Advice for Early-Stage Hardware Startups

by Sam Altman2/10/2015

This is a guest post from Luke Iseman and Jeff Chang.

Hardware Is Easy

As we slog away at our soldering irons, it’s become de rigueur in the early-stage hardware startup world for us to wearily mutter: “Hardware is hard.” Our software-centric compatriots are understandably worried that we might weld their MacBooks shut if they mention their fifth multivariate test of the day.

But compared to any other time in history, hardware is easy.  Finding product-market fit remains as tricky as ever, but prototyping new physical things is faster and cheaper than ever before. Here are some guidelines we’ve picked up so far:

1.     Form A Posse.  Hardware is the Wild West: we’re just getting started in the quest to make smarter things faster, now that every material (and everybody) is at most a mouse click away.  Nobody yet knows the full potential of this exponentially growing community.  But you’ve got to take advantage: Join or start a Hackerspace, work from Techshop, contact makers who post interesting projects: find people working on hardware like yours and ask how they dealt with challenges you’re facing.  These conversations have led to me discovering faster and cheaper ways to make SMD stencils, casting aluminum parts from 3D prints, sourcing cheap components direct from China at in-country prices, and taught me everything I know about making things.

2.     Interview Your Industry.  After 5+ years of working on smarter gardens, I finally met with a giant gardening supply company. I learned more about what customers will actually buy (aka what I should build) in half a day than in half a decade. Figuring out what real people actually purchase at quantity is infinitely more useful than another coffee with another potential investor.

3.     Fast Over Fancy.  The speed at which new components are released, hardware devkits kickstarted, and novel 3D printer filaments formulated is gradually nearing the speed of software.  Just like with programming languages & frameworks, there will always be newer faster better tinkering toys, and you can spend forever researching them.  Instead, find things that work and use them to build a functioning prototype.  Nobody cares if you’re using an Intel Edison or a 555 to blink the LED in the prototype you show them: people care about whether you’ve made something that they want.

4.     Separate Prototypes.  I’ve used too many glue sticks and hours stuffing components into decent-looking half-functional prototypes.  Instead, make these 2 different objects.  Make an ugly black box ‘works like’ prototype, and create an entirely non-functional ‘looks like.’  Add a reasonable story for how you’ll get the black box shrunken down to fit in a manufacturable version of the sexy enclosure, and start selling.

5.     Finite Iteration.  Break down the elements of your separate prototypes further, and iterate on the most discrete units of functionality that you can.  Do a majority of the people you ask to play with your prototype tell you that the button is awkward?  Getting buttons with the right click-feel is one discrete element to iterate on, button placement is another.  Well-designed products don’t just pop from Steve Jobs’ brain into mass production: they’re iterated into existence through many rounds of fast experimentation on each element that matters.

6.     Selective Inattention.  Pre-selling an impossible product will get your project pulled and company sued, if anybody even bothers pledging towards it.  Having a million units ready to ship prior to telling anybody what you’ve built will make you bankrupt just as fast.  I think you’re ready to start to scale (meaning raising $ to fund production, from investors or crowdfunding) when many people who see your prototype want to buy one.  If you’ve got a data-driven story about why you should raise money or begin production at a different point in time, try and convince your team that it makes sense. If they’re sold on the idea, then go for it.  Just be ready to defend your variance.

7.     Achievably Exceptional.  I can make a reasonable argument about why I will be the first to make something new, maybe even spinning a sound story about why people will buy this new thing.  If I’m particularly lucky, I can find others to spend time/money on this vision.  This can add up to a believable pitch about why I’m going to be the exception, the startup that doesn’t go to 0 within a year.  However, it’s absurd to imbue myself with magical powers that defy the realities of global supply chains.  Look at the hardware kickstarters you’ve backed, and add a buffer.

Manufacturing guys smarter than me say it’s at least a year from locked-down, working prototype to delivery to customers at any significant scale.  Unless you’re doing under 1,000 units, you should have a really compelling argument for why you’re better at manufacturing than the 50+ kickstarters I’ve backed and waited too long to receive. If you’re making less than 1,000 units and they are not nuclear reactors, ask yourself if it’s worth your time.

Hardware is easy, and hardware is a complicated minefield of company-killing disasters.  This has led to most of your entrepreneurial competition staying in the purely digital world.  Meanwhile, many talented engineers had their desire to make physical things scared out of them by an education insisting everything they build be as reliable as a bridge.

It’s wide-open here in hardware startup country, full of opportunities to do things worth failing at.  The Nests and Teslas of the world are just starting to skim the surface of what we can make in our connected, sensor-filled, AI-enhanced meatspace.  Set aside your keyboard, and get to work remaking the real world as dramatically as we’ve reinvented the digital one.

– Luke Iseman (ex-Edyn, W14) makes smarter gardens and tiny homes

Hardware is Easy … Except For the Hard Stuff

Alright, you’ve got your production-ready prototype built, you’ve got some investor or Kickstarter money in the bank, and after talking with a few hundred customers, you’ve got a good idea of how to make ‘em really happy.  Now you’re looking to build, test & ship 1,000, 10,000 or 100,000 pieces of your product.

This is where Fitbit and many others almost died — for Fitbit, the 15 months of turmoil in the manufacturing & QA abyss, the lots of times they were “pretty close to being dead.”  But in the past seven years, these hardware startups have paved the way — they’ve made it much easier for the rest of us to scale hardware.

