Product, Transport, Installation or Use? A Window Failure Responsibility Matrix

1. Define the observed condition and affected population

Identify opening marks, unit and component configurations, locations/elevations/orientations, discovery time and operational state. Describe symptoms neutrally: contact location, movement, visible gap, water path observation, crack pattern without causal label, finish appearance, loose item, noise or missing component. State measurement methods and uncertainties only from controlled plans.

Determine population without assuming common cause. Compare same type, batch, elevation, installer, shipment, storage zone, interface detail, use pattern and unaffected controls. Record inspected and uninspected populations, concealed limitations and selection method.

Apply immediate safety or containment only through authorised project procedures. Preserve pre-disturbance evidence and changes. A safe temporary action can alter the condition; record it without treating improvement or persistence as proof of cause.

2. Keep a multi-hypothesis matrix open

Create separate hypothesis families: product design/configuration, component/material/manufacture, packaging and transport, storage, installation and adjustment, building/interface and adjacent trades, environmental exposure, operation/use, cleaning/maintenance, later modification and combined mechanisms. The categories are prompts, not findings.

For each hypothesis state a possible mechanism as a question, predictions that would be expected if it were relevant, evidence that could support or weaken it, evidence owner and competent discipline. Include disconfirming evidence. Do not score responsibility percentages before mechanisms and contract scope are independently reviewed.

Allow interaction. Transport movement may damage a package that installation inspection should have detected; interface geometry may amplify a product drainage limitation; maintenance access may depend on design. Technical contribution and legal responsibility are not identical.

3. Reconstruct product and custody states across the lifecycle

Map released design and requirements, material/component production, factory assembly and inspection, pack/closure, loading and transport, arrival inspection, storage, installation, interface closure, commissioning/handover, occupancy/use, cleaning/maintenance and modifications. At each state record identity, custodian, configuration, observed condition and change.

Separate absence of evidence from evidence of absence. A factory photo may not show concealed machining; arrival inspection may sample package exteriors; commissioning may not test the later exposure. State coverage before using a clean record to weaken a hypothesis.

Track revisions and deviations: drawing changes, substitutions, NCR/rework, repacking, site adjustments, protection removal and repairs. Do not assume the latest document describes the affected unit.

4. Map technical interfaces to contract boundaries without deciding liability

Collect signed contract and incorporated specifications/drawings, approved submittals, scope and responsibility matrices, Incoterms and carriage/insurance documents where relevant, inspection and acceptance clauses, installation/interface allocation, protection/storage duties, handover/manuals, warranty/component warranties, notice and remedy routes, governing law and authorised changes. Do not use tender drafts as signed scope.

For each physical interface record who designed, supplied, approved, installed, inspected, protected, operated and maintained according to the controlled documents. Several parties may hold different duties. A technical deviation does not automatically establish breach, causation, waiver, acceptance or cost responsibility.

Keep statutory or mandatory duties and third-party rights for legal review. Contract labels such as by others, approved or warranty exclusion require interpretation in context. The matrix exposes documents and gaps; it does not adjudicate.

5. Commission scoped expert review and controlled testing

Write an investigation brief that identifies units, conditions, hypotheses, questions, available records, evidence custody, access/safety constraints, authority and prohibited destructive work. Select disciplines appropriate to product, glass, hardware, corrosion/materials, packaging/transport, façade/interface, installation, building science or other issues. Declare conflicts and information limits.

Use staged methods: document and chronology review; non-destructive observation and controlled measurements; manufacturer or laboratory examination; then invasive opening or testing only if authorised by responsible safety, design, contract, insurer and legal routes. Define method, calibration, environmental conditions, sample, acceptance source and restoration before action. This article provides no test pressure or method.

Require reports to separate facts, assumptions, methods, limitations, findings, alternative explanations and conclusions within the expert's discipline. Preserve raw data and contrary evidence. One failed test can show performance under test conditions without allocating responsibility.

6. Move from evidence to separate technical and responsibility decisions

Maintain a live matrix showing each hypothesis as open, supported, weakened, not testable or resolved within a stated scope, with evidence references and expert authority. Do not translate confidence into blame percentages. Record combined mechanisms and residual uncertainty.

Route technical disposition separately from responsibility: safe containment, repair/replacement engineering, reinspection and population action can proceed under authority while contract, insurer and legal parties assess cost and remedy. A practical settlement does not rewrite technical causation; a technical finding does not automatically decide legal allocation.

Close only with affected population, evidence inventory, expert reports, approved technical action and verification, contract/insurance/legal decision where reached, notices and residual monitoring assigned. Preserve disagreement and unknowns. Never call silence or an expired commercial discussion proof that one party was responsible.

7. Test each hypothesis against counterfactuals and population patterns

For each supported-looking hypothesis, ask what else should be observed if it were materially relevant and what findings would be surprising. A proposed manufacturing mechanism may predict a batch or geometry pattern; transport may predict package-position or contact evidence; storage may predict exposure distribution; installation may correlate with installer, opening geometry or interface; use and maintenance may correlate with access or event records. These are investigation questions, not diagnostic rules.

Compare patterns without erasing confounders. Same elevation can share weather and installer; same batch can share shipment and storage; one opening type can share both hardware and user behaviour. Record competing grouping variables and missing populations before treating correlation as cause. Absence in a small inspected sample should not close a mechanism without a justified sampling and expert basis.

Use counterfactual records carefully: unaffected units, retained samples, factory controls, pre-install photographs and post-repair observations. A repair that removes a symptom can support a mechanism question but may change several variables at once. Record every intervention and do not call a successful practical remedy proof of original responsibility.

At decision meetings, require each proposed cause statement to cite identity, population, lifecycle timing, method, expert discipline, alternatives considered and limitations. Require every responsibility statement to cite the separate contract/legal/insurance authority. If those chains cannot be completed, label the point unresolved rather than converting commercial urgency into certainty.

FAQ

Continue the project review

Connect this guide with the proposed window and door systems, then review how Cavorlen approaches manufacturing and quality control. When your schedule is ready, send it through the project review form.

Related buyer guides

• Evidence for a Window Damage Claim: Opening Mark, Context and Timeline
• A Window Factory Quality-Control Checklist Buyers Can Actually Use
• An Aluminum Window Specification Checklist for International Projects