The tender documents of the perfect competition — meta-structure

For those who write tender documents (project owners, technical advisory teams, municipalities, SRU). Distilled from 63 normative documents analysed in depth + 105 jury reports + every clarifications cycle in the OA archive (705 competitions, 2010–2026). Every structural claim stands on measured frequencies.

Which procedure is this? (read before anything else)

The first decision changes the criterion, the documents and the strategy — and the archive shows they are often confused:

This document is written for the first column — but PART IV (EIR) applies to all three; it is in services and design-build that the OA-SRLVT documents the worst quality pathologies [measured].

1. Principles (what separates the perfect from the average)

1. One document, one numbering. The archive is full of material split between CE and TR with no equivalence table (CML, the Lisbon municipality: security in the CE, criteria in the TR) and technical clauses restarting at "1." — two clauses numbered 1 in the same document [measured]. The perfect tender document is one single articulated body with continuous numbering and an index.
2. Dates and placeholders verified. An official TR has been published reading "Deliberação da CML de **/07/2025" — asterisks included [measured]. Nothing goes out with placeholders.
3. Technologically neutral. Naming proprietary software/brands/formats is forbidden; bases delivered in IFC/DXF, never .rvt alone [measured].
4. Price is not a design criterion — but it decides almost everything else. In the archive's 105 design jury reports, "fees" has 0 occurrences [measured]. Scope: this characterises the design competition, not the market. Public procurement of architecture as a whole is dominated by price: the OA-SRLVT Observatory 2023–2025 measures 89.4% of 386 procedures with price in the criterion — and the 10.6% without price are, with two to spare, exactly the 38 design competitions [measured]. And the Observatory 2020–2022 splits by advisory support: with OA advisory support, price was a criterion 0 times in three years (0 of 34); without, 89% (93 of 104) — across all 138, 67% [measured]. The difference between the two worlds is the document — which is what this meta-document is. This diagnosis now has a proposed answer in Brussels: art. 98 of COM(2026) 590 makes the best price-quality ratio mandatory, with a minimum of 30% quality — 50% in labour-intensive contracts — and leaves to pure price an exceptional derogation, when quality is ensured by specification or execution conditions [reading of the proposal]. If adopted, these 89.4% will come to measure the before.
5. Deliverable by a 3-person atelier. Every requirement (incl. BIM) must be satisfiable with free software; qualification evidence only at the suitability stage, never in the proposal [principles P1–P5].
6. Every formal rule must be worth the exclusion it causes. In 2025 one competition had ALL proposals excluded and was revoked [measured]. Duplicated formal rules (physical AND digital panel, each "under penalty of exclusion") are traps, not rigour [measured].

2. The structure (skeleton consolidated by frequency)

The perfect document has 4 parts in a single body — they correspond to the 4 genres today scattered across separate files [measured]:

PART I — Procedure (skeleton of ~25 articles, present in ≥3 of the 6 TRs analysed with stable order [measured])

Systemic hole: in 12/19 documents the deadline lives outside the main document (announcement/platform) — date corrections are the most common reason for clarification [measured]. The perfect one carries the complete calendar in a table of its own.

Good practices the average TR loses: (a) proposal validity with a number — 120 days is the current reference in service procedures; in the OA documents it is the 1/19 hole; (b) model annexes as part of the skeleton — a library of models (proposal declaration, terms of responsibility per role, standard team table, three forms of security) that removes submission ambiguity; (c) articles almost nobody carries that are worth adding: variant proposals, signature of the proposal, publication of the list of entrants.

PART II — Contract (clauses, when the design is awarded)

Chapters: I General provisions (object, base price) · II Obligations (phases, deadlines per phase — preliminary study (EP) 30–60 d, preliminary design 60 d [measured] — clarifications, responsibility for errors, acceptance, IP, confidentiality, payments) · III Penalties and termination · IV Security (8/19; release 30 days, art. 295 CCP) and insurance · V Final provisions (subcontracting, communications, counting of deadlines, disputes, GDPR) [measured]. In the mature version: the contracting authority's obligations in their own chapter and retention in payments as a standalone clause.