1.     Don’t do it from scratch.  Hundreds of resources, extensive manufacturing & fulfillment infrastructure, & consultancies already exist.  Don’t jump in with the first solution you find, no matter how easy they may make it seem.  If you don’t know enough to question them & keep on top of things at every turn, they’ll often take you for a ride or get you bogged down in an endless series of delays and bottlenecks.  Before you begin, talk with as many hardware startup founders, consultancies & manufacturers as possible.  After you start, talk with as many founders, consultancies & manufacturers as possible.  Learn enough to get a sense of the many things you don’t know and need to find out.

2.     Don’t forget the certifications — they each cost at least a couple thousand bucks, and some of them require bunches of pre-testing & carrier testing.  Make sure to get started on necessary certifications early, and skip the nice-to-haves.  Especially when cash is short & you need a shippable product yesterday, use easy-to-substitute off-the-shelf and pre-certified components wherever possible.  Extra certification costs don’t help you build a better product, and you need every dollar working to make your startup a success.

3.     Always think about bringing your capabilities in-house, if still outsourced — EE, ME, firmware, ID, apps, frontend & backend dev.  Search for, or wait for, the right person — put out your feelers (Angel.co, job postings, HN if you’re part of YC, friends of friends & friends of first employees), and whenever you can afford it, put candidates through trial projects and start hiring the best.

4.     Realize it’s not cheaper to hire people (unless you’re comparing with high-end design firm pricing), but things move a lot faster in-house, especially for fast prototyping & debugging.  Once you’ve been through the “our firmware designer’s in Europe, and build testing’s in Shenzhen” mess, you realize that a single email a day from each consultant can’t cut it — it’s delaying your progress by months.  And a successful company is always prototyping & debugging — once the first iteration’s out the door, you’ve got to start revising the next production run & prototyping future iterations.  While you’re head-down blinders-on busy preparing production runs & iterations, don’t forget to anticipate & create where the market’s headed – it can easily leave you behind, if you don’t imagine & build new awesome products when the time is right.

5.     Trial projects for candidates will help you understand your own business better too — just like with a rubber ducky, the act of explaining your hardware or firmware to someone will help you understand it better.  Not to mention, your candidate should hopefully know a whole host of things about their field that you don’t.  Versus a consultant (especially for firmware), where you often only know that it’s done once it’s done, and learn almost nothing about how it works.

6.     Hardware costs are very spiky — and with each revision, mistake, or market shift, those costs will rise.  This will be particularly distressing to those of us accustomed to software startups, since your burn rate may swing wildly month to month.

From talking with founders of many hardware startups, there’s no hard & fast rule on hardware vs. people costs.  You’ll spend more on hardware, if you’ve got a remarkably innovative and complex product, if your tooling goes south, if your first couple manufacturers screw you over, or if you decide to go the high-end design firm route.  You’ll spend much more on people, if you have the capital to bring your team in-house, if you share equipment for prototyping, if you need to spend several years iterating prior to launch, or if you find an efficient & responsive manufacturer right off the bat for production and QA.

7.     With every startup, the people are key. No matter how much your hardware costs, the hardware isn’t more important.  An awesome team can do wonders with a run-of-the-mill concept, while an average team will tend to pull failure from even the most brilliant idea.  Your hardware’s being designed and built, app bugs are being caught and fixed, marketing copy’s being created and fine-tuned — all by the people on your team.  No matter how complex your hardware is or how many patents you might have, a dozen other teams are probably building something similar.  Your team is what makes the difference.

Of course, don’t forget QA, distribution & fulfillment, custom packaging, the right kind of sales team for your product, and customer relationships.  Nice thing about being a part of YC – there’s a YC startup for each of these things.  If you’re a YC startup that provides services for other startups, you can rustle up dozens of interested new customers with an email.  Building a company that makes hardware creation better?  Consider applying to YC.

Finally, standard rules of building a company never expire.  If you want to become a good CEO or CTO, read awesome books & blogs, go implement & see which ideas work in your startup, read some more, learn from an executive coach, ask everyone you meet lots of questions, and listen more than you speak.  Not to get too Zen, but the mind can be either a full or empty cup, and full cups don’t do too well.  Only someone who always keeps a bit of the newbie mindset can continually adapt and build an incredible organization.

Always remember, you’re not building this organization, or even your own product – that’s up to the people you hire.  The lone inventor model almost never works.  Yes, you’re building a product, but the company is all about the people, not the hardware.  Eventually, you probably won’t even be deciding the vision — just hiring the best people in the world to hash out the details for you.  So really focus on perfecting your hiring processes, spend the time to create & live out your culture, do your best to keep everyone on the same page, and just keep the boat moving forward.  And don’t forget — the Pareto principle applies to everything.  So do what matters, and ignore the rest.

Our world is made of physical objects, stuff you can touch & manipulate.  Bytes are awesome, but many of the world’s biggest ideas need a tight weave of both hardware & software.  All of the largest Internet-based companies are now building hardware – this is where the next Google, Facebook or Amazon will arise.  How will your team help create our future?

  • Jeff Chang (Doblet, S14) is building a network of portable batteries for your phone

Author

  • Sam Altman

    Sam Altman is the CEO of OpenAI. He was the president of YC from 2014-2019. He studied computer science at Stanford, and while there, worked in the AI lab.