A tower photograph and a client logo may show that a project exists. They do not show which entity supplied which windows, under what configuration, or how the work performed.
Past-project evidence is useful only after two different questions are answered. First, is the reference authentic and attributable: did the named supplier entity, plant or proposed key team actually perform the stated role? Second, is it relevant: are the product, complexity, performance pathway, volume, programme, market and responsibility comparable enough to inform the new decision? A reference call can help, but it must be client-authorised, factual and structured. The result should be a bounded relevance assessment, not a testimonial assembled by the supplier and repeated by the buyer.
Reference boundary: This method does not certify project history, ownership, product performance, client satisfaction or future capability. It does not authorise publication of client names, contact details, drawings or photographs. Obtain consent and comply with confidentiality, privacy, procurement and competition requirements.
1. Convert the project name into claims that can be checked
Ask the supplier to state the exact legal entity that contracted or supplied, its role, factory or plants used, project location, client type, completion period, product scope, approximate quantity range if authorised, key configurations, performance or certification route, installation responsibility, subcontracted processes and proposed reference contact. A brand name is not a legal entity; ‘supplier’ can mean designer, system licensor, fabricator, trader, installer or component provider.
Separate factual claims from evaluation. ‘Completed in 2024’, ‘supplied operable aluminium windows’ and ‘the assembly plant was X’ can be checked. ‘Premium’, ‘successful’ and ‘zero complaints’ need definitions and evidence. A photograph can support visual context but rarely establishes the supplying entity, contract role, configuration or completion. Ask what original record links the supplier to the claim.
Record the supplier's permission basis. Client names, contacts, contract values, drawings and problem records may be confidential. The supplier should confirm what can be disclosed and obtain client authorisation before a buyer makes contact. A refusal to disclose can leave evidence limited without proving deception; use anonymised records, independent verification or a less sensitive reference route where proportionate.
| Claim field | Supplier statement | Primary record requested | Permitted disclosure |
|---|---|---|---|
| Legal supplier entity | [exact name/identifier] | Contract/PO/official project record | full / redacted / independent confirmation |
| Role and scope | designer/fabricator/trader/installer/component provider | Scope schedule, invoice, handover or client confirmation | [record] |
| Factory and external providers | [names/sites/processes] | Production, inspection and supply-chain records | [record] |
| Dates/status | award, supply and completion period | Dated transmittals/delivery/handover records | [record] |
| Product/configuration | opening types, system, glass, hardware and size range | Approved schedule/drawings/BOM extracts | [record] |
| Performance/evidence route | tests/listings/project criteria | Applicable reports/directories/approvals | [record] |
| Reference contact | name, employer, project role | Client-authorised contact route | contact / written-only / unavailable |
| Problems/corrective actions | factual examples | NCR/claim/closeout extract | [record] |
2. Define similarity before the supplier chooses its showcase
Similarity is multidimensional. A project can share product appearance while differing in opening types, maximum sash mass, wind and water requirements, thermal or acoustic pathway, glass safety basis, hardware, finish exposure, certification regime, interface design, installation scope, volume, logistics and programme. Decide which dimensions predict the new contract and their minimum relevance before reading the reference list.
The current U.S. Federal Acquisition Regulation treats past performance as an indicator rather than a guarantee and requires consideration of currency, relevance, source, context and general trends. It also recognises that relevant information may concern predecessor companies, key personnel or major subcontractors. That is a public-procurement rule, not a mandatory private window-buying procedure, but it supplies a useful discipline: identify whose experience is being relied upon and why it is relevant.
World Bank procurement guidance and standard qualification documents likewise use specific similar-contract experience and define similarity through factors such as physical size, complexity, methods or technology. Do not copy public tender thresholds into a private purchase. Use the principle to predefine a proportionate relevance matrix and avoid selecting only by project prestige or headline value.
| Dimension | New requirement | Reference facts | Relevance and gap |
|---|---|---|---|
| Entity/team/plant | Who will perform the new order | Who actually performed reference | same / transferred / different / unknown |
| Product family/opening | [record] | [record] | high / medium / low / unknown |
| Configuration extremes | Size, mass, coupling, glass and hardware risks | Actual supported range | Do not infer from average units |
| Performance/evidence | Methods, values and acceptance route | Actual project route | Comparable method and configuration? |
| Destination/regime | Code, certification, language and document needs | Reference jurisdiction | Transfer limits |
| Scope/responsibility | Design, supply, install and interfaces | Supplier's actual role | Relevant work was or was not performed |
| Volume/programme | Release batches and constraints | Reference quantity/dates where shareable | Capacity inference remains bounded |
| Recency/change | Current proposed resources | Reference date and subsequent changes | Current applicability |
| Outcome evidence | Acceptance, issues and service needs | Factual closeout/performance record | Context and remaining unknowns |
3. Attribute the work to the right entity, plant and team
The first identity match is legal: compare the name on the claimed contract, purchase order, invoice or authorised client confirmation with the bidding and proposed manufacturing entities. If the past project was delivered by a parent, affiliate, former employer, joint venture or subcontractor, record that structure and the share of work actually performed. Common ownership or branding does not automatically transfer operational experience.
