xyzbim.eu / pesquisa / boeing-777-da-construcao · essay · 2026-09-06

The Boeing 777 of construction — why it does not yet exist in 2026

Aerospace, automotive, semiconductors, logistics and animated film integrated their chains end to end. This essay gathers what the literature says about how they managed it, why construction fell behind — and a simple test: we searched the scientific literature and the public repositories for a building with the documented process and an accessible as-built model. It does not exist (verification of 2026-09-06).

In one sentence: the sectors that got there had three things construction does not have — an open and stable data standard, someone with contractual power to impose it on the whole chain and a product that repeats — and the result shows in the simplest number of all: while the Boeing 777 has been designed entirely in digital since 1994, there is not a single building in the world whose complete process is published and whose as-built model can be downloaded today.

1 · The aircraft that construction still has not had by 2026

Between 1990 and 1994, Boeing designed the 777 entirely in CAD — the first commercial aircraft without a final physical mock-up, with digital pre-assembly and the Pratt & Whitney, Rolls-Royce and General Electric engines coordinated in model. What was learned there was not about software.

Software was the least of it. What made the difference was the chain: Boeing agreed with the three engine makers to use the STEP standard “as the production data exchange process” [C, NIST Planning Report 02-5]; the US Navy began writing into contracts the delivery of data in the ISO 10303-239 profile, and the Department of Defense requires 3D models from its suppliers [C, NIST TN 1753]. The standard was born of a measured pain: the lack of interoperability cost the North American automotive industry on the order of a billion dollars a year (RTI 1999, via NIST 02-5). And the doctrine that came out of it is explicit:

“a digital three-dimensional (3D) model … serves as the authoritative information source for all activities in the product's lifecycle.” Model-Based Enterprise, NIST Technical Note 1753 · the model as the single source of truth, not the exchange of documents

Two more pieces construction envies. First: STEP has been in use since 1994, and the AP242 (2014) merged the aerospace and automotive variants into a single profile — a mature standard before most BIM programmes were born. Second: the product that repeats — Ford's platforms cut 15–20% of development cost per model [C, NIST 02-5]; each aircraft, each car amortises the digital investment over thousands of units. A building does not repeat itself, and it is here that the analogy starts to break — we return to that in §5.

2 · How the others managed it

Sector	How it integrated the chain	The mechanism
Aerospace	Boeing 777, 100% digital; the OEM imposes STEP and tools on suppliers	standard + integrator with contractual power
Automotive	product platforms (Ford −15–20%); interoperability costing $1bn a year led to STEP	repeated product + open standard
Semiconductors	IEEE 1800/1076 as the contract between stages; common roadmap ITRS→IRDS; a consortium pays for the standard	standard + roadmap + someone to fund it
Maritime logistics	ISO 668 container — the interface was standardised, not the product; the gain came from handling, not freight	interface + data (EDI)
Process industry	ISO 15926 for the design→operation handover — but with no consensus on criteria and no conformance testing: partial adoption	negative lesson
Cinema/VFX	USD (Pixar, open-sourced in 2016) — AOUSD alliance under the Linux Foundation; Core Specification 1.0 ratified in Dec 2025, with a path to ISO	consortium keeps the format open

Two notes of honesty, because portraits of these sectors almost always lack the shadow. First: even in the leading sectors, benefits are not always measured — in a systematic review of MBSE, two thirds of the claimed benefits had only “perceived” evidence, not measured [C, DOI 10.1002/sys.21566]; in North American industrial practice, 85% of firms use 3D models but only 41% as the primary source of information [C, NIST survey]. Second: the process industry is the experimental control — ISO 15926 was created by asset owners for the handover of engineering data to operations, and even so NIST concluded there is “no consensus as to the criteria for achieving interoperability” and “modest” conformance verification effort: half a standard does not stick [C, NIST IR 7917, read in full]. The parallel with IFC and buildingSMART is uncomfortable and direct.

3 · The test: we searched for the most documented building in the world

If the other sectors have the complete digital thread, buildings like this should exist. We set the bar on three legs: documented process end to end, as-built model (reflecting what was built, not what was designed), public access today.

We put the question to the scientific literature (OpenAlex, ~1,700 works in four corpora [measured]) and to the public repositories (Zenodo, GitHub, state programme archives). The result: zero (verification of 2026-09-06). No real building meets the three legs — and the pattern of the failures is more informative than the count:

• Whoever documented everything closed the model. The Crossrail programme (Elizabeth line, London) published the best process documentation that exists — half a million assets registered in the information system — and zero model files [C, Crossrail Learning Legacy]. The Sydney Opera House twin is official and has no route of access [C, Sydney Opera House]. The data from The Edge's 28,000 sensors, in Amsterdam, stay with the operator [C, Bloomberg 2015].
• Whoever publishes does not document the process. The open models are research datasets — the capture of an instant, not the narrative of the cycle.
• What circulates freely is synthetic. The buildingSMART example files exist for purposes “educational, implementation and certification” [C, buildingSMART] — precisely because they belong to no one.

