The Trial Drawing Test: Measuring a Window Supplier’s Technical Response Before Award

1. Define what the exercise is meant to predict

Start with the award risk, not the test format. A buyer may need to know whether the proposed supplier can read an architectural package, reconcile schedules and details, expose interface gaps, control assumptions, route questions to competent people, identify evidence needs and return a revision-controlled response. Those behaviours are observable. The exercise cannot prove that every later engineer will perform, that the factory can manufacture the system, or that a suggested detail meets the project requirements.

Write the evaluation questions before selecting the drawing. ISO's Auditing Practices Group paper on evidence collection emphasizes planning evidence against criteria and warns that failure to find evidence does not automatically prove nonconformity. This buyer exercise is not a certification audit, but the discipline transfers: publish criteria, collect comparable evidence, record what was and was not demonstrated, and avoid a supplier-wide verdict from a narrow sample.

Separate technical intake from drafting productivity. The test should measure the quality of the path from input to controlled question and output. If CAD speed or a specific file format is commercially important, score it separately. Otherwise a bidder with reusable libraries and attractive rendering can outrank a bidder that correctly stops on a missing sill datum, unsupported performance claim or unresolved design responsibility.

2. Build one fair, representative and legally usable test pack

Choose a small part of the real product and project risk: for example, one operable unit at a wall transition, a coupled assembly with a schedule/detail conflict, or a door threshold where drainage, finished-floor level and accessibility inputs interact. Use information the buyer is authorised to share. Redact unrelated client, security and personal information without removing the technical context needed to answer safely.

Do not use an enormous tender set or deliberately impossible puzzle. A good pack contains enough information for useful progress and a controlled number of genuine omissions, ambiguities or pre-identified conflicts. State that the exercise includes unresolved information and that the supplier is expected to raise questions rather than guess. Give every bidder the same baseline, issue time, permitted clarification route, deliverables and evaluation window. Record any clarification sent to one bidder and decide whether fairness requires issuing it to all.

The pack should say what the buyer owns, what the bidder may reuse, confidentiality rules, whether any output will be paid for, and that submission does not transfer design liability or guarantee award. If the exercise could produce commercially valuable design work, obtain procurement and legal review. A trial should not extract a finished unpaid solution from multiple bidders under the label of due diligence.

3. Write the expected-issue key before responses arrive

The reviewers should identify the important issues and their consequences before seeing supplier answers. Otherwise the score will drift toward the most persuasive presentation. For each expected issue, record the controlling documents, nature of the conflict, affected opening marks, likely downstream consequence, authority needed and acceptable evidence of recognition. Include credit for a valid issue the reviewers missed; a bidder should not be penalised for finding a real problem outside the key.

Weight by consequence, not by count. A missing sheet title is not equivalent to an unresolved structural support path, inconsistent handing, missing glass safety basis, blocked drainage route or unexplained interface responsibility. Avoid embedding technical answers in the key unless a competent authorised reviewer has established them. Often the expected behaviour is to ask the right person, hold an affected output and state what evidence would close the question.

ISO's APG paper on design and development describes design as transforming requirements into product or service characteristics and highlights planning, inputs, interfaces, verification, validation and review. It is educational guidance rather than a mandatory buyer rubric. Use those concepts to ensure the key examines the response process, while project-specific architects, engineers, specifications and applicable rules supply the actual criteria.

4. Run the exercise as a controlled communication sequence

Use a stable issue channel and record four times separately: pack issued; receipt and revision acknowledged; first substantive clarification submitted; and final deliverable submitted. Response speed should never reward guessing. A fast acknowledgement can confirm the pack is readable and identify a missing attachment. A substantive response should show that a competent person has reviewed the technical question. Agree whether the clock pauses while the buyer answers a clarification.

The ISO/IAF APG customer-communication paper discusses product information, enquiries, contracts or orders including changes, customer feedback and relevant contingency requirements. It also states that the APG material is non-definitive educational guidance. For a trial, the useful adaptation is to observe whether communication gives correct product information, controls changed inputs and closes the loop rather than simply producing frequent messages.

Do not coach one supplier through the answer. The trial administrator can clarify rules and distribute common information. Technical reviewers should log interventions. If a defect in the pack makes the intended comparison unfair, issue a controlled addendum and adjust timing. Preserve both the original and revised baselines so the final assessment can distinguish a bidder error from an assessment-input error.

