Which governs when the plan sheets and the 3-D model disagree

The answer, up front

It depends on the agency, and on some projects it depends on the project. The trade's folk answer — the plans always govern — is not reliably true across the documents we read.

We read the published documents of six bodies, every one of them fetched from the publishing body's own server, and the five that reach the question answer it differently. We are not going to sort them into two camps, because the camps are what produce the overstatement. Here is each one in its own words:

• PennDOT has defined a delivery mode in which "a model(s) comprises the primary construction contract document, preeminent in importance as defined by the Specifications or Special Provisions" — and it assigns that status file by file, so the LandXML export of a governing DGN is itself marked "For information only."
• WisDOT says its construction data packet "does not have construction contract document status for construction projects."
• SUDAS furnishes electronic support files "for information only," and where one conflicts with a contract document, "the contract documents shall govern."
• TxDOT posts cross sections and 3D models "For Information only" behind a mandated disclaimer — while on its Digital Delivery pilot projects a special provision brings digital files inside the definition of "plans."
• Caltrans ranks six contract parts for resolving a discrepancy, and no entry in that list names a model.
• FHWA's TechBrief on 3D data is not contractual authority: its own last-page Notice says it "does not establish policies or regulations."

Other states are moving on this question and we have not read them from primary sources, so treat this as six documents rather than a survey.

The question is concrete: you have a plan set and a 3-D design model — a corridor, an alignment, one or two triangulated surfaces, usually LandXML, sometimes native DGN — and at some station they disagree by a few tenths. Which one are you building to, and which one are you paid against? Five of these bodies have addressed that question in their own documents — some in binding specification text, some in program guidance that is not contract language at all — and FHWA's brief does not reach it. The disagreement is genuine, not a failing of any one of them, and it means the answer for your job lives in your bid documents.

This is a documentary reading, not guidance. We quote what published documents say and cite where they say it. Your project is governed by your contract documents, your special provisions, and the edition of the standard specifications in force at your letting. Read those, and ask the owner in writing before you price the work. A contractor acting on somebody else's reading of a specification carries an exposure we cannot carry for them.

What we read, and where it came from

All six were fetched from the publishing body's own server: FHWA (fhwa.dot.gov), Caltrans (dot.ca.gov), WisDOT (c3dkb.dot.wi.gov), TxDOT (txdot.gov and dot.state.tx.us), PennDOT (pa.gov), and SUDAS (iowasudas.org). The language we quote is contract text in only three of the six: SUDAS's Standard Specifications, Caltrans's Standard Specifications, and TxDOT's special provision. PennDOT's Interim Guidelines and TxDOT's PS&E manual section are program and procedural documents, WisDOT's operative sentence sits on a knowledge base page rather than in its specifications, and FHWA's TechBrief says of itself that it establishes no policy — and the rows below say so. This is not a fifty-state survey.

PennDOT: a mode in which the model is the primary contract document

The Digital Delivery Interim Guidelines (Version 2.4, February 28, 2023, Chapter 1) define a form of digital delivery in which "a model(s) comprises the primary construction contract document, preeminent in importance as defined by the Specifications or Special Provisions." The definition goes on to elevate the project design models, 2D and 3D, "to primary authority for construction." On the name: the Guidelines use "Model as the Legal Document," while PennDOT's own program webpage reads "Model as Legal Deliverable." The acronym is MALD either way.

Read the qualifier, because it is the whole hinge. Primacy is conferred "as defined by the Specifications or Special Provisions." The Guidelines do not themselves make the model govern; they point at the contract documents that do. The word "precedence" appears nowhere in their 180 pages, and we did not retrieve the project special provisions that would carry an operative clause.

What PennDOT does instead of a ranking is a per-file binary. Table 14 labels each delivered file "Legal document" or "For information only," fixed at PS&E in a signed and sealed Certification Memo and documented in the Project Index File. The native DGN design models, corridors, bridge and drainage models, and the project PDF roll plot are legal documents; the LandXML exports of those same surfaces, alignments and drainage are "For information only," as are the Right of Way, Utility and Traffic Control DGNs.

