The Prefab Pivot

One of the main themes of this newsletter is that the construction process is significantly constrained by geography. For work done on-site, workers and equipment need to be able to reach the construction site in a timely fashion, and that site will be different for each new building project. For work done off-site (prefabricated components, etc.) the same constraint applies - the material needs to be transported to the jobsite in an efficient manner. Since building materials are bulky and low-value, this has traditionally limited how far components can be shipped cost-effectively.

Different geographies will have different permitting jurisdictions, which may have different code requirements, different building departments with different levels of stringency, and different inspectors who care about different things. They’ll also have different design requirements based on different climates, seismic or wind conditions, or local infrastructure. And at the micro-geography level, every building site will be a little bit different, and impose its own set of constraints and uncertainties - a blocked view, an inconvenient boulder, a 40 inch gas line that must be built around, etc.

These geographic constraints have shaped the US construction industry (particularly residential construction) into one that is largely capital light and decentralized - builders have little vertical integration, instead preferring to subcontract out much of the building work to local trades, rent equipment as it’s needed, etc. Even large homebuilders like Lennar, which build tens of thousands of homes a year, subcontract out much or all of the actual construction work.

This sort of structure allows for maximum flexibility - a business can scale its capacity up and down as needed, based on the number and type of projects they’re currently doing. While many of them would undoubtedly prefer to keep more trades in-house, this is a tough proposition. Construction takes dozens of different trades, each one of which might only be needed for a small fraction of the overall project (the building might take 10 months to complete, for instance, but the drywalling might only take 2 weeks.) There’s a big difference between ‘having a lot of projects’, ‘having so many projects that you can always keep your drywall team busy’, and ‘having so many projects that variation averages out and you have a relatively uniform amount of drywall work’. 

Geographic constraints exacerbate this - driving distances limit how efficiently you can reallocate workers between projects, and building department interactions (‘Have we got the permit yet? Nope.’) mean it's often unclear exactly WHEN work will start. Combined with the naturally clumpy nature of construction work (a comparatively small number of high-dollar-value projects) and the natural cyclicality of real estate, it means a lot of variation in output, and limited means to damp that variation.

A decentralized, flexible industry structure works with this constraint - independent subcontractors, materials suppliers and equipment rental companies serve all builders in a local geographic catchment area, and builders can use locally available subcontractors without having to carry them on their books. These wheels are greased by a community of practice, where everyone shares similar ideas about how a building will go together, reducing coordination costs and making it possible to use a new team for each building that goes up.

And of course, it also results in what’s considered an ‘archaic’ fabrication process, where a large fraction of the work is done on-site, by hand.

Efforts to modernize this process have traditionally centered around prefabrication - doing as much work as possible in off-site factories. But the most common implementation strategy - a large, centralized factory that achieves economies of scale by maximizing production volume - works against the geographic constraints and high output variation in construction. Centralizing production in an off-site factory adds potential efficiencies, but it also adds transportation costs, as well as overhead expenses that become a liability if (slash when) the market shifts. The most popular prefabrication methods in the US thus tend to be either low-capital, or flexible enough that they can serve a wide output mix, or both (such as wood trusses), which tends to mean simple, low-level components.

The history of ambitious factory-built efforts in the US is one of trying to overcome this dynamic - of trying to crack the nut of getting a large volume of reliable work through your factory. But what would a modernized fabrication method that instead tried to work with it look like? A new class of construction startup has the potential to be that.

The Fabless Prefabbers

These companies describe themselves in different ways, but we might call them ‘fabless prefabbers’ - companies that offer a prefabricated building system, but outsource the actual fabrication to others.

• Modulous - A UK-based construction startup that is pursuing multifamily construction with a volumetric modular building system

• PT Blink - An Australia-based construction startup that uses a unique post-tensioned steel deck system

• Juno - A US-based construction startup that is utilizing mass timber components.

(We might also include Hypar here, which is trying to build an API to connect building design with fabrication. But they’re operating at a slightly different level in the abstraction layer.)

Despite the different framing strategies, these companies are all pursuing a similar strategy - offer up a prefab ‘kit of parts’ (manufactured by others) that developers and builders can use to create their buildings. Here’s the cofounder of Juno in an interview for dezeen:

the first thing that we really did was decided to look to see what were the smallest number of parts that we could design to actually keep the skew count low, keep the quality really high.

And then take those very small number of parts to build a supply chain around delivering those parts to our project sites. So that we could put these things together.

And those parts are made such that they can live independent of each other, they can be second or third sourced if need be, we can always chase the best technology, the best vendor who's doing the best work, we can learn from their techniques.

We then take those parts, make them very predictable, make them relate the same way every time they're using the project, and then allow a lot of flexibility in creating many, many different buildings on different sites.

Then make it such that the assembly of the project on the site is really like a factory assembly line, but it's on a construction site. So you've really simplified down, potentially, the assembly of the project because much of it is done ahead of time.

