Amanda Williams, Bruno Nadeau
The democratization of manufacturing has been called the Third Industrial Revolution [1,2], and indeed its global significance may prove to be enormous. Along with advances in prototyping and knowledge sharing, easier routes to manufacturing have enabled the burgeoning "maker movement" and allow HCI researchers to innovate faster than ever before [3,4].
So you have a working prototype for your research project: Do you want to make 100 to deploy and evaluate in situ? Do you want to explore alternative sources of funding? Do you want to spin off a project into a hardware startup? Or do you simply want a dozen or so gorgeous, robust devices for demos?
Chris Anderson has written that standardized design files "allow anyone, if they desire, to send their designs to commercial manufacturing services to be produced in any number, just as easily as they can fabricate them on their desktop" . But while manufacturing hardware may be much easier than it used to be, that does not make it, in any absolute sense, easy. Spokespeople for the maker movement rarely go into detail about the mundane challenges of translating a great prototype into something manufacturable at scale. But we will.
In the spring of 2013, we prototyped a tangible interactive device named Clyde. Clyde is a programmable, Arduino-compatible lamp whose removable hardware modules and open source design allow users to customize how it reacts to its environment. Out of the box, it changes color based on touch and light levels. We raised money and reached early customers on Kickstarter, which allowed us to tackle manufacturing, and we shipped our first batch of 1,000 this past summer. It's been an exhilarating, frustrating, and educational year. We want to share some experiences that few people will have encountered during the prototyping phase: sourcing components and managing a supply chain, designing for ease of assembly and testing, and communicating with your factory.
We manufactured Clyde in Shenzhen, China, because we had a unique opportunity through a hardware startup accelerator to develop our prototype and visit factories there. Known internationally as a manufacturing hub, Shenzhen is a great place to prototype, with diverse components and fabrication techniques available for affordable prices. Even more important, Shenzhen offers a great deal of locally available design and engineering expertise. We quickly discovered that designing and manufacturing our product would not be a simple case of making decisions in North America and then throwing schematics "over the wall" to China.
While prototyping, we began to work closely with a lamp factory. When we first met with them, we had just finished our first prototype, made entirely of wood. The factory engineer pointed out that an all-wood lamp would have problems with heat dissipation from the LEDs. During that meeting, he also pointed out places where we should cut some holes to improve air circulation. After we left, he took our 3D models, ran them through a program that simulated the LEDs' heat dissipation, and sent us results showing where some of the problem areas might be. Over the next couple of weeks, we collaborated on our models, switching to aluminum and adding internal ridges to improve heat dissipation. The resulting enclosure still met our own goals of beauty and ease of disassembly, but also performed well in their heat-map simulator, and later in actual use.
Judging from other designers and entrepreneurs we knew in Shenzhen, this collaborative exercise was the rule rather than the exception. Far from "designed in California, manufactured in China," what we often saw on the ground was an exchange of know-how between makers and manufacturers.
If you're prototyping electronics, sourcing components is usually a matter of visiting an online supplier, ordering a few of each, and receiving them within a day or two, ready to assemble. Shopping for components can become more complex when working with factories. It centers around a Bill of Materials (BoM), which specifies what components and quantities you need, and possibly includes target prices and assembly notes. The BoM is necessary to get quotes; it's usually wise to send it to several different factories to compare their estimates. It is also possible to source components from a third party, though you may prefer to keep your supply chain simple (more on that in the next section).
Every piece that you have to artfully jiggle into place will present a problem on the assembly line.
Settling on a final BoM may require more prototyping. Locally available components can be equally good but much cheaper, making it worthwhile to tweak your design to test them out. In our case, we redesigned our PCBs late in the game because our microprocessor suddenly became expensive and difficult to source. We made our board layout compatible with two different sizes of the same processor, which gave us more flexibility in sourcing. That revision saved us about $2,000. Sourcing difficulties can happen for a variety of reasons: A large company might be buying up huge quantities, a fire at a factory might destroy stock, or a politically motivated clampdown on imports can limit supply.
Clyde is composed of several different sorts of parts: injection-molded plastic, die-cast aluminum, silicone, PCBs (with dozens of different basic electronic components), and gooseneck tubes, not to mention USB cables and power supplies. Each of these pieces comes from a different supplier. Some of them require surface treatments to make them shiny, color them, or prevent rust, which can involve a subcontract with yet another factory. With this many moving parts in our supply chain, it's practically inevitable that a single manufacturer in our critical path will create major delays for everyone if they run into a problem or fail to deliver on time.
If we had it to do over, we would have designed with simplicity in mind and aimed for fewer different materials and manufacturing processes. As interaction designers, we're already familiar with the notion that design must work creatively with constraints; design for manufacture introduces a new family of constraints related to an entire system of sourcing and assembly.
While you were prototyping, you probably knew that prototype well and did a lot of work with your own hands. When it comes to manufacturing, however, every piece that you have to artfully jiggle into place will present a problem on the assembly line. While still prototyping, consider what it means for your device to be assembled by a stranger who may not know or care much about it, who might be bored or sleepy, or who might be thinking of finding a new job. Once you decide to manufacture a product, your documentation of the assembly process will be incredibly important. Expect to see some of your photographs and (translated) instructions posted at workstations on the assembly line (see Figure 1).
We learned a few concrete design lessons that can reduce difficulties during assembly. For example, we now try to design asymmetrical attachment points (e.g., screw holes, snap fits) to ensure that there is only one way to attach one part to another.