So even here, "the model governs" is too coarse. A contractor loading the LandXML export is loading a file PennDOT designated non-governing, while the DGN it was exported from is designated legal. That is the finding: not a new rank in an old list, but a different mechanism — status assigned file by file, in a signed memo, at PS&E. It is a pilot with a stated trajectory. An August 7, 2025 Implementation Policy memo commits each District to a minimum number of Digital Delivery Projects let each year beginning in 2027.

Where the files are furnished for information

WisDOT. The Civil 3D Knowledge Base page on the construction data packet (last updated 2026-06-02): "While the CDP is a contractual design project deliverable, it does not have construction contract document status for construction projects. If there is a discrepancy between the information in the plan and the information in the CDP, the contractor is required to build the project according to the plan." Note the two-sidedness: the packet is owed by the designer to WisDOT, and is not a contract document as between WisDOT and the contractor. Note also that the sentence governs "the information in the CDP," a package-level term; it does not say "model." The surface models are what the packet carries, so the inference is short — but it is an inference. And this language lives on a knowledge base page: it is not in FDM 19-10-43, the Facilities Development Manual procedure the page itself names as its requirement source, and the CDP does not appear in the specifications' precedence list. WisDOT Standard Specifications 105.4(3) (2025 Edition) ranks addenda, special provisions, plans, additional special provisions and standard specifications, and the CDP is not in that list.

SUDAS. The specifications used by Iowa's local agencies, Section 1040, Article 1.05 (2026 Edition; page stamped "Revised: 2018 Edition"): "Electronic support files, if available, will be provided prior to letting and are for information only. Should there be a discrepancy between an electronic support file and a contract document, the contract documents shall govern." The order of precedence at 1.03.A runs eight items, change orders down to the SUDAS standard specifications, and electronic files are not among them.

That binding text never says "3-D model." The only SUDAS document naming a 3D file is the Design Manual, Section 1D-1 ("Revised: 2021 Edition"), written in "should" throughout, which advises the designer to attach "a disclaimer statement... that indicates the paper copy on file with the agency is the official copy." The document that names models is not the one that governs, and the one that governs does not name models. Note too that plans sit fifth of those eight: they govern over an electronic support file, and are themselves subordinate to change orders, addenda, the contract and special provisions.

TxDOT. The PS&E Preparation Manual, Chapter 7, Section 7.3 (Manual Notice 2024-1, effective October 2024) is titled "Provide Cross Sections and/or 3D Models to Contractors – For Information only," and mandates a disclaimer on transmittal: "The following data is for non-construction purposes, only. It is the responsibility of prospective bidder to validate the data." TxDOT does not say the model "is not a contract document" — that is a reasonable inference from the disclaimer, not the manual's words. And the disclaimer is introduced as applying to "the transmittal of the cross section data" while the heading says "cross sections and/or 3D models," so the operative language reads as drafted for cross sections with models added around it.

There is a live counter-current at TxDOT. On Digital Delivery pilot projects, the Special Provision to Item 000, "Important Notice to Contractors" (2024 Specifications, 000-079, dated 12-24), states that the 2024 Standard Specifications definition of "plans" includes "digital files" as part of the documents that make up the Contract. And the Model Development Standards Guidance — cover-stamped "DRAFT- August 2025," still being piloted — runs from interim for-information-only deliverables to "Models as the legal document (MALD) on select projects."

Caltrans: a rule for discrepancies, not a statement of what is contractual

Standard Specifications, 2025 Edition, Section 5-1.02 ranks the contract parts in descending order: special provisions, project plans, standard specifications, revised standard plans, standard plans, supplemental project information. No entry names a model or an electronic design file.

Two things follow, and only two. First, that list operates "If a discrepancy exists" between contract parts. It is a conflict-resolution rule. It does not say that a model is part of the contract, and it does not say that a model is outside it — read either way, it is being made to answer a question it was not drafted for.

Second, Section 2-1.06B describes electronic design files as containing "cross sections, digital models, and roadway design alignments and profiles," but the only thing that subsection affirmatively declares to be supplemental project information is logs of test borings. The definition at 1-1.07B is conditional on its face: information is supplemental project information when it is "specified as supplemental project information." That specifying act happens in the project's special provisions, not in the Standard Specifications. So whether a given project's design model has rank 1.6, or no rank in this list at all, is answered by that contract's special provisions.