The second match is operational. Which plant assembled frames? Who fabricated insulating glass? Who applied the finish? Who produced drawings and calculations? Who installed? Which proposed people participated, in what roles and for what period? A supplier may legitimately rely on key personnel or major subcontractors, but the new proposal must show that those resources are actually committed and that their prior work is relevant.
Use stable records and independent contact routes. A supplier-created case-study PDF can organise the claim but should not be the only evidence. Compare lawful redacted contract or delivery documents, controlled drawings, inspection reports, product labels/listings, project directories or public authority records where genuinely applicable. Absence from a public website is not proof the project or role is false.
| Reference role | Entity/person claimed | Evidence of actual participation | Link to new proposal |
|---|---|---|---|
| Contracting/selling entity | [record] | Contract/PO/client confirmation | same entity or documented relationship |
| Window assembly plant | [record site/operator] | Production/inspection/traceability record | same plant and current controls? |
| Technical design/detailing | [record team/entity] | Authored/reviewed transmittals and revisions | named proposed lead/availability |
| Glass/finish/external process | [record providers] | BOM/purchase/certification/inspection evidence | same route or controlled change |
| Installation/interface work | [record] | Method, inspection and handover evidence | included or excluded in new scope |
| Key personnel | [name/role with permission] | Project organisation and authorised confirmation | committed role, deputy and employer |
| Warranty/after-sales | [record entity] | Issue/closeout route | new obligation owner |
4. Build an evidence stack rather than trusting one artefact
Use corroboration. A contract extract can support identity and scope but not completed performance. Delivery documents can support movement of goods but not acceptance. A completed drawing set can support technical involvement but not fabrication. A test report can support a named specimen and result but not every installed unit. A client reference can provide context but may rely on memory or cover only the contact's workstream.
Prefer records created during the project over a retrospective marketing pack, while checking authenticity, authorisation and completeness. Connect them by project code, date, entity, opening marks, drawing revision, shipment or unit identity. Redact prices, unrelated parties and proprietary details. If the supplier cannot share a document, allow an authorised client or independent reviewer to confirm specified fields without giving the buyer the whole file.
Apply the ISO/IAF evidence-collection caution: evidence must be evaluated against criteria, and failure to locate evidence is not automatically proof of failure. Use ‘supported’, ‘limited’, ‘unresolved’ and ‘contradicted by identified evidence’ rather than ‘real’ or ‘fake’ unless a competent finding warrants it.
| Evidence | May support | Cannot establish alone | Cross-check |
|---|---|---|---|
| Contract/PO extract | Entity, customer, scope and date | Completion, quality or exact manufacturing route | Client confirmation and delivery/closeout |
| Approved drawings/schedule | Configuration and technical participation | Actual production or installation | BOM, inspection and release records |
| Invoice/delivery/shipping record | Commercial supply or movement | Acceptance, performance or defect status | Opening/package trace and client record |
| Inspection/test report | Named sample/unit, method and result | All units or new-project applicability | Configuration, lab/scheme and acceptance |
| Project photograph | Visual context at a point in time | Supplier identity, hidden build-up or outcome | Metadata/authorisation and project records |
| Client reference | First-hand role-specific facts | Everything outside contact's knowledge | Structured questions and documentary corroboration |
| Public listing/news | Public existence and selected facts | Contract role or product detail not stated | Primary project/client and supplier records |
5. Contact references through an authorised and independent route
Ask the supplier to obtain permission for contact and to identify the referee's employer, role during the project, subjects they can discuss and permitted channel. Verify the channel independently where possible: an organisation's official switchboard, corporate directory or a contact supplied directly by the client is stronger than an unverified mobile number in a sales deck. Do not pressure a referee to disclose confidential performance, personal data or commercial terms.
Send a short factual agenda in advance. Confirm the referee's project role and period before asking about the supplier. Ask what the supplier entity actually did, which product or interface scope the referee observed, whether delivery and issue handling met defined expectations, what changed, and whether the account is based on direct knowledge. Invite ‘not known’ answers. Avoid leading questions such as ‘Was the supplier excellent?’