The most important thing we could have done was to anticipate the PCB assembly process. Depending on the PCB, assembly may require some or all of the following steps: surface mount placement (first side), solder reflow in an oven, surface mount placement (second side), reflow (again), manual placement of through-hole components, wave soldering (this involves a wave pool of molten metal!), and finally, hand soldering if necessary. If you can fit all your components on one side, you'll save a couple steps. If you need wave soldering, you'll have to orient your through-hole components so they're less likely to be knocked askew by that deluge of melted solder. If you can rely on automated processes and minimize manual soldering, you'll save labor costs and reduce your error rate. To the inexperienced (as we were), this can sound like an intimidating list of considerations, but early factory visits can teach you a lot about these processes, especially if you are thinking ahead about what kind of impact it should have on your design.
No matter how well you design your product, there will be a small percentage of errors in production that need to be caught and fixed. For every PCB you design, you'll need to make another custom PCB to connect to and test it. You'll need to make a test jig that runs those tests quickly and reliably. And you'll need to write testing software. Along with your assembly instructions, your test plan will be one of the most important pieces of documentation you write; you should start working on it early, refining it as you refine your product. Testability can even be a consideration in your choice and placement of components: Will a line worker need to plug and unplug a connector by hand, or can they just pull a lever to automatically make a connection with a row of pins?
Along with your assembly instructions, your test plan will be one of the most important pieces of documentation you write.
An important consideration in designing your test plan and equipment is to eliminate or at least minimize manual intervention. Testing identical PCBs over and over is a boring job, and workers' attention will wander. Even simple tests that require someone to check if a light is on and confirm with "yes" or "no" can invite human error. Putting a light sensor in the test jig will take a bit more time, but it will be time well invested.
Despite the rigors, we enjoyed the process of building our testing equipment. We tend to draw contrasts between mass manufacturing and the small-batch production that characterizes maker culture. But just as makers depend on mass-produced electronics to prototype affordably, mass manufacturing (in Shenzhen at least) depends on small businesses that can mill customized test jigs to order. Watching our own machines take shape was, in a way, as magical as some of our breakthrough moments in the early stages of hand-prototyping our product.
If you've read this far, we can finally admit that all of the advice and anecdotes we've related here can be summarized in just a few concise sentences. First, document everything religiously, because you'll be working with other people whose know-how doesn't necessarily overlap much with yours. Second, collaborate with your manufacturers and suppliers, and respond to their capabilities. Third, though you're still fundamentally working on a design problem, manufacturing will throw a new set of constraints at you that you will need to work with.
The underlying commonality to all of these points is that manufacturing a product is really an exercise in good communication. We benefited enormously from spending time in Shenzhen and meeting our manufacturing partners face to face, complete with tea, really strong rice liquor, and occasional karaoke outings. Regular meetings allowed us to resolve problems and miscommunications in minutes that would have taken days over email with a 12-hour time difference. Factory visits allowed us to see for ourselves how our potential partners treated other clients' products and intellectual property, what assembly processes we needed to anticipate, and what their working conditions were like. Clean facilities, clear safety procedures, and a collegial relationship between line workers and engineers are, of course, an ethical issue, but at the same time, these are indicators that your product will be made with care, and that any mistakes will be reported and resolved rather than hidden.
And last but not least, maybe it's good for us to see for ourselves how the sausage is made—to meet the people who do the (at best) boring jobs that enable us to realize our dreams, to see how much or how little plastic waste our creations generate, and to understand how our inventions really are a team effort.
2. The Economist. The third industrial revolution (Apr. 21, 2012); http://www.economist.com/node/21553017
3. Tanenbaum, J., Williams, A., Desjardins, A., and Tanenbaum, K. Democratizing technology: Pleasure, utility and expressiveness in DIY and maker practice. Proc. of Conference on Human Factors in Computing Systems. ACM, New York, 2013, 2603–2612.
Amanda Williams is co-founder and CEO of Fabule Fabrications. She's in charge of creating beautiful interactions and hardware. Indecisively, she loves both qualitative user research and hardware design. firstname.lastname@example.org
Bruno Nadeau is co-founder and CTO of Fabule Fabrications, in charge of product design and software wizardry. With a degree in computation arts, he's made some creative interactive installations and physical computing pieces. email@example.com
One force behind the democratization of manufacturing is the increasing availability of funds to pay for it, through crowdfunding. If you want to crowdfund a project, here are a few tips to keep in mind:
It's not just about the money. Crowdfunding can help you establish whether there is any demand for what you're making. You can even A/B test the popularity of individual features by making them available at different reward levels. You'll get great feedback from your backers. And if your campaign does well, you'll discover that journalists trend-hunt on crowdfunding sites.
Prepare. A lot. A good-quality video is highly correlated with success, and a mature, beautiful prototype will also be a huge help. Scour successful and failed campaigns similar to yours to research appropriate content, funding goals, and reward tiers. You should simultaneously be developing your network, months ahead of time if possible. Build an email list, which has relatively high conversion rates. Talk to media and decide whether you want them to embargo the story until you launch or generate hype beforehand. When you launch, send personalized email messages to everyone you know; your campaign's performance on the first day is crucial.
Be prepared to scale, or put limits on it. Set your goal higher than you think it needs to be; manufacturing can have hidden costs. Ask yourself what you'll do if your campaign blows up: Are you using a manufacturing process that scales easily? Alternatively, you can set a hard limit on all of your reward tiers, or set them to deliver at different dates.
Be honest and transparent with your backers. If you're new to manufacturing, you will make mistakes and encounter delays. Most people on crowdfunding sites are not looking for perfection; they're looking to be involved in the creation of innovative products. Invite them in, and they can make your crowdfunding experience incredibly rewarding.
Copyright held by authors. Publication rights licensed to ACM.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2014 ACM, Inc.