#####10/4/2013— [Jon McKay](/blog/"http://technical.io" target="_blank">hardware more accessible to our fellow web developers, but it’s a hardware company nonetheless. We’d all worked at startups before and knew how to move fast with software, but making a hardware company just as agile was a completely new experience.
There are eight ideas that helped us move fast, and if you come from a web background and are looking to build a hardware company like I did, hopefully they’ll help you move faster too:
1. The Lean Startup Still Applies
All the rules for lean startups still apply for hardware companies. Make the easiest, quickest prototype and get user feedback on it. Once you can prove you’re heading down the right path, iterate as quickly as you can. We prototyped with an Arduino for the first few weeks to get developers’ feedback on our direction. Only after ensuring we struck a chord did we start using more specialized hardware.
2. Use Evaluation Boards to Save Money
A great way to create hardware prototypes iteratively is to order “evaluation boards” for the chips you’re going to use. Most integrated circuit (IC) manufacturers sell pre-made circuit boards for some of their chips that include all of the support circuitry the chip needs to work, along with nice connectors for integrating the board into a more complete system. As you build up your prototype, you can just wire those eval modules together. The major draw here is that you can put off designing your own circuit board as long as possible. When small manufacturing runs and assembling your own boards cost you thousands of dollars, eval modules can really spread out your runway.
3. Use Reference Designs to Maintain Sanity
You’ll have a lot less trouble with weird hardware bugs if you stick to reference designs. Every (good) manufacturer will release documentation with their ICs that contains a section with a reference design. The reference design gives you an idea of all the other components (resistors, capacitors, etc.) that you’ll need to support the IC (see example for our WiFi chip below). Stick as close to that design as possible (especially for high frequency chips) and your technology will work much more reliably.
4. Use the Web for Validation
Driving traffic to your hardware’s website is crucial for gaining early mindshare and validating your ideas. When we got started with Tessel, we felt like we couldn’t get an accurate picture of how our users would like the device by reading about it on a website. We felt that they wouldn’t “understand the experience”. As if there was something magical about using the device that words just can’t describe. The fact is, if you can’t get across the value of your product with words, then you’ll never be successful. Almost no one will have the opportunity to try your product before they buy it. Make a website and get it out there as soon as possible (well before you’ve finished actually making it). When our website was picked up on HackerNews (a month before we planned to launch ourselves), we got over 10,000 awesome, excited people on our mailing list that we could notify when our crowdfunding campaign began.
5. Crowdfunding Is Not For Revenue
Crowdfunding is completely revolutionizing the hardware world, especially hardware startups. By gauging interest and essentially eliminating the need for up-front inventory, crowdfunding has changed the way entrepreneurs start (and even run) their hardware companies. But for the time being, crowdfunding isn’t so much a means of making money on a finished product, as it is a way of getting the most important pieces of validation: how many people (and who) will click the buy button, how many won’t, and why. You’ll hear it in comments, tweets, articles, and blogs, and you’ll need to analyze the feedback to make your product better. An entire crowd of early adopters become accessible to help you grow your business in the right direction. Crowdfunding makes you data rich, not money rich – use it to grow.
6. Test Driven Driven Development
Test Driven Development is not just a software development process; it speeds up hardware development as well. When the start date for our crowdfunding campaign was approaching, we had a two-week turn-around time for each hardware revision. Hardware companies in China can get that turn around time down to days, but for U.S. manufacturing (at least in Boston), two weeks is fairly fast. Since it takes so long to get the boards, the key is to reduce the amount of time manufacturers are not spending making your boards. Here is Technical Machine’s process:
Create a test suite. Before you have even sent out a board revision to be made, you should have a document describing each feature that needs to get tested, each test for each feature, and how to run those tests. When you begin manufacturing thousands of units, you’ll also need to create hardware test rigs.
Know what you want to get out of each test. Taking the time to decide exactly what you want to measure and how you plan to measure it before powering on the device will save you time, tears, and inventory. In hardware it’s exceptionally rare that “it’ll be different next time” you power it unless you change something physical on the board. More importantly, if the product on your desk breaks, you often need to go out and build a new one from scratch, not just hit compile.
Create a F*#!-Ups Document. There is something wrong with every revision. Expect it. Stop crying. It’s just going to happen. Create a document to keep track of every problem with the board, because as soon as you plug it in, you’ll find them. Run through the whole suite of tests, adding every problem to the list.
Revise continuously. While one engineer is testing the revision and noting mistakes, another should be fixing those problems in the design.
Each board revision consisted of about a week fixing the previous revision and a week waiting for the new boards to come in. During the manufacturing week, we would write all the firmware for new features and the test suites firmware.
7. Free Samples (Shhh!)
Most manufacturers will give you free samples of their ICs if you just ask for them. Some manufacturers even offers samples on their website and will overnight ship them for free. Take advantage of it – they will dramatically cut prototyping costs. We’ve probably used hundreds of dollars worth of WiFi module samples alone.
8. Open Source Hardware Is Growing
As most software developers know, the ability to use open source code dramatically increases productivity and learning. The same principle holds true for open source hardware. Great open source hardware companies like Sparkfun and Adafruit allowed us to
adopt their designs for Tessel so that we didn’t have to come up with the billionth RFID antennae on our own. In turn, we give all of our hardware designs back to the community so that we can grow and innovate together.
Those eight ideas helped us keep our overhead low and get our minimum viable product in front of potential customers as fast as possible. Of course, Tessel is only a circuit board with no industrial designed casing or external components so I can’t speak to the problems you’ll encounter there. As (open source) hardware startups become more and more popular, I hope that we’ll hear from more entrepreneurs who will shed some light on those topics and how we can work to make hardware as agile and iterative as software development has become.