During the summer of 2015, my wife Kaia Dekker and I ran a Kickstarter campaign for the Model 01: a new type of heirloom-grade computer keyboard. Like many people, we spend most of our days at a keyboard for work. But despite having tried most of the keyboards out there, we hadn’t found one we really loved.
The Model 01 isn’t like other keyboards—it’s made from two blocks of solid maple with tactile mechanical keyswitches, similar to those found on classic Macintoshes. We custom-sculpted each of the 64 keycaps to gently guide your fingers to the right keys. From the time we launched the campaign, it took us about two and a half years to get the first production units into customer hands. During that time we ran into a lot of issues and learned how to make a product that folks really love.
Since early 2014, Keyboardio has been our full-time job. Before launching our Kickstarter campaign, we made dozens of prototypes, began building a community, and worked our way through a hardware incubator. During the campaign, we hit our initial goal of $120,000 in just a few hours and ended up raising just north of $650,000.
We didn’t set our goal arbitrarily. $120,000 was the amount we'd need to deliver the smallest number of keyboards we could reasonably expect a factory to be willing to produce. Before setting delivery dates, we consulted friends and mentors who collectively have over a century of manufacturing experience. Many suggested that—if nothing went wrong—our design was far enough along that it could be rolling off the assembly line four months after the campaign ended on July 15, 2015. Having some sense of how little we knew—and that things do go wrong—we doubled the estimate to nine months. We planned to ship by April, 2016. I'm pleased to report that we managed to deliver rewards to more than 95% of our Kickstarter backers by April… April 2018.
Looking back, we’ve had more than our share of challenges. Three stand out in my mind, though: factory selection, the design for manufacturing process, and sourcing our LEDs.
The first difficult moment came when our manufacturer dumped us. We probably should have seen it coming, but we didn’t.
In manufacturing, as production volume ramps up, there’s a dramatic change in the techniques and technologies that are cost effective (or even available). The "right" factory to make 500 of something is almost never the right factory to make 5,000 or 50,000. Because of this, we decided not to sign with a manufacturer before launching our campaign.
Shortly after the campaign, I flew to Taiwan and Shenzhen to meet with about a dozen factories. A few seemed like they were interested in the project. Typically, they were smaller shops, for whom our run of a few thousand keyboards would be a measurable percentage of their annual revenue. One, however, was different.
A partner had introduced us to one of the largest keyboard companies in the world. They brought an entire engineering team to our first meeting. They'd studied our design and had a detailed sales quotation, including proposals for design improvements and cost reductions. They'd even found a supplier who could CNC mill the keyboard's wooden enclosure. Their pedigree was excellent, having made keyboards for many top computer brands.
The biggest worry we had was that we would be one of their smallest customers. Bunnie Huang has written in detail about why it's dangerous to work with a factory that's too big for you. When we raised these concerns, they told us that they saw us as a “prestige” project—and since the Model 01 was a unique design, they’d be able to show off their manufacturing capabilities to future clients.
We thought we'd lucked out. The day we told them that we’d like to work with them, our sales contact informed us he was leaving the company, but that we'd be in good hands with our project manager. As we began spinning up the project, they pushed us to change our design to better match their manufacturing capabilities. They said they didn't need a deposit or pay for engineering support until the design was 100% final. In retrospect, this should have been a red flag.
We ended up spending three to four months doing significant reengineering based on their feedback. Throughout this period, their replies became slower and slower, and we began to suspect that something was wrong. A few weeks after Chinese New Year, the factory admitted that they were breaking up with us. We were basically back to square one. After the ceremonial pint of Ben & Jerry's, we told our backers. They were crushed, but generally understanding. They knew that we were all in this together.
In hindsight, we should have known that a sweetheart deal from a supplier who was completely out of our league was too good to be true. We ended up having to throw out much of the design we'd redone for that manufacturer's capabilities, and completely rebid the project. This misadventure alone probably set us back about nine months.
Design for manufacturing (DFM)
The second major hurdle we faced was that we are not trained mechanical engineers. We've been lucky to get help from talented folks along the way, but there were some things that we just couldn’t figure out, including how to adjust the keyboard’s position and angle. That's what led us to choose factories that claimed to have design expertise.
The factory that ended up manufacturing the Model 01 showed some promising ideas in our early design meetings. As they refined those ideas, however, many of the clever ones didn't pan out. Still, we were making progress, and we agreed that they should build a full working sample of the product.
It took a while for that sample to turn up. When it finally did, we discovered that the keyboard tipped over if you rested your palms on it. Needless to say, that wasn't going to work. As soon as we understood what was going on, we told our backers about the issue.