PART III — Preliminary programme (technical annex)

1. Preamble · 2. Context · 3. Location and boundaries · 4. Characterisation · 5. Objectives · 6. Functional programme with an areas table per cell · 7. Construction cost estimate (the base value + construction cost pair is in 15/22 OA records) · 8. Construction sustainability · 9. Explicit urbanistic constraints ("max. 3 floors above ground…") · 10. Annexes [measured]. Good practice: number the specialties with stable codes across the whole tender document and the price map — it gives traceability between programme, prices and invoicing and kills the pathology of "duplicated clause 1".

PART IV — Exchange Information Requirements (EIR)

The section that does not exist in any of the 1,059 documents of the archive with normative vocabulary (0 mentions of ISO 19650/COBie/LOIN/CDE) — but that already bursts through in 2025 (SRU Lisboa Ocidental: ANNEX II EIR, BIM manager 90 h, exclusion for an incomplete BIM Competencies Declaration) [measured]. The ready-to-adopt text rests on: principles P1–P5, exchanges per phase, LOIN without generic LODs, naming, PT-TM06/EPSG:3763, open formats (IFC4/BCF/IDS/CSV), 10 automatable acceptance criteria, calendar, enforceability checklist.

EIR survival rules: (1) it has a price line — its own chapter in the price map, with an express decision on whether it enters the minimum percentages per component and the tie-break (an EIR without price is unpaid work; without scoring it is unevaluable); (2) it has a named owner in the team (a line in the standard table) — otherwise there is no one to address the rejection of a non-conforming deliverable to; (3) specification by performance, never by means — what comes out of the requirement: coverage, accuracy with numbers, formats; never the prescription of instruments; (4) "a requirement without a number is not enforceable" — tolerances, quantities and thresholds fixed in the tender document, not delegated to the BEP; (5) required qualification proportional to the real market — if nobody in the country can demonstrate the experience, the requirement moves to a post-award test, not to a CV.

3. Award criteria (the design the juries actually use; scale descriptors translated)

Measured frequencies in 46 reports with legible criteria [measured]:

Recommended scheme (the most tested in the archive, ≥15 IHRU/Lisboa Ocidental competitions 2021–2024): 30 / 25 / 15 / 15 / 15, with subfactors along the lines of the 2025 model (35/25/25/15 with C1.1–C1.3) [measured]. Rules: weights sum to 100% (verified classic failure: sums ≠100 [measured]); anchored 1–10 scale with descriptors per level published in the tender document (10 = "notable clarity and consistency"; 6 = "many aspects without development"; 4 = "formal weaknesses") [measured]; tie-break = best score on the heaviest factor.

4. Prizes (real ranges)

1st prize €2,000–30,000: small Azorean competitions €2–3k; IHRU/Santa Engrácia €7–8k; Évora/Castelo Branco/MNAC/Academia Ciências €9–12.5k; Martim Moniz €30k. Structure: first prize + participation prizes + non-monetary mentions; the winner also receives the contract by direct award (art. 27(g) CCP) [measured]. Practical rule: prizes ≥ the investment asked of the entrants (the OA's promoter manual asks for exactly this balance [measured]).

5. Anti-pathology checklist (before publishing)

• [ ] One continuous numbering; zero placeholders; dates in a table of their own
• [ ] No duplicated formal rule (physical+digital with double exclusion)
• [ ] Required documents = documents listed exactly once, with format and size
• [ ] Anonymity with explicit separate-packaging instructions (it is the real exclusion no. 1 [measured])
• [ ] Criteria: weights sum to 100%, scale with descriptors, no price
• [ ] Clarifications cycle with deadline and form (it generates a dated document)
• [ ] EIR under principles P1–P5 ( evidence only at the suitability stage)
• [ ] Proposal validity defined with a number (current reference: 120 days)
• [ ] Security/penalties proportional (or absent in an ideas competition — 0/22 OA records have them [measured])
• [ ] Model annexes included (declarations, terms of responsibility, team table, forms of security)
• [ ] The EIR has: a price line, a named owner in the team, numbers on every tolerance

6. Errors the archive documents (do not repeat)

1. TR published with an asterisked date [measured] · 2. OA "tender documents" that are post-competition catalogues without the brief [measured] · 3. Prize regulations mixed up with commissioning competitions [measured] · 4. Templates cloned between entities, propagating errors (Loures/SGMAI/Abrantes; 4 CML schools with the same CE) [measured] · 5. Substantive decisions taken in clarifications ("the wall is outside the intervention area") instead of in the programme [measured] · 6. BIM requirements without reciprocity (demanding IFC, delivering .rvt; or demanding BIM and not delivering the BIM survey) [measured].

