Failure Ledger¶
Status: Customer-evidence container Validation status: Initial customer-described themes captured; no artifact-seen or pilot-confirmed entries yet
This page is the core learning asset for the pivot. A real entry here is worth more than ten polished scenarios.
The ledger records release failures where an artifact was wrong, incomplete, ambiguous, or missing engineering intent. Each entry should explain what happened, why the normal process missed it, who suffered, what it cost, and what RapidDraft could have shown before release.
Failure Classes¶
| Class | Meaning | Example | Product implication |
|---|---|---|---|
| Consistency | two fields disagree | drawing Rev C, BOM Rev B | MVP-ready wedge |
| Completeness | required field or file missing | ERP material group blank, missing STEP | MVP-ready wedge |
| Standards | customer, product-family, or industry rule not satisfied | required coating note missing for part class | needs customer rules |
| Intent | artifact lacks something required by function | bearing seat has no fit callout | judgment layer, high trust bar |
| Historical | past lesson not reused | same supplier failure as 2019 repeats | judgment memory |
| Process | ownership or approval unclear | supplier package sent before quality approval | workflow/accountability, maybe not a check |
If most stories are consistency and completeness, the wedge may be enough but the moat is thin. If painful stories are intent or historical, the judgment layer has a stronger case.
Customer-Described Discovery Entries¶
These entries come from July 2026 calls and should guide demo selection. They are stronger than AI-generated bait scenarios, but they are not yet artifact-seen or pilot-confirmed.
Several entries now have account-specific workflow breakdowns: Volocopter, Bauer Kompressoren, Theegarten, Avilus, Rohde & Schwarz, and Deutsche Aircraft.
| ID | Evidence status | Source | Failure class | Pain | Product implication | Next proof |
|---|---|---|---|---|---|---|
RR-CUST-001 |
customer-described | Volocopter / Vinicius | standards / completeness | NX drawing setup varies by engineer across layers, dimensions, reference sets, and validation targets | drawing setup standards checker | one concrete NX standard matrix, validator output, and failing drawing |
RR-CUST-002 |
customer-described | Volocopter / Vinicius + Aaron | historical / search-reuse | engineers cannot search by actual geometry/dimensions across PLM/ERP | geometry and dimension search over CAD/PLM/ERP | one bolt/screw search example with expected hits and known false positives |
RR-CUST-003 |
customer-described | Rohde & Schwarz / Robin | standards / intent | functional reviews miss requirements and guideline compliance | requirements-driven review | one material/area/environment requirement and artifact |
RR-CUST-004 |
customer-described | Bauer Kompressoren / Wolf Goetze | historical / intent | senior engineers map unstructured custom requests to prior designs from memory | engineering search, design reuse, decision memory | one anonymized request, one expected prior design, and one near miss |
RR-CUST-005 |
customer-described | Deutsche Aircraft / Evgeny | completeness / process | CATIA/3DEXPERIENCE drawings loop through manual completeness checks before SAP release | drawing completeness and release-cycle reduction | one checklist, one drawing package, and one SAP/BOM metadata export |
RR-CUST-006 |
customer-described | Ulrich | standards / intent | DFM checks must reflect actual machine park and quantity | machine-aware DFM and quantity-dependent costing | one machine constraint and one quantity-sensitive rule |
RR-CUST-007 |
customer-reacted | Dennis Schmitz | standards / intent | review should account for material density, AM orientation, CO2 footprint, process, rendering, and drawing-analysis errors | material/process-aware review | one part with material/route/scrap calculation |
RR-CUST-008 |
customer-mentioned, needs confirmation | Theegarten/Marcel signal, Bauer fit | completeness / standards | ERP/CAD metadata and engineering master-data fields are likely manually maintained | Engineering Master Data Assistant | new-part creation walkthrough: fields, systems, time, owner |
RR-CUST-009 |
customer-described | Avilus / Franz signal | standards / completeness | drawings should match company standard; notes mention missing dimensions, missing surfaces, missing material, bend parameters, Teamcenter | drawing standards and completeness checker | one drawing standard, one sheet-metal drawing, and one Teamcenter package/export |