And here’s copy from the PT Blink site:

The Blink DMI™ process delivers multistorey buildings as a kit of configurable parts that are manufactured offsite and integrated very quickly and safely onsite.

By designing, manufacturing and integrating, with PT Blink you maximise speed and reduce costs, while maintaining the ability to configure the final product for a wide range of design expressions.

And here’s language from Modulous’ site:

Designs are based on our proprietary Kit of Parts (TM), which is a series of sub-assemblies procured from the established supply chain that can be brought together to create modular homes in facilities local to site. 

(good luck enforcing that trademark, guys!)

The idea of construction using a standard ‘kit of parts’ is also not new. But it’s previously taken shape differently - as a strategy a prefabber can use to increase their factory volume, by using a small number of repetitive components to produce a large number of possible building types. Blokable and Plant Prefab both pursue a kit of parts approach.  But with the fabless companies, the potential here is slightly different - rather than a way to maximize production volume through a factory, the kit of parts becomes a way to build a construction platform.

Construction as a platform

Briefly, a platform is a type of business that aims to facilitate transactions between different parties, and takes a cut of the resulting transactions. A classic example of a platform is a credit card - American Express signs up merchants and consumers with the card, and then takes a processing fee from the merchant, and a lending fee from the consumer. Platforms have traditionally added value by solving coordination problems - instead of having to negotiate a short term loan with every possible merchant, American Express goes and does that for you. The last 20 years of the tech industry has in many ways been a story of building platforms on top of internet infrastructure, and many of the largest tech companies - Amazon, Facebook, Google, Uber, eBay - are platforms in one way or another.

So what is the potential for a construction platform? What coordination problems are we trying to solve?

The most natural way of thinking about the potential for a construction platform is that of a computer operating system (OS) (operating system metaphors are extremely overdone, so let’s see if we can’t do better.) An OS provides an interaction layer that allows hardware and software components to talk to each other. With an OS, a hardware or software builder can design for the requirements of a single operating system, rather than trying to navigate the requirements of thousands of possible components that might be in millions of different configurations. By defining the rules that hardware and software must play by, and managing component interactions, an OS can abstract away much of the complexity. 

Likewise, the potential for a construction platform is based on having a standard set of components and well-defined (and enforced) rules for their interaction, making it easier for building component manufacturers, prefabbers, and designers to interact.

The current system of construction ‘solves’ this coordination problem by having converged on an informal set of typical ways that a building goes together - our ‘community of practice’. But this leaves a great deal to be desired. For one, it makes it very hard to introduce new elements into the building ecosystem - since no one controls the standard, the only way is the brute force, one-customer-at-a-time approach. For another, ‘standards’ often end up being someone loose, and encompass a lot of meaningful variation - different component manufacturers will have different preferred materials, different component sizes, etc (‘we use an 8ft wide panel, but our competitor uses a 10ft wide panel’.) It also becomes hard to introduce improvements that would be globally beneficial but locally costly. It would benefit nearly everyone, for instance, if all light framed wood construction was framed using OVE methods, but any individual builder trying to make the change will face a cost for doing so.

By building a standard interaction layer between hardware (building products and systems) and software (the architectural design of the building), a construction platform can theoretically solve these coordination problems. On the ‘software’ side, architects and GCs can design their buildings from a kit of parts without worrying about having to adapt their design to a particular manufacturer's system, or figuring out how to make the different components play nice with each other. On the ‘hardware’ side, by tailoring their components to the platform requirements, fabricators instantly get access to a huge group of potential users that will be dramatically easier to sell to (because much of the coordination work has already been done.)

Power of the platform

A lot of interesting things become possible once this structure is in place.

New building product introduction becomes easier, because the risk for both product developers AND users is greatly reduced. The effort done to coordinate components by the platform means higher quality, fewer errors, and less wasted coordination time for practitioners. Jobsite feedback can be incorporated into product improvements, which in turn can be deployed to the entire ecosystem at once (‘based on feedback that the XP panel is hard to hold, we’ve added a lifting handle to the bottom of it in version 2.1.1) Discoverability of new products is improved, since the platform provides a method for finding them. The potential for user feedback (“this factory totally dropped the ball, 1 star”) and gatekeeping could enforce better behavior, better outcomes, and result in higher-trust interactions.

By not doing the actual fabrication themselves and instead just taking a cut, the platform has incentive to maximize the value of all the transactions that are taking place - in other words, to try to ensure the best outcomes for everyone. There are any number of things a platform owner might do for this:

• Offering a training and certification program for users and installers of this system, to try to enforce a standard of quality.

• Fund (or otherwise encourage) the development of new products that they see a need for.

• Offer or help arrange financing for projects that might otherwise be difficult (since prefab manufactures components early on in the process, it has different financing needs than conventional construction.)

• Encourage/require the use of high level-of-detail models as standard project deliverables.