7. External validation

The structure was also confronted with acquisition of design services procedures (CCP regime, award to the most economically advantageous proposal): the four parts hold — the procedural document maps onto Part I, the legal chapters onto Part II, the technical programme (embedded or standalone) onto Part III, and information management is again absent from the published ones and present in the improvement proposals — the same asymmetry as the OA corpus. The differences are of regime (price vs prize; suitability by standard team tables) and of numbering discipline, not of structure.

7.2 International standards (7 systems, official sources)

Confrontation with UIA/UNESCO 2020, Germany (RPW 2013), France (loi MOP/CCP via MIQCP), the United Kingdom (RIBA), Switzerland (SIA 142/143:2025), the Nordics (DK/NO/SE) and Spain (Ley 9/2017 arts. 183–193) [31 official sources]:

What is confirmed: Part I is equivalent to the "Regulations" that all 7 systems have (jury, anonymity, clarifications, criteria, prizes, calendar) and Part III to the universal "Programme" — the UIA even asks for the budget in the programme, just like this structure [measured]. Part II (a contract inside the tender document) is uncommon: FR/DE/ES award to the winner through their own route and the UIA only demands a commitment to contract within 24 months with minimum compensation if the works do not advance — add as an article of its own [reading].

What the international systems have and the Portuguese tender documents do not: (a) a review phase for the prizewinners (the German Überarbeitungsphase — the placed entrants rework with paid support); (b) public exhibition and published report as a normalised closing step of the procedure; (c) normalised model declarations (UK/RIBA); (d) an odd jury with a majority of independent experts (UIA ≤9; RPW; SIA ≥½ independent; DK/FR quotas of ⅓, ES ⅔) — the even 5+5 jury practised in Portugal is singular in this landscape [measured]; (e) prizes in international rules: UIA 3–10 prizes + mentions for ≤30% of the entrants; Germany: Wettbewerbssumme ≥ the fee of the preliminary planning phase, to be used up by the prizewinners [measured].

Part IV is the common gap: none of the 7 systems has a standalone information requirements block in the brief — Spain has only a generic suitability requirement (DA 16 of Ley 9/2017) and Norway keeps the brief purely procedural [measured]. Whoever legislates on the BIM side does so outside the competition rules: Denmark BEK 119/2013 §6 (digital models in design competitions), Spain Orden PCM/818/2023 (BIM requirements in pliegos by typology, including "Concursos de proyectos" [design competitions]; OC 1/2025 chain: Objetivos→Usos→Modelo→Entregables→ECD→BEP), and the European basis was, until now, art. 22(4) of Directive 2014/24/EU — a permission, not an obligation. The proposal replacing it changes the verb (see 7.3).

International reinforcements for Part IV: (1) the clause "automatically verifiable = automatically verified" (Statsbygg/SIMBA principle); (2) a requirements matrix discipline × phase (SIMBA 7×4; COBim 13 series); (3) express legal basis in the articles (art. 22(4) of Directive 2014/24 while it stands; art. 65 of the COM(2026) 590 proposal when applicable; DL 10/2024 in building permitting; RCM 89/2026 in the national strategy); (4) an Information Protocol as contractual annex (the UK successor to the CIC BIM Protocol); (5) an EIR→operation bridge (maintenance AIR/COBie-lite — Spain/Denmark).