RR-CUST-010 |
customer-described | Deutsche Aircraft / Jose V. C. | consistency / historical / process | correct CAE load report was lost, analysis used wrong loads, and the correct report was found later | stress input readiness and load provenance | one sanitized load report, one analysis input table, and one expected correct source |
Ledger Template¶
Use this format for every real discovery story:
| Field | Description |
|---|---|
| Failure ID | RR-YYYY-NNN |
| Source | interview, artifact, pilot, internal hypothesis |
| Evidence status | unvalidated, customer-described, artifact-seen, pilot-confirmed |
| Industry | e.g. packaging machinery |
| Product family | e.g. wrapping module |
| Release event | e.g. supplier RFQ package |
| Persona affected | e.g. manufacturing engineering lead |
| Artifact | drawing, BOM, CAD, PDM metadata, ERP template, PPAP binder |
| What was wrong | exact defect |
| Why normal process missed it | process or technical gap |
| Who found it downstream | role |
| When found | supplier RFQ, manufacturing, inspection, field, audit |
| Cost or delay | quantified if possible |
| What RapidDraft could show | issue card and evidence |
| Required data | exact files and fields |
| Detection method | rule, extraction, RAG, geometry, customer rule, human review |
| False-positive risk | low, medium, high |
| Product implication | check, workflow, rule-learning, not productizable |
Seed Hypothesis Entry¶
This entry is not customer-confirmed. It is included only as a template example.
| Field | Value |
|---|---|
| Failure ID | RR-HYP-001 |
| Source | AI-generated packaging machinery scenario |
| Evidence status | unvalidated |
| Industry | packaging machinery |
| Release event | Konstruktionsfreigabe to Einkauf and Fertigung |
| Artifact | shaft drawing |
| What was wrong | 35 mm bearing seat has no fit/tolerance callout |
| Why a consistency checker missed it | title block, BOM revision, material, and part number all agree |
| How a senior engineer recognizes it | shaft function and mating bearing imply a controlled fit |
| Downstream consequence | ambiguous machining target, scrap, assembly delay |
| Required data | drawing geometry, feature classification, mating context, customer rule or past example |
| Detection method | not generic MVP; possible with customer-provided rules/examples |
| False-positive risk | high |
| Product implication | judgment layer, not first demo opener |
Seed Commodity Checks¶
These are candidate wedge checks, not customer-validated pain.
| ID | Check | Failure class | Required input | Feasibility |
|---|---|---|---|---|
RR-CONS-001 |
Drawing revision matches BOM revision | consistency | title block + BOM revision column | high |
RR-CONS-002 |
Drawing part number matches BOM part number | consistency | title block + BOM row | high |
RR-COMP-007 |
All BOM children have part numbers | completeness | BOM | high |
RR-COMP-008 |
All make parts have released drawings | completeness | BOM + file inventory | high/medium |
RR-COMP-011 |
ERP item group populated for every new part | completeness | ERP item template/export | high |
RR-PKG-015 |
Supplier package includes PDF, STEP, BOM, and process notes | completeness | package inventory | high |
RR-CONS-010 |
STEP file revision matches released drawing revision | consistency | STEP filename/metadata + drawing revision | medium |
RR-CONS-003 |
Drawing material callout matches BOM material field | consistency | note extraction + BOM material | medium |
Intake Rule¶
Do not add polished invented scenarios as ledger entries. Add real failure stories, customer corrections, or pilot findings. If the entry is hypothetical, label it as hypothetical and keep it out of validation counts.
Open Questions¶
- Which failure class appears most often in real conversations?
- Which failure class creates the clearest budget owner?
- Which real failure can be converted into the first one-check demo?
Sources¶
- Customer Discovery Use Cases
- Volocopter Use Cases
- Bauer Kompressoren Use Cases
- Theegarten Use Cases
- Avilus Use Cases
- Rohde & Schwarz Use Cases
- Deutsche Aircraft Use Cases
_sources/pivot_files_claude/Judgment_Failure_Ledger.md_sources/pivot_files_claude/Validation_Plan.md_sources/pivot_gpt_files/critique_output.mddocs_pivot/_sources/usecases-erpdrawing/rapiddraft-usecases-scenarios.mddocs_pivot/_sources/usecases-erpdrawing/bauerkompressorsearch.md