• Create software (API, Revit plugins, etc.) to make design easier

And of course, there’s the natural flywheel of the platform at work - each new user, and each new use of the platform makes it a little bit better, making it that much more enticing.

Level of integration

There’s multiple ways that a platform could take shape. One potential axis of variation is ‘level of integration’ - how much control does the company maintain over the parts and how they work with each other? Typically there will be a tradeoff - the tighter your integration, the less flexibility and breadth your system has, but the higher potential performance.

A smartphone analogy is illustrative here. On one end of the spectrum you have iPhones and the iOS platform. Apple maintains a very tight control over this platform - you can’t install iOS on any random smartphone, and Apple spends a huge amount of coordination effort vetting hardware suppliers and making sure components meet their standards. This lets them very tightly integrate the hardware and the software, and achieve a high quality user experience.

On the other end of the spectrum you have Android, which is much more flexible. The core Android components are open source, allowing anyone to use it on any sort of device they want, from high end smartphones to simple tablets designed for booking conference rooms.

A construction platform likely faces a similar tradeoff - the more control you exercise over your components, the better performance and user experience you can achieve, at the expense of flexibility and breadth of your offering. Some companies may opt for a small library of components that they’ve carefully vetted (or even maintain ownership of), while others may opt for a breadth strategy, trying to have as large a library of components and systems as possible.

Is there a strategy that’s superior? It’s hard to say - the smartphone example shows that an industry can clearly support both (though it’s notable that while Android is on many more devices, most of the profits in the industry have accrued to Apple.) Clay Christensen’s work suggests that as a product’s performance improves past the level that most buyers need, emphasis shifts towards more modular, flexible product architectures that can accommodate a broader array of use cases. And current methods of construction are decidedly skewed towards the flexible, modular end of the spectrum. On the other hand, changing performance criteria (such as energy efficiency or construction speed) may require a tighter integration.

In some ways, this becomes a refinement of the Katerra model - linking material suppliers and developers with a huge offering of potential product offerings, but omitting the part where you do the design and fabrication, leaving that to others:

Reverse TSMC

In talking about the potential for fabless prefabrication, we would be remiss if we didn’t mention the semiconductor industry, Taiwan Semiconductor (TSMC), and the rise of ‘fabless chip designers’. 

TSMC is a contract semiconductor manufacturer. Unlike Intel, which integrates chip design and fabrication, TSMC manufactures the designs of other companies - AMD, nVidia, and Apple all use TSMC to manufacture their chips.

This contract manufacturer model was relatively novel when TSMC was founded, per Ben Thompson: 

at that time all chip manufacturing was integrated a la Intel; the few firms that were only focused on chip design had to scrap for excess capacity at Integrated Device Manufacturers (IDMs) who were liable to steal designs and cut off production in favor of their own chips if demand rose. Now TSMC offered a much more attractive alternative, even if their manufacturing capabilities were behind.

The rise of TSMC enabled a new type of semiconductor company - the fabless chip design company:

Meanwhile, the fact that TSMC existed created the conditions for an explosion in “fabless” chip companies that focused on nothing but design. For example, in the late 1990s there was an explosion in companies focused on dedicated graphics chips: nearly all of them were manufactured by TSMC. And, all along, the increased business let TSMC invest even more in its manufacturing capabilities.

The opportunity of TSMC and a construction platform both stem from decoupling design and fabrication - in both cases it’s difficult to have an integrated design/manufacturing that can effectively cover all a customer’s use cases. TSMC centralized fabrication, and decentralized design, which was instead done by a group of companies each focusing on their particular use-case (their ‘geography’.)

A construction platform, on the other hand, reverses this dynamic - instead, it centralizes design (in the form of a standard kit of parts) and decentralizes fabrication (lots of fabricators, each covering their own particular literal geography that they can effectively serve.)

Zero to one

This is an optimistic view of what a construction platform could theoretically achieve. It’s an open question if any of these companies will be able to achieve it, or even if they’re aiming at such an expansive vision. The initial building systems these companies are using (post tensioned steel, volumetric modular, and mass timber) are targeting a comparatively narrow swath of the market (though I suspect their VC backing will tilt them towards aiming bigger.) Most of the benefits I described are predicated on a platform that encompasses a fairly large volume of construction work.

I suspect that the flywheel effect will be quite strong once it gets going (both because I think there’s likely a lot of low-hanging fruit to pick, and because once enough people sign on it becomes an easy sell to risk-averse builders adopting a ‘wait and see’ approach), so it may simply be a question of whether these companies are able to execute and successfully build momentum. So I’ll be watching them with interest.

---

These posts will always remain free, but if you find this work valuable, I encourage you to become a paid subscriber. As a paid subscriber, you’ll help support this work and also gain access to a members-only slack channel.

Construction Physics is produced in partnership with the Institute for Progress, a Washington, DC-based think tank. You can learn more about their work by visiting their website.

You can also contact me on Twitter, LinkedIn, or by email: briancpotter@gmail.com

I’m also available for consulting work.