The two closest cases are complementary half-proofs. The Schependomlaan, a real Dutch house, has the entire collaborative process documented — weekly IFCs, clash reports in BCF, event logs — and the whole set is downloadable today (CC BY 4.0; we downloaded and opened the IFC, 49 MB) [C, buildingSMART mirror]; but the as-built is only a point cloud — there is no as-built IFC. The NZERTF, the NIST laboratory house, has the opposite: design, construction, two years of operation with 379 instrumentation channels and the data open with DOI for a decade [C, NIST] — without any BIM. And the epitaph of the genre: the IFCs of Norwegian public buildings were once listed on an FTP of the HIBO programme, which died without a copy in the web archive [C, verification made on 2026-09-06].

4 · “But what is the problem with sharing an FM model?”

It is the right question — and the answer starts by undoing a confusion: equipment brands and warranties are the innocent part. They already travel in the O&M manuals the owner receives and in COBie; of all the information in the model, they are what benefits the building's owner most. What blocks sharing is something else, in order of weight:

• The model is evidence with legal value. An as-built records who specified what, when, and where reality diverged from the design — it is a potential means of proof in litigation. In Portugal, the decennial liability (art. 1225 of the Civil Code) keeps that exposure alive for ten years [inferred — the house's legal reading, not one from the reviewed documents]. Nobody voluntarily publishes their future prosecution evidence, in an industry that McKinsey describes as contracting “based on … the lowest bid price” [C, MGI 2017].
• Physical security — there is a standard just for this. An FM model contains positions of video surveillance, access control, door schedules, voids and network routings. The ISO 19650-5:2020, “Security-minded information management”, specifies the classification and the controls for sensitive information of built assets [C, ISO 19650-5]. It is also for this reason that the British programme's prison models are not public.
• Privacy. A twin in operation carries occupancy data — presence patterns of identifiable people. The Edge is the limiting case [C, Bloomberg 2015].
• Commercial value and lock-in. Maintenance data are the cost of switching for the FM contractor — retaining the model is retaining the client [inferred]. And the model is a contested asset: copyright of the designer over the design embedded in it, the contractor's records, deliverables the client paid for — the fashionable critique of “platforms” points exactly at the missing governance and ownership [C, Aksenova & Oti-Sarpong 2023].
• Publishing creates a duty to maintain. A public as-built dies at the first renovation, and an outdated model someone trusts is worse than none. After handover, there is no budget to maintain — the Norwegian FTP above is the literal demonstration.

The sum of this is not secrecy — it is a collective action problem: the benefit of sharing is diffuse (the industry learns), the cost and the risk are private. That is why the only almost-open cases came from academia, where it is research funding that pays the bill for sharing.

5 · What can be mirrored — and what cannot

What cannot be mirrored was said by the founders of the transfer itself. Lauri Koskela wrote in 1992 that the peculiarities of construction — “one-of-a-kind projects, site production, temporary organization” — prevent flows as efficient as manufacturing's [C, CIFE TR 72]. Six years later, Ballard and Howell closed the argument:

“construction is a fundamentally different kind of production… Adopting a single-minded strategy of transforming construction into manufacturing would be precisely the wrong thing to do… Manufacturing is becoming more like construction.” Ballard & Howell (1998), IGLC-6 · the convergence runs both ways

And copying without adaptation is documented as failure: DfMA guides imported from manufacturing “without adequately considering important differences between the two sectors” — six axes of difference, from site to process [C, Tan et al. 2020, read in full]. The right mirror is not “being a factory”: it is importing the mechanisms — imposed standard, platform that amortises, closed data loop — and letting them live in the form construction has.

Still missing is the number that frames everything: in the US, construction is the second least digitalised sector, ahead of only agriculture — and in Europe it is in last place, behind it: “In the United States, construction comes in second to last, ahead of only agriculture. In Europe, construction is in last position.” [C, MGI 2017] — with productivity falling about 1% a year between 1970 and 2020 — in 2020, below the 1950 level [C, Goolsbee & Syverson, NBER 30845]; in Europe, 94% of firms have fewer than 10 workers and “no firm is large enough to pioneer and lead major innovations” [C, MGI 2017]. The first building with the three legs — published process, verifiable as-built, open licence — is one contractual clause away. The 777 of construction is going to be a small building.

6 · Frequently asked questions

7 · The next step