Long before Amazon could effectively deliver on it’s promise of 2-day shipping for Prime customers (early Prime users will remember it often took 3+ days), the industrial supply company McMaster-Carr was selling a large catalog mechanical components with shipping times over ground that were regularly faster than Amazon Prime. You could even get toilets and night vision overnight!
I was designing and building a lot of machines at the time (2008–2015) in coursework, internships, research, and jobs. I noticed that quick access to parts along with a reasonable user interface for looking up parts had created a design methodology that was ubiquitous but not much talked about. I call it DfMc (Design for McMaster-Carr), a thoroughly tongue-in-cheek reference to the much more widely known concept of Design for Manufacturing.
A feature of DfMc is being unable to make complex custom parts. This often is because of both budget constraints (common in student projects and university research), constrained access to machine tools (shared lab CNC mill, etc), and, frankly, design engineer inexperience (myself included).
When you have quick access to a large catalog of stock parts and limited access to custom parts, design decisions take on a specific character. I found that as I became better at designing machines with the aforementioned constraints, I would actually go to McMaster-Carr as the very first step in the design process. If you had to make your machine work with mostly off the shelf components, the McMaster-Carr catalog became essentially a directory of all the possible machines you could build. In DfMc, McMaster-Carr is your package manager, to use a lazy software engineering analogy. In software, you had better be sure you can get adequate functionality from an API before you build something that connects to it.
To connect the functions of stock parts, custom parts would be minimized down to essentially flat plates with holes (sometimes taped). An undergraduate with a drill can make these quite easily. A machinist can make these in an almost completely automated process with a waterjet. To use the software engineering analogy again, these custom plates are the application/middleware code connecting your various APIs.
Another factor that drives machine designers to DfMc, which I have neglected to mention until now, is short timelines. These come in many forms — project deadline, grant application deadline, conference/journal submission deadline, or investor demo. DfMc, more than anything else, can be extraordinarily fast. When your custom parts are dead simple and your stock parts can come overnight, you can build machines at unparalleled speeds. I’m convinced that 90% of any reputation I have as a rapid prototyper is due to being good at navigating the McMaster-Carr catalog. That, and developing a good working relationship with my machinist, but that’s a topic for another post.
Deadlines can produce fast results, but it is well known that they have many unintended consequences. With DfMc, machines can be built quickly and with low manufacturing experience. This can fool you into thinking it’s a linear progress towards the final product. A machine built with Design for McMaster-Carr should be treated as a proof of concept and should be completely reconsidered when viewed through any other Design for X lens. The Juicero juicer was much lampooned for being poorly engineered, among other criticisms. I suspect an ineffective transition away from a DfMc design contributed here.
The exceptions to needing to redesign as you scale are if the machine only needs to be built once and/or it can be capitalized. Ironically, while DfMc is often motivated by budget constraints, the end product is often much more expensive than other methods. McMaster parts are horribly overpriced, and the simple custom parts are often more numerous and use more material as compared to a cleverly designed stamped or injection molded part.
Think I’m dead wrong? How have you transitioned away from DfMc designs after prototyping? Please feel free to reach out!