

A real use case

A slurry batch is ruined by a mixing-sequence error, conductive additive charged before the binder solution, an order the recipe sheet specifies but nothing enforces. D5 retrains the crew and corrects the batch records. D7 is where the plant actually changes: the mixer recipe system is reconfigured so material addition steps unlock only in sequence, the wrong order is no longer possible, not just prohibited, the control plan and PFMEA detection scores are updated, and the read-across finds the same unenforced sequence on the sister mixer and on the pilot line, which get the same interlock the same week. The error that cost a batch can now not recur anywhere in the building, including with operators not yet hired.

Common mistakes

• Stopping at D6 under time pressure. The 8D closes, the system that produced the failure is intact, and the same failure returns wearing a different lot number.
• Training as the only countermeasure. Training decays with turnover and time; it is the weakest rung of the prevention hierarchy, not the whole ladder.
• Skipping horizontal deployment, fixing one line and leaving its identical sister running the identical weakness.
• Updating the control plan without verifying anyone audits it. A document change nobody checks is a prevention nobody has.
• Not feeding the lesson back into FMEA scores and the defect library, so the next team rediscovers the failure mode from scratch.