Datum
All skills
QualityRev. Q3 2026

DFMEA builder

Facilitate and draft a Design FMEA: decompose functions, enumerate failure modes with effects and causes, score severity, occurrence, and detection honestly, and turn the top risks into specific design actions. Use whenever the user needs an FMEA or DFMEA for a design review, customer requirement, or APQP/PPAP package, asks what could go wrong with a design, or wants failure modes brainstormed for a component or assembly.

Stefan Vogt portraitStefan VogtQuality Systems LeadBosch logoBosch

Likes are peer vetting. Like this skill if it held up in your work.

DFMEA Builder

Builds a DFMEA that engineers actually use, not a checkbox spreadsheet: the value is in the failure mode brainstorm and the actions, so the skill front-loads those and keeps scoring honest.

Required inputs

  1. The item and its boundary: component or subsystem, what is inside vs. outside the analysis.
  2. Functions: what the item must do, with measurable requirements where they exist. If the user only has a part name, derive candidate functions and confirm.
  3. Operating environment and duty: loads, temperature, vibration, contamination, user behavior, life target.
  4. Format required: classic RPN (Sev x Occ x Det) or AIAG-VDA Action Priority. Default to whichever the user's customer requires; ask if unknown.
  5. Known field or test failures on similar designs. History is the best occurrence data there is.

Method

  1. Function decomposition. List functions as verb-noun with a requirement ("transmit 40 Nm without slip," "seal to IP67 for 10 years"). Every failure mode must trace to a function; this prevents the generic "part breaks" rows.
  2. Failure mode enumeration. For each function, enumerate modes across the standard categories: full loss, partial/degraded, intermittent, unintended function, and delayed/early. Then sweep physics-of-failure prompts: overload, fatigue, wear, creep, corrosion, thermal cycling, loosening, contamination, tolerance drift, material degradation, assembly error enabled by the design. Aim for completeness over elegance; merging comes later.
  3. Effects and severity. Trace each mode to the end effect at the next level and at the vehicle/system/user level. Severity scores against the end effect (10 = safety without warning, 9 = safety with warning or regulatory, 5 to 6 = degraded function, 2 to 3 = annoyance). Severity belongs to the effect and does not change with design controls.
  4. Causes and occurrence. Causes must be design-level and specific ("wall section below 1.2 mm allows creep at 85 C"), not restatements of the mode. Occurrence reflects the design controls preventing the cause: proven design with margin scores low, new material or new load case scores high. Push back on rows where everything scores 3.
  5. Detection / controls. Split prevention controls (design rules, derating, tolerance analysis, material spec) from detection controls (DVP&R tests, analysis, reviews). Detection scores how well the control finds the cause or mode before release, not after. "Inspection at supplier" is a process control and does not belong here; note it for the PFMEA.
  6. Prioritize and act. Rank by RPN or Action Priority, but apply the overrides: any Severity 9 or 10 with Occurrence above 1 gets an action regardless of RPN. Actions must name an owner type and be verifiable ("add fatigue test to DVP at 2x life, 6 samples," "increase fillet to R2 and rerun stress analysis"), never "monitor" or "review."

Output format

  • The FMEA table with columns matching the requested format, delivered as a spreadsheet-ready table
  • Top risks summary: the 5 to 10 rows that matter, each with its action in one line
  • A parking lot of items that belong in the PFMEA or DVP instead
  • Open questions where scoring needs data the team has and the assistant does not

Guardrails

  • Scores proposed here are drafts for the team to challenge; a DFMEA is a cross-functional judgment, and the output should say so on the sheet.
  • Refuse to score occurrence as low solely because "we've always done it this way" when the duty cycle or environment changed; call that out.
  • Keep severity honest: if a mode can plausibly cause injury, it scores in the 9 to 10 band even when it makes the sheet look bad.
  • Do not let the table balloon past usefulness: merge duplicate modes, and keep causes at the level the design team can act on.

More Quality skills