"Caltrans ranks the model sixth of six" asserts a classification the provision never makes, unconditionally, where the governing definition is conditional. That sentence is what got the earlier version of this article pulled, and it is why the section you are reading claims less than you might expect.

Where FHWA fits, and where it does not

FHWA's TechBrief "Utilizing 3D Digital Data in Highway Construction" (FHWA-HIF-17-031, April 2017), authored under contract by WSP | Parsons Brinckerhoff, is not a contractual authority: its own Notice says it "does not establish policies or regulations," and the words "precedence," "govern" and "binding" appear nowhere in it. It reports an engineering problem — the data "is often not sufficient for construction due to a variety of reasons," of which "the most notable" is that "the original ground basis for the design differs to field conditions." That is a statement about model quality, which is a different proposition from contractual status. Note also the direction it leans: it recommends "design practices that prioritize the 3D model as the source of the contract plans."

What these documents do not say

The negatives are as load-bearing as the quotes, and they are easy to lose in summary:

• PennDOT's Guidelines do not make the model govern. They confer primacy "as defined by the Specifications or Special Provisions," and "precedence" appears nowhere in 180 pages.
• WisDOT's operative sentence does not say "model." It says "the information in the CDP," and it lives on a knowledge base page, on a knowledge base page rather than in FDM 19-10-43, the procedure that page names as its source, or in the precedence list at 105.4(3).
• SUDAS's binding specification does not name a 3-D model; it says "electronic support files." The document that names 3D files is the non-binding Design Manual, written in "should."
• TxDOT does not say the model "is not a contract document," and the disclaimer it mandates names cross section data.
• Caltrans does not say whether a model is or is not part of the contract. 5-1.02 says what happens if a discrepancy exists.
• FHWA does not address which document governs at all, in any form.

The one measurement of our own

Mathyra is pre-launch: no customers, no pilots, nothing shipped, and we have never shipped a surface. We have one number relevant here, and state it with its limits.

On one corridor — FHWA Federal Lands, OR FLAP DOT CRGNSA 100(9), public domain via SAM.gov — our reading of the solid pen family off the cross-section sheets sat at a median RMS of 1.503 ft against the finished TIN in the same bid package, across 80 scored sections, and within 0.50 ft on 19 of those 80.

One corridor, one designer, one plotting convention. The larger limit: cross-section sheets are plotted from the same corridor model the bid-package LandXML is exported from, so a comparison like this measures how faithfully a drawing was plotted and read back — not how close either the drawing or the model is to the ground. That is why scoring against the designer's model never produced an accuracy number.

It matters here for one reason. If a plotted line and an exported model can differ by that much on one corridor while both honestly represent the same design, then which of the two your contract points at is not a formality. And when our own software found sheets that appeared to contradict each other, the contradictions turned out to be our own chaining rather than the designer's — chase an apparent disagreement to its cause before it is priced as a conflict.

What to ask before you bid

• Which article of the standard specifications covers digital or electronic files on this letting, and what edition is in force? Within a single agency's specifications, the article that ranks contract documents and the article that allocates risk on electronic data can be two different provisions doing two different jobs.
• Does that list name digital files at all, and if it names a category, does anything define what is in it? An undefined category and a named file are different situations.
• Is there a special provision that changes the ranking? On a Caltrans job, special provisions rank first and can rewrite 5-1.02 entirely; on a PennDOT MALD job, the Guidelines point at the Specifications or Special Provisions for the operative clause, so that is where to look.
• Where files are designated individually, which are legal documents and which are for information only? A native model file and its LandXML export can land on opposite sides of that line.
• If the model and the sheets disagree at a station, which document will the owner answer an RFI against? Ask before the first bust, not after.

None of that is exotic — it is data prep and pre-bid diligence, and in most organizations it has no owner. That is what we found looking for the person whose job this would be.

The honest summary

Anyone who tells you the plans always govern is describing some of these documents and not others. Anyone who tells you the model is the new authority is describing one pilot program and reading past its qualifier — PennDOT's own Table 14 makes the native model a legal document and the LandXML exported from it explicitly not. Nobody is describing all six, and that is the finding.

Read your own contract. Ask the owner in writing. For the mechanics of how a plan set becomes data before any of this comes up, that is the problem Mathyra is being built to solve.