In the second of a multi-part guest series from Elaine Chen, we explore sourcing strategies for building 2000 units, the “uncanny valley” of consumer electronics manufacturing. These posts are meant for new hardware teams going through the process for the first time.
In the previous post, we discussed the components that make up a hardware product, including COTS and custom parts. We discussed how to source commodity and critical COTS parts. All this is non-obvious but still relatively manageable.
Where it gets much less manageable is when you come to custom parts: The custom electrical / electronics and mechanical components.
Let us deal with the easy parts first: Printed Circuit Boards (PCBAs) and cable harnesses.
Fortunately, a very robust industry exists where companies can build and test custom PCBAs and cable harnesses at various quantities. Assuming you want your product to be durable, you will almost never want to assemble your custom PCBAs and cable harnesses in house. These days almost all PCBAs are surface-mount (SMT), not through-hole. Assembling an SMT PCBA requires placement robots and special equipment that are very expensive and hard to justify. You can use a variety of prototyping services to create hundreds of fully tested boards very cheaply.
Cable harnesses are a little less obvious. It would seem that having summer interns crimp wires into pins and then insert pins into connectors would be a cheap way to create your cable harnesses.
There are two problems with this approach. First, you are looking at a lot of cable assembly equipment, and good equipment is not cheap. Second, you will have tremendous challenges with quality control. You will be amazed at how easy it is for even trained technicians to run into quality problems that result in intermittently malfunctioning cables.
For low quantities, find a prototype-quantity cable harness supplier to create your lots of 50 or 100 for you. For higher quantities, outsource to Asia. This is the one area where Asia is a viable choice even for moderately low quantities if you can tolerate the turnaround time.
Sourcing custom mechanical parts: Plastic Parts
Now let us come to custom mechanical parts, starting with custom plastic parts. This is where most of the head-scratching occurs in the quest to build 1000-2000 units.
Let's say you have a design of moderate complexity, with something like 10-20 custom plastic parts. How do you make the plastic parts? 3D printing is not viable.
First, the surface finish is not there. Second, the structural strength is often not there due to the limitations in materials and processes. Lastly, if your product needs to pass regulatory approval for sale in Europe - i.e. you need the CE mark - plastic parts must have flame retardants built in, and that is very hard to find in RP parts. Cast urethane parts come closer in surface finish and structural strength but the per-part cost is still prohibitive. This leaves molding and forming as the viable processes.
Injection molding is by far the most practical way to make small to medium sized plastic parts. The cost, once you invest in molds, can be a few cents per part at scale, and potentially less than $1 per part at lots of 2000. The problem is that you must invest in molds. There are really only two choices.
Generally speaking, injection molded plastic parts will cost less if made in China than if made in the USA. Making your plastic parts in the USA has the obvious benefit of proximity between R&D and manufacturing. It is much easier to have your engineering team live there for a couple of weeks to tweak all the molds so everything fits correctly.
However, it is significantly cheaper to do it in China, and the minimum order quantity for plastic parts in China can be quite low in the beginning (500-1000 range is not uncommon), provided that the product team can convince the supplier that higher quantities are on the horizon.
The "right-shoring" decision is up to the product team and the unique needs of the project. It is sometimes faster and easier to do things in the USA the first go-around, and sometimes this is worth the price premium.
As for metal parts, the first question to ask is at the design phase. Why do you need metal?
Sometimes the choice for a metal part might be a measured decision to minimize plastic parts. For instance, you may elect to use sheet metal instead of plastic for internal structures. Each part will cost more, but you will have a much faster turnaround time for production parts, and you also get to postpone any significant investments in tooling. It is a very good strategy if you are test-marketing a first design and are quite sure you will have design changes before scaling up to mass production.
In other cases, metal is chosen for structural reasons. For instance, you are building an oddly-shaped, high-torque gearbox that will transmit high loads to the housing of the gearbox, and plastic just does not have the material properties to support these loads.
The question to pose is, can you change the system architecture so you can swap the metal parts for plastic parts? Modern carbon-filled or glass-filled plastics can have very good material properties at a fraction of the weight and cost of an equivalent metal part. Sometimes you need to think out of the box during the architecture phase and consider the consequences of certain design choices all the way to the end of the process. What is fast and easy now may create a permanent headache once you go into production.
Let's say you have very good reasons to need metal parts. Sheet metal is a good choice for reasons previously stated.
If you must create a complex 3D shape, you will need to prototype the part with a CNC process. This part can cost hundreds of dollars. The long term goal might be a cast part that is subsequently post-machined to achieve the tolerance needed for that part. But tooling for die casting aluminum is astronomical, and the lead time is usually 4-5 months. In this case, you might have a heuristic for whether to invest in die cast tooling.
Let's say you are making 500 copies of this product. Can you make back the per-part cost in 500 parts? Given the high cost of CNC machining, the answer is often yes. If you can come up with a clear ROI that you can count on, and you can wait 5 months, go for it. If not, you will just have to grin and bear the CNC cost.
This is more than enough sharing on component sourcing at the 1000-2000 unit level. Now you need to assemble them. What are your options here? This will be the topic of the next post.
About the Author
Elaine Chen is a startup veteran and product strategy and innovation consultant who has brought numerous hardware and software products to market. As Founder and Managing Director of ConceptSpring, she works with executives and leaders of innovative teams to help them set up and run new product innovation initiatives with the speed and agility of a startup. She is also a Senior Lecturer at the MIT Sloan School of Management and the Martin Trust Center for MIT Entrepreneurship. Follow her at @chenelaine.