5. Require outputs that expose the supplier's reasoning

A drawing alone hides the path to the answer. Require a compact input register, marked-up source, issue/RFI log, assumptions register, responsibility map, proposed drawing or detail, evidence plan and final open-item list. Ask the bidder to distinguish information copied from the pack, its own proposed values, system-owner limits, project-authority decisions and items still unknown. The distinction matters more than graphical style.

Use placeholders instead of invented numbers where the project input is missing. If the bidder proposes an option, it should state the basis, affected requirements and who must approve it. If calculations, testing or system-provider confirmation will be required later, request a plan, not a fabricated result. A bidder should receive credit for declining an unsupported conclusion and identifying a safe path to closure.

Configuration control is part of the evidence. ISO 10007:2017 provides guidance on configuration management from concept to disposal and describes planning, identification, change control, status accounting and configuration audit. The trial is not an ISO 10007 conformity assessment, but its files should still show which input revision produced which response and how an addendum changed the output.

6. Score interface thinking without turning the supplier into the designer of record

Windows sit inside other systems. The response should identify the line between product geometry and building structure, air/water layers, insulation, cladding, interior finishes, sill pans or flashings, sealants, anchors, fire or acoustic assemblies where relevant, electrical controls, access, lifting and maintenance. Recognition does not confer design authority. The bidder should mark required inputs and responsible parties rather than producing an unsupported whole-wall solution.

Score whether the bidder follows paths. Where does load go? Where can water exit? Which layer provides continuity? What space is needed for movement and installation? Which dimension controls fabrication? Which unit weight affects hardware, packing and lifting? Which change would invalidate test or certification applicability? These are question paths, not universal details. The correct answer depends on the project, product and responsible professionals.

Pay attention to the relationship between proposal and evidence. A system may be capable of a particular configuration while the bidder has not supplied an applicable report. A generic detail may show intent while the site-specific substrate remains unknown. A product certificate can be current while the proposed variant falls outside its scope. Credit the bidder that separates these states and proposes an authorised closure route.

7. Use an anchored rubric and independent reviewers

Use at least two appropriately competent reviewers for material technical procurement. Score independently before discussion, then reconcile large differences with written reasons. An architect may detect coordination problems; a window engineer may understand system limits; a quality reviewer may see configuration drift; procurement should protect fairness and commercial boundaries. No reviewer should score outside their competence as if personal confidence were evidence.

Anchor each score. A zero can mean the issue was missed or contradicted. A middle score can mean the issue was identified but not tied to impact, authority or affected records. A high score can mean it was correctly located, prioritised, bounded and routed with a controlled next step. Add a critical-failure rule for an unsafe or dishonest action, but require documented review rather than automatic rejection for a harmless formatting error.

Avoid one total that hides fatal weaknesses. Keep critical-risk recognition, input/configuration control, RFI quality, interface reasoning, evidence discipline, output consistency and timing visible. A bidder with a lower total may remain viable with a defined condition; a high total cannot neutralise a critical unsupported assumption. Record any weighting chosen before receipt.

8. Convert the result into a bounded award decision

Write the conclusion in layers: demonstrated in the sampled test; partially demonstrated; not demonstrated; not tested; and requires post-award verification. Explain any critical concern and the exact record. Do not label the supplier generally ‘technical’ or ‘non-technical’. The proposed team may have performed well on input control but not shown configuration-specific performance evidence. Another may have found the most issues but failed to maintain one output revision.

Connect conditions to procurement gates. A viable supplier might need to nominate the actual technical lead, close a defined competency gap, complete a second configuration, provide a design-control procedure, demonstrate the drawing-to-BOM handoff or accept first-off witness points. A condition needs an owner, due date, acceptance evidence and consequence. Avoid vague promises of ‘more attention after award’.

Retain the issued pack, clarification history, response, score evidence, reviewer conflicts, decision and supplier feedback under the procurement identifier. ISO 10013 provides guidance on documented information and ISO 10005 on quality plans for a product, project or contract. Citing them does not make this exercise certified or ISO-compliant; the practical purpose is a reproducible decision and a usable post-award control plan.

Safe conclusion: The proposed team demonstrated the following behaviours on the issued trial configuration and revisions. The exercise did not validate the design, product performance, factory capability, project capacity or untested configurations.

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

• A Window Factory Quality-Control Checklist Buyers Can Actually Use
• An Aluminum Window Specification Checklist for International Projects
• How to Evaluate an Aluminum Window Manufacturer in China