7.3 The legal framework coming (2023–2030, with a declared limit)

Which piece mandates what — and when? It is worth separating three instruments that the public conversation mixes up. Portaria 255/2023 (a ministerial order), already in force, introduces parametric BIM models into the instructions for public works projects, under the "whenever possible" of art. 62 of the Public Contracts Code (CCP) — it does not mandate, it has no threshold, it has no vocabulary. DL 10/2024 (Simplex Urbanístico), amended by DL 108/2026 (29 May), requires architecture projects in building permitting under the RJUE (Portugal's building permitting code) to come BIM-modelled from 1 January 2030 — the regulating ministerial order still unpublished as of 14-09-2026. The amendment removed the 2027 pilot and added the first value threshold: projects of public works contracts above €5,404,000 (art. 474 of the CCP) come to owe a BIM model "whenever possible" — permissive, on the works side, without vocabulary. PortugalBIM (RCM 89/2026, of 21 May) is the national strategy: four pillars, IMPIC as implementer, an Action Plan due in August 2026; it promotes the adoption of BIM in public procurement (target 5/g) without mandating it. None of the three demands ISO 19650, LOIN or CDE of anyone.

The fourth piece is on its way and is the biggest: the Public Procurement Act (COM(2026) 590, proposed on 9 September 2026), a Regulation of direct application repealing the three 2014 Directives. Three points matter to this study: art. 65 makes BIM mandatory in the execution of works contracts ≥ €25M "in open and interoperable formats", with the threshold lowerable by delegated act of the Commission — the verb changes from allowing to mandating, but the vocabulary remains undefined; art. 98 orders award by the best price-quality ratio, with a minimum of 30% quality (50% in labour-intensive contracts) and a derogation only when quality is ensured by specification or execution conditions; and an absence: in 200 pages, "design contest" appears once, in a lateral cross-reference — the design competition, which arts. 74–77 of Directive 2014/24 still anchor in European law and which the CCP regulates in arts. 219-A et seq., disappears from the proposed framework [reading]. My argument is that this does not weaken the Portuguese design competition — it frees it: no longer coming from Brussels, the regime can be rethought from the evidence of this archive — an odd jury with independent experts, a paid review phase for the placed entrants, a published report as a normal step — instead of from transposition.

Where did the number come from? The study backing the proposal — "BIM and Public Procurement", commissioned by DG GROW and published on that same day, 9 September 2026 — had drawn the path with a different anchor: a graduated obligation starting from the European works threshold (€5,404,000), a coordination step "around €1M" below it and reinforced requirements above €15M and €50M, all in open formats and with progression conditioned on evidence reviews — its scenario 3. About 75% of the stakeholders consulted asked for clear regulatory guidance. Art. 65 settled for €25M — between the study's €15M and €50M bands — with the descent referred to a delegated act: the study's dynamic architecture without its starting point [reading]. For those writing tender documents in Portugal, the consequence is the same in both scenarios: the vocabulary still does not come from Brussels.

Limit to declare: everything said above about COM(2026) 590 is a proposal in first reading — with no committee and no rapporteur in the European Parliament as of 11 September 2026 — and weights, thresholds and the fate of the design competitions can change in trilogues; application only takes place two years after entry into force (art. 149), with a realistic horizon in 2029–2030. Until then, whoever writes tender documents still writes them under the CCP and Directive 2014/24 — and whoever wants to set the standard of vocabulary will do it in the EIRs, which is where the power has always been.

Sources of this section: Portaria 255/2023 · DL 10/2024 · DL 108/2026 · RCM 89/2026 (PortugalBIM) · DG GROW study "BIM and Public Procurement" (2026) · COM(2026) 590 · status of procedure 2026/0265 (COD). Checked on 14-09-2026.

Sources

Analysis of 1,059 documents from 705 architecture competitions published in Portugal between 2010 and 2026, of which 105 jury reports; and the confrontation with seven international systems through their official sources — UIA/UNESCO 2020, RPW 2013 (Germany), MIQCP (France), RIBA (United Kingdom), SIA 142/143:2025 (Switzerland), the Nordic rules and Ley 9/2017 (Spain) — and with the applicable legislation: Directive 2014/24/EU art. 22(4), DL 10/2024, DL 108/2026 (amendment to the RJUE), ISO 19650-2, ISO 7817-1, Orden PCM/818/2023, Portaria 255/2023, RCM 89/2026 (PortugalBIM), the COM(2026) 590 proposal (Public Procurement Act, in first reading) and the DG GROW study "BIM and Public Procurement" (2026) that backs it.