Record the date, participants, permissions, question, factual response, qualification and any follow-up evidence. Give the supplier a fair opportunity to explain material negative or inconsistent information. The FAR source-selection framework explicitly considers context and allows offerors to identify problems and corrective actions; that is a sound fairness principle even outside U.S. public procurement.
| Question | Why it matters | Acceptable bounded answer | Follow-up |
|---|---|---|---|
| What was your role and period on the project? | Tests first-hand knowledge | Exact role/date or not known | Limit later answers to observed scope |
| Which legal entity and team did you engage? | Tests attribution | Entity, names/roles where permitted | Compare with proposal |
| What product/scope did they provide? | Tests relevance | Opening/system/supply-install scope at useful level | Request authorised record if ambiguous |
| Which difficult requirements were involved? | Tests complexity | Factual configuration/process/interface | Map to relevance matrix |
| Were agreed programme and quality controls met? | Tests performance in context | Defined outcomes, exceptions and causes | Ask what evidence/contact knows |
| What problems occurred and how were they handled? | Tests corrective behaviour | Issue, ownership, response and closure | Do not treat any problem as automatic failure |
| What has changed since then? | Tests currency | Team, plant, product or process change | Revalidate current resources |
| May this statement be retained or quoted? | Controls disclosure | Permission scope | Restrict record/use accordingly |
6. Evaluate problems and corrective action, not a perfect story
Every material project has constraints and changes. A reference with no questions, delays, damage, nonconformities or service events may reflect a narrow contact or curated narrative. Ask for one representative difficulty and trace detection, containment, technical and commercial authority, correction, verification and prevention. The objective is not to collect gossip or blame; it is to understand control under stress.
Context matters. A delayed release caused by late client approval differs from a supplier missing a confirmed commitment, yet both require communication and programme control. A broken unit in transit requires packing, carrier, receipt and responsibility evidence before a cause conclusion. A design revision may be normal project development or uncontrolled change. Record facts, sources and unresolved attribution separately.
Look for trends across more than one reference when risk justifies it. Repeated late technical responses, unannounced component substitutions or unresolved closeout may be more predictive than one severe but well-corrected event. Equally, three references selected by the supplier do not form a statistically representative population. State the selection bias.
| Field | Reference fact | Evidence/source | New-order implication |
|---|---|---|---|
| Reported issue | Symptom and affected scope | First-hand record/contact | Comparable risk? |
| Detection/notification | When and by whom | NCR, minutes, claim or inspection | Visibility and escalation control |
| Immediate containment | Work/product held or protected | Dated action evidence | Can current team contain similarly? |
| Cause/authority | Confirmed cause or unresolved hypotheses | Competent conclusion | Do not convert allegation into fact |
| Correction/disposition | Action and authorised acceptance | Closure record | Relevant technical/service capability |
| Prevention/change | Process, design, provider or training change | Updated controlled record | Is control still current? |
| Outcome | Closed/open and observable consequence | Client/supplier evidence | Condition, monitoring or no inference |
7. Handle new suppliers and non-disclosable work proportionately
A company with no directly comparable disclosed project is not automatically incapable. It creates an evidence gap that can be addressed by other controls: proposed key-person experience attributable to the individual; major subcontractor experience tied to a committed role; a controlled trial drawing; applicable product evidence; a paid prototype; first-off inspection; a pilot batch; staged release; stronger supervision; or a smaller initial scope. Do not present these as equivalent to a completed project reference.
Public procurement guidance commonly requires relevance and proportionality so that experience criteria do not unnecessarily exclude smaller or newer suppliers. The buyer should likewise ask only for references that meaningfully predict the risk. Requiring a prestigious local hotel logo when the actual need is a repeatable window configuration can distort competition and encourage superficial claims.
Confidential reference work can be verified through a named independent adviser or client confirmation of limited fields. Record who performed the verification, the fields checked and what remained unseen. ‘Verified by third party’ is not useful without a bounded scope and conflict check.
