Revise It: A Moat Analysis

Revit’s moat is that the contract documents it generates are consistently generated for a project. It is stable at revision clouds. Its ability to change-control that it shows the same export again and again despite heavy software engineering on top of it, is its moat, in my opinion. Here I have tried to unpack it using OpenAI Deep Research as my study partner.

Some terms, because the whole argument hangs on them. A moat is the advantage a competitor can’t easily copy. Contract documents are the drawings (and specs) that become part of the construction contract: the set the contractor bids and builds from. A revision cloud is the scalloped outline drawn around whatever changed on a sheet, tagged with the revision number. And a set is never issued once. Permit comments come back, the owner changes the unit mix, the structural engineer moves a column line, and the set goes out again as Revision 1, Revision 2, Revision 7. Every reissue has to be the same set as last time except for what actually changed, and every change has to be clouded. A door number that silently changes on a sheet nobody clouded gets built wrong.

So the question for this page: which parts of Revit built that moat, and how early?

To cross the moat, one must understand it better than those currently defending the moat. Let’s begin.

One distinction to carry through the release history. Some features make a better model: massing, rendering, content libraries. Others keep the set consistent when the model changes: one change propagating everywhere, every view drawn from the same model, 2D detailing attached to that model, many people editing it without forking it. The second kind is the moat. After each release below there’s a note on which kind it shipped.

A note on the record: the dates, prices and feature lists below come from that Deep Research session, not from a line-by-line reading of Revit Technology’s own release notes. Treat each date as approximate and each feature as “arrived around this release.”

I will skip the private releases and jump right into the 1.0 Feature Sets

Revit 1.0 April 2000

Released 25 years ago the month I wrote this, April 2025.

• Parametric Change Engine – Changes to one element in the model meant changes to every element in the model. Moving a wall would update: walls, floors, roofs, the flipping dimstrings associated with the wall, and annotative notes. This feature alone is nearly impossible to build. (Parametric means elements are defined by rules and relationships rather than fixed coordinates; dimstrings are the chained dimension strings measured off that wall.)
• Graphical Family Editor – A graphical interface for designers to create parametric objects – making parametrics easy to access for designers for the first time. Dynamic blocks only came to AutoCAD in 2006. (A family is Revit’s parametric object: a door, a window, a casework type, with its sizes and rules. Dynamic blocks are AutoCAD’s comparable feature; “2006” is the AutoCAD 2006 release.)
• Automatic Coordination – Any change made in a view was reflected across every view and related documentation

What it did for the set: almost everything. The change engine plus automatic coordination is the moat in its first form. A plan, a section and a schedule are three views of one model, so they can’t disagree with each other the way three separately drafted CAD sheets can. In CAD, keeping the sheets in agreement is a person’s job. In Revit it’s the software’s.

Revit 2.0 August 2000

At the time shipping a new version FOUR months after the first one was seen as lunacy. Are these guys nuts? Check out their velocity:

• Massing Study Tools – Simple 3D form modelling
• Building-Scale Detailing – Wall section editing and the ability to do larger scale drawings (DETAILS)
• Family Editor Improvements – custom doors, windows, edit in place
• AccuRender Integration – Raytracing & basic photo-real rendering
• Content Libraries and Catalogs – Predefined components (Doors and Windows) and gearing up for manufacturer catalogs. This enabled architects to place common elements without needing to model them from scratch
• Version 1 was impressive but very buggy. V2 Had a whole lot more stability
• Speed and Model Regeneration, even for larger models
• Revit 2.0 File type (non-backwards compatible); this was a free upgrade to everyone that bought in the first year
• UI: everything was edited in a dialog box, there was no parameters panel with immediate canvas updates yet

All the hallmarks of moving the Technology “Parametric Propagation Engine” towards product market fit are apparent here. They billed $200/ Mo & with no options for annual or perpetual licenses. The early “BIM” took a much larger mindshare from the AEC industry–and there was nothing known as BIM at this time. (BIM, Building Information Modeling, is the coordinated 3D model with data attached to every element; the name came later.)

What it did for the set: mostly model features (massing, rendering, content), plus two that matter more than they look. Stability and regeneration speed are what let a team trust the set: a set you can’t regenerate on a large model, or that crashes mid-issue, isn’t change control. And building-scale detailing is the first step toward the big-scale drawings a CD (construction document) set needs.

Revit 2.1 October 2000

• Bugfixes and Performance Improvements

What it did for the set: more of the same trust. Unexciting, and exactly the kind of release a document tool needs.

Revit 3.0 February 2001

The multi-role demands upon the software that AEC has? The need to allow more than one editor into the machine. (AEC: architecture, engineering and construction.)

• Multi-User Collaboration in Parallel (Worksets) – Users could check out a subset of the model to edit it, while others could only view it. Once checking the subset back in, it would engage the parametric propagation engine and get everyone up to speed on changes.
• 4D Construction Phasing (Phases) – Setting existing, demolition, new construction started here. Keeping all phases in one model was unheard of at the time
• Compound Walls, Floor, Roofs – Using multiple materials for how these elements are comprised of physical materials. (So a wall knows it is stud, sheathing and gypsum board, not just two lines.)
• Detail views and Drafting views – 2D specific detailed linework attached to named views–used for call out views and section blowups. This enabled construction documents on par with CAD
• Rendering – added radiosity for more natural lights within spaces. They also leveraged Accurender procedural plants to create geometry without needing to create geometry
• Expanded Library – drag and drop placement of doors, windows, fixtures from family libraries

What it did for the set: this is the release where, by this record, the moat closes. Worksets let a whole team edit one model without forking it into copies that drift apart. Phases put existing, demolition and new work in the same model, so the demo plan and the new plan come from one source. Detail and drafting views let the hand-drawn part of a CD set (the details) live in the same file as the model and get referenced from it. Compound assemblies make the wall tags and sections tell the truth. Only rendering is purely a model feature.

Count the dates as recorded here: 1.0 in April 2000, 2.0 in August, 2.1 in October, 3.0 in February 2001. Four releases in ten months, and by the fourth the whole loop was in place: one model, many views, many editors, every phase, details attached, and a set that reissues without drifting.

I will leave this analysis here for now. It shows what is accomplishable by a startup in the early 2000s, and we should all learn from it and respect its velocity. It derailed CAD and moved us towards parametric modeling.

Derailed, not replaced: plenty of the industry still runs on CAD and PDFs (that’s the argument of BIMposter Syndrome). What Revit moved was the drawing set, from something drafted to something generated. That’s the judgment this history points to: the moat isn’t the model, it’s the reissue. Anyone trying to cross it has to beat Revit at the boring guarantee: change one thing, reissue the whole set, and only the clouded thing changed. A tool that generates better buildings but can’t make that promise ends up upstream of Revit, not instead of it, which is where TestFit sits today: it fits the building to the site, and the geometry goes through Dynamo (Autodesk’s visual-scripting tool for Revit) into Revit, where the set gets made and reissued.