The factory told us that they thought this problem was only true of the prototype they'd made and once we made an injection-molded version, it’d be a non-issue. This didn’t make any sense to us, since the final version ought to function exactly like the prototype. More importantly, the factory had agreed in writing that they wouldn’t start to making an expensive 1,000-pound milled steel injection mold until we'd signed off on the design.
When we pointed this out, the factory told us that they knew we were tight on time and pretty sure the design would work, so they’d had the subcontractor start making the injection molds. When we got the 'final' parts, they had the same problem as the prototypes. Thus began a months-long saga of redesigns of the steel tooling, culminating in our factory firing the subcontractor.
They've since told us that for anything but the simplest possible designs, they recommend letting a factory do the entire design from scratch or not engaging until we have a complete design. That's not to say that the design will be final—any factory worth its salt will work with you on the DFM process to reduce manufacturing costs and to improve product quality.
The last major issue we ran into was around sourcing the bright, colorful RGB LEDs underneath each key of the Model 01. After a fair bit of research, we selected the APA102C, a high-quality LED from Taiwan. However, our factory identified this LED as a potential supply-chain risk, saying that they’d rather use Chinese-made options that would be easier to get and less expensive.
We told them that we'd be happy to evaluate the other options, but that we were most comfortable with the LED we'd already selected and tested. Some time in 2016, the factory told us that APA was quoting them three months of lead time and that they weren't willing to take that kind of a risk, as they expected to start manufacturing in two months. This is the point where we should have stood our ground and demanded that they stick to our approved vendor. Instead, we embarked upon a months-long long saga of seemingly unsolvable electrical issues, documenting it along the way for our backers.
We agreed to try other LEDs and, at first, everything seemed ok. The factory sent samples. They looked great. They worked just fine—or so we thought. We recorded a quick video showing off the new LEDs and posted it to Twitter. Our friend (and backer) @scanlime quickly replied, asking about the buzzing noise they seemed to be making. Paying careful attention, we discovered that the replacement LEDs did indeed make a maddening buzzing noise. We asked the factory to please check both the original LEDs and the ones they were proposing. They agreed that the new LEDs buzzed. But they were adamant that the old ones did, too.
From there, we cycled through four different alternate LED options. All of them buzzed. Several of the LED vendors brought in their chip designers to talk through the problems. In one case, a chip designer admitted that nobody should be using the product they'd sold us. It was low quality, out of date and there was some sort of IP issue. One vendor even made us samples of a custom LED that they said wouldn't buzz. It buzzed. A consultant redesigned our electronics to minimize the buzzing, but the circuit boards got hot enough to boil water. Finally, in desperation, we rebuilt a circuit board by hand with LEDs from our original supplier. No buzzing.
About three months in, we told the factory that we had to switch back to the original LED. By this point, the lead time for these parts had grown to four months. We were able to get the LED vendor to ship us a small part of our order sooner, but for the most part, we just had to wait.
What we got right
In the end manufacturing the Model 01 took us about two and a half years—far longer than we ever thought it would. I've thought a lot about how we managed to get through the project with our sanity and reputations intact. What I think we did right revolves around two major themes: money and communication. Good financial planning and clear communication with our backers made a big difference.
From the beginning, we've been incredibly conservative about money, always budgeting with worst-case scenarios in mind. Except in cases where we were pretty sure that the money would solve a problem, we opted for slow, careful, and cheap. That gave us the breathing room to be able to handle unexpected costs or problems when they came up. This careful planning helped us get past several moments that could have left us dead in the water.
Before we launched, other creators told us that frequent communication with backers tended to encourage others to pledge. This fit well with who we are. During the campaign, we wrote daily updates. Afterward, we slipped into a monthly cadence.
When things are going well, writing updates is easy. When things go off the rails, communication becomes difficult. It's really easy to convince yourself to delay a backer update until you have some positive news to share.
It's a trap. Don't do it.
Our backers committed to helping make our dream a reality. They were in it with us. Every time we’ve shared a catastrophic failure with our backers, we’ve been blown away by the outpouring of compassion, respect, and understanding.
Nearly three years after our campaign, it’s been fantastic to hear from backers as they’ve fallen in love with their Keyboardio Model 01s. The experience has been more of a roller-coaster than we’d anticipated, but we can’t imagine doing it any other way. With thousands of keyboards in daily use by our customers, the work of fulfilling our first Kickstarter campaign is nearly complete. We’re just starting to think about our next product and our next campaign. We might take a couple weeks off before we get started, though.