| Gap | Alternative evidence/control | What it supports | Residual risk |
|---|---|---|---|
| New legal entity | Named committed team and plant evidence | Resources may carry relevant experience | Organizational processes not yet demonstrated together |
| Confidential project | Client-authorised limited confirmation or independent review | Specified identity/scope fields | Underlying detail remains unseen |
| New product family | Trial drawing, system-owner review, prototype/test plan | Current technical pathway | No completed field history |
| New factory/provider | Process audit, sample, first-off and staged release | Sampled execution at new route | Volume and long-term consistency |
| No comparable scale | Pilot batch and resource-loaded capacity plan | Controlled ramp evidence | Full programme still conditional |
| No client contact | Corroborated project records and independent project source | Documented facts only | No first-hand performance context |
8. Issue a bounded reference disposition
Conclude separately on authenticity/attribution, relevance, performance context, currency and availability of the relied-upon resources. ‘Supported’ can mean primary records and an authorised contact corroborate the stated entity, role and selected facts. ‘Limited’ can mean the project exists but the supplier's exact role, configuration or outcome is only partly evidenced. ‘Unresolved’ means necessary evidence was unavailable. ‘Contradicted’ requires a cited conflict and a fair response process.
Do not translate one supported reference into ‘approved supplier’. Combine the result with legal identity, certificate scope, technical response, subcontracting, product evidence, factory process, capacity, commercial, payment and contract reviews. Define conditions: retain a key person, use the named plant or subcontractor, repeat a test, disclose a route change, provide first-off evidence or limit the initial order.
Set a refresh trigger. A five-year-old project may remain technically relevant while the team, plant, system, component supply chain or certification has changed. Conversely a recent project may not be comparable. Recheck material facts before award and when the proposed delivery route changes.
| Dimension | Disposition | Evidence relied on | Limit/condition |
|---|---|---|---|
| Project existence | supported / limited / unresolved / contradicted | [primary record/contact] | Public presence is not supplier attribution |
| Supplier entity and role | [record] | [contract/client/trace] | Affiliate/team transfer stated separately |
| Configuration relevance | high / medium / low / unknown | Relevance matrix | Unreviewed extremes excluded |
| Performance context | [record] | Factual source and corrective-action evidence | No universal client-satisfaction claim |
| Currency/resources | [record] | Current proposed team/plant/provider link | Revalidate after change |
| Procurement use | informative / conditional / insufficient | Integrated reviewer note | Not supplier approval by itself |
Safe conclusion: The evidence supports the stated entity's role in the reference project to the recorded scope and supports relevance only for the compared dimensions. It does not prove identical product performance, current capacity or future project success.
FAQ
Is a project photo enough to validate a reference?
No. It may support visual context. It normally does not establish the legal supplier, plant, contract role, hidden configuration, performance or client outcome.
Should the buyer call every reference?
Use a proportionate risk-based sample. For material references, obtain client authorisation and verify the contact route independently. Record the contact's actual knowledge and permission.
What makes a window project similar?
Predefine relevant dimensions such as entity/team/plant, product and opening type, configuration extremes, performance route, scope, destination, volume, programme and recency. Prestige or appearance alone is weak.
Can an affiliate's project count?
It can inform the decision when the relationship, actual performing entity, transferable resources and their commitment to the new order are evidenced. Do not present affiliate history as the bidder's own without qualification.
Is a project with problems a bad reference?
Not automatically. Review context, severity, containment, corrective action, verified closure and recurrence. Controlled recovery can be more informative than an unexplained perfect narrative.
What if confidentiality prevents disclosure?
Use client-authorised limited confirmation, lawful redacted records or an independent reviewer with a clearly stated verification scope. The result remains limited to fields actually checked.
How should a new supplier without references be treated?
Record the gap and use proportionate alternative controls such as team evidence, trial drawing, prototype, first-off inspection, pilot batch, staged release and tighter monitoring. Do not relabel these as past-project experience.
Does one validated reference prove future capacity?
No. Past performance is one indicator. Current capacity needs a separate resource-loaded review of the proposed team, plant, providers, bottlenecks and commitments.
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.
Official sources used
U.S. FAR 15.3 — Source selection and past-performance evaluation ↗U.S. FAR 42.15 — Contractor performance information ↗U.S. FAR 12.206 — Use of past performance ↗World Bank — Current Project Procurement Framework and documents ↗World Bank — Procurement Regulations for IPF Borrowers, seventh edition ↗World Bank — Archived similar-contract qualification example ↗UNCITRAL — Model Law on Public Procurement guide ↗UK Government — Supplier selection: experience and past performance ↗ISO 9001 Auditing Practices Group — Evidence collection ↗ISO 19011:2026 — Guidelines for auditing management systems ↗ISO 10013:2021 — Guidance for documented information ↗Related buyer guides
This guide does not validate or imply any Cavorlen project, client, factory, delivery record or capability. It does not treat a reference, award, photograph, invoice, test report or client statement as universal product evidence. All names, quantities, dates, roles, configurations and results must come from authorised project-specific records; unknowns remain unknown.