Every structural claim stands on a frequency measured in the corpus. The markers [measured] and [reading] separate what has been counted in the archive from what is interpretation. The exclusion cases cited appear in published jury reports.

Primary provenance: the historical competitions archive of the Ordem dos Arquitectos, consulted in August 2026.

The perfect response to an architecture competition — meta-structure

For those who compete (small ateliers, 1–10 people). The other face of the META-DOCUMENT: what the real structure of the PT competitions and the language of the 105 jury reports teach about how to respond.

0. First test: which procedure is this?

Before any strategy: design, acquisition of services, or design-build? The archive shows that 89.4% of 2023–25 procurement had price in the criterion — and that the 10.6% without price are exactly the design competitions [measured]. This document answers the first case. If it is acquisition of services: the technical+price proposal follows another logic (price justification prepared, price proposal in a table). If it is design-build: the OA-SRLVT documents the pathologies [measured] — competing requires a construction partner from the start.

1. The mathematics of the game [reading]

Median: 21 proposals per competition (IHRU/SRU 20–32; emblematic competitions 40–75; small municipal ones 3–18). The jury reads everything in a few weeks. The battle is decided on two factors that add up to 55–60%: concept with exemplary clarity (30–35%) and constructive rationality with a credible construction cost (25%). The other factors (programme 15%, surroundings 15%, sustainability 15%) are not won — they are lost: answer to everything without heroics.

Price: irrelevant in design — and dominant outside it. "Fees" has 0 occurrences in the 105 design jury reports; but in the market at large price decides ~90% [measured]. Within design, prepare a price justification only if the procedure is by invitation/direct award (art. 146 CCP — real case: exclusion for not answering a request to justify an abnormally low price [measured]).

2. What the jury says about the 1st-ranked entry (real language, our translation, to mirror)

• "Clarity in this proposal reflects formal consistency… clarity of the construction solutions… technical rationality materialised in a main structure [that] receives pre-fabricated, modular, repeatable elements of reduced maintenance" (S. Francisco de Borja, 2021)
• "high conceptual coherence, revealing clarity, formal consistency and quality in the materialisation of the programme" (Martim Moniz, 2023)
• "it integrates into the site, assuming a concealed yet deeply transformative presence" (Estrela, 2017)
• "great formal simplicity, intentionally contained… balance between architectural 'protagonism' and urban 'harmony'" (MNAC, 2024)
• a simple structuring idea: "an axis that structures the entire sense of the intervention" (Castelo Branco, 2016); "transfiguration of the area into a park, where the programme materialises in clearings" (Arquivo Municipal, 2025)

Pattern [reading]: (1) a single, legible idea; (2) a complete answer to the programme without gratuitous gestures; (3) sensitive insertion into the pre-existing; (4) constructive credibility (common materials, modularity, maintenance). And the inverse — what kills: "contrived formalism", "the formal overlaying the space", "reference ideas implanted gratuitously", "a less-than-careful reading of the place" [measured].

3. Structure of the proposal (the canonical set of documents)

From the real skeleton of the tender documents [measured], the perfect proposal organises itself into 3 logically separated envelopes (physical separation depends on the tender document — anonymity demands it):

A. Design work (anonymous) 1. Panels (A1, light self-supporting material, opaque envelope): the idea on 1 panel; the development on the rest; scale 1:200/1:500 as required — never identify authorship (not even as "support team": real exclusion for exactly that [measured]) 2. Descriptive and justifying report: concept → programme (an area matrix matching the programme's table) → insertion (photos/analyses of the place) → construction strategy and materiality → sustainability/maintenance (the "easy" 15%) → cost estimate coherent with the areas table (deviation <1% — juries check [measured]) 3. Technical elements as the tender document requires (sections, elevations, summary areas table, estimated cost declaration — missing these last 2 excluded real proposals [measured]) 4. Pre-contract BEP where an EIR exists (section 5 below)

B. Identification form (it identifies, outside the anonymous envelope) — in a strictly separate file/envelope. The archive's most common exclusion is exactly the mixing of the two [measured].

C. Suitability requirements (winner only, afterwards): certificates, OA registration, social security — and the BIM evidence if the tender document defers it to this phase (2025 practice [measured]).

4. The 7 errors that exclude (all with a real case in the archive)

1. Anonymity breach — the identification form accessible with the entries; "Support team" written on the proposal; the BIM Competencies Declaration submitted together with the entries (2025). Exclusion no. 1.
2. Missing documents — no identification form, no areas table, no cost estimate.
3. Upload swapped on the platform — identification form uploaded as the "work"; .xlsx instead of .pdf; panel repeated/missing vs. the digital version.
4. Deadline — 15 minutes late = excluded, no discussion (documented case, 2013).
5. Shared teams — the same team on two proposals; "false declarations" of registration number.
6. Duplication — the copy is excluded.
7. Formal excess — too many pages, non-conforming code (9 of 63 proposals in a single competition, 2016).

Final check before submitting [reading]: every document asked for in the tender document exists, in the format asked, in the right place on the platform, with no author metadata in the files (clean "created by" in the PDFs/IFC).

5. The competition BEP (the answer to the EIR, if one exists)

An 8–12-page structure ready to fill in: identification + commitment, a team with ISO 19650 roles and no fictitious job titles (3 people are enough), 100% free tools (Bonsai/Blender, IfcOpenShell, FreeCAD, QGIS, LibreOffice, Nextcloud as the CDE), modelling strategy per phase (the 2D derives from the model — the sentence that distinguishes an openBIM atelier), mini-MIDP, a point-by-point compliance matrix against the EIR (never "complies" in general: each line cites the ref., the verb and the evidence), risks, pre-flight.

The real 2025 BIM trap [measured]: if they require a qualified BIM manager and you do not have one — a declaration of commitment (hire/assign at the suitability stage) or a consortium with a BIM company (the "atelier + Work in BIM" model, observed); the 90 h evidence can only be demanded at the suitability stage. If they deliver the survey in .rvt, ask for IFC in writing in the clarifications cycle (there is a documented precedent).

6. The clarifications cycle, and what it changes

The archive proves that clarifications change competitions: extensions obtained for lack of a BIM manager; requirements kept/revoked; scopes corrected ("the wall is outside the intervention area") [measured]. Rules of the perfect entrant [reading]: ask early, in writing, about (1) formats/IFC, (2) scope ambiguities with cost impact, (3) deadlines. Every published answer is the law of the procedure — archive everything.

7. Final pre-flight (checklist)

• [ ] All the documents of the "Proposal documents" article present, in the exact format
• [ ] Identification form envelope/file separate from the entries; zero author metadata
• [ ] Areas table = programme (±1%); coherent cost estimate
• [ ] Physical panels identical to the digital ones (real exclusion case over a difference)
• [ ] BEP with a complete compliance matrix; optional IFC validated (ifcopenshell.validate → 0 errors)
• [ ] Submission ≥24h early (a deadline is a deadline, 15 min = exclusion)
• [ ] Keep a receipt/protocol of everything

Sources

Analysis of 1,059 documents from 705 architecture competitions published in Portugal between 2010 and 2026, of which 105 jury reports; and the confrontation with seven international systems through their official sources — UIA/UNESCO 2020, RPW 2013 (Germany), MIQCP (France), RIBA (United Kingdom), SIA 142/143:2025 (Switzerland), the Nordic rules and Ley 9/2017 (Spain) — and with the applicable legislation: Directive 2014/24/EU art. 22(4), DL 10/2024, DL 108/2026 (amendment to the RJUE), ISO 19650-2, ISO 7817-1, Orden PCM/818/2023, Portaria 255/2023, RCM 89/2026 (PortugalBIM), the COM(2026) 590 proposal (Public Procurement Act, in first reading) and the DG GROW study "BIM and Public Procurement" (2026) that backs it.

Every structural claim stands on a frequency measured in the corpus. The markers [measured] and [reading] separate what has been counted in the archive from what is interpretation. The exclusion cases cited appear in published jury reports.

Primary provenance: the historical competitions archive of the Ordem dos Arquitectos, consulted in August 2026.