Designing for mass manufacturing is essentially impossible to learn without actually building several products. The challenges of negotiating with suppliers, verifying components, orchestrating supply chain logistics, and maintaining quality standards will swiftly kill a cash-constrained startup. But there's good news: unless you're on the cutting edge of a specific technology (like Apple with displays or metalwork), the problems you encounter at scale have very likely been solved by others. So the task then becomes to find those individuals or companies and hire them quickly.
PCH International is probably the most well-known of these firms. They notably make Kindle, products for Jawbone, and all the accessories for those dudes in Cupertino. Because they often work with big players, they're able to leverage large contract manufactures' (CMs) quality standards and economies of scale for smaller production runs. Currently PCH is also aggressively going after hardware startups and recently launched PCH Accelerator, which is essentially a strategic investment group to fund partnerships with startups.
Another option is Dragon Innovation, who works with Pebble, Makerbot, Sifteo, Leap Motion, and others. They're based in Boston and have recently partnered with the brand-new hardware accelerator/investor BOLT. They're certainly less flashy than PCH, but still know how to deliver.
HAXLR8R (Shenzhen) and Lemnos Labs (San Francisco) are other startup "accelerators"-- kind of like a hardware-flavor of YC. They seem to be more focused on the R&D stages (pre-Kickstarter) than production. For some personal anecdotes by others, see What startup incubators exist for hardware companies? [link]
One way to think about contract manufacturing in China is with the "smile curve" popularized by Acer in the early 90s. It's essentially a graph of a product's extractable value as a function from creation to sale. On the left is brand, high-level design (IxD, ID, PD) and any specialized R&D. In the middle you have component sourcing, supply chain management, manufacturing, assembly, and shipping. And then on the right you have distribution, retail sales, marketing, and partnerships.
... it goes on with more examples. Full post: http://www.quora.com/Michael-Grinich/Posts/Product-manufactu...
Funnily enough, Liam Casey actually often calls PCH the "APIs for manufacturing."
I'm sure that is true. Possibly because the people with experience of this come from a different community (established consumer electronics) than some of the new people coming onto hardware via the startup scene.
Bunnie Huang recently wrote a series of detailed blog posts about how to take prototypes to manufacturing in China. There is a lot of hard-won knowledge in them.
Part 1: The Quotation (or, How to Make a BOM)
Part 2: On Design for Manufacturing
Part 3: Industrial Design for Startups
Part 4: Picking (and Maintaining) a Partner
I both agree and disagree. For cutting edge products and extremely large volumes, yes you need experience. For a very complicated product, I again agree. But for the majority of what we're seeing with "Kickstarted, low to mid-volume production"? Disagree. If you have a mechanical or manufacturing background you can absolutely get a new product launched without paying Dragon or PCH exorbitant amounts of money. Will there be hiccups? Yes, but nothing catastrophic if you follow good engineering principles.
Design, validation and simulation can all be carried out in advanced CAD packages (or by not-advanced CAD + lots of labor).
Depending on your timeline, you can prototype locally for higher costs but faster iteration (design, build, test), or you can prototype with suppliers more slowly but save on production verification and testing. The speed benefits are variable too: I've had US proto houses come back with ridiculous lead times and Chinese factories turn around a production verification sub-assembly in less than a week. YMMV.
I love the smile curve; I've never seen it before, but that is exactly how things work. But the situation is changing as more and more Chinese factories of all sizes integrate engineering teams and work to become OEM/ODMs for the local market. The other changes are in the labor rate: as China gets more expensive that curve is going to flatten.
For small scale stuff in the 100's to 10's of thousands I think that you can go it alone for the right product class, which happens to be most of what we see here on HN/Kickstarter. Find factories on Alibaba, spend a minimal amount of money testing them out, place small orders to protect yourself and voila! I think the hardest part is dealing with demand prediction/logistics, but that's inherent to any hardware enterprise that doesn't have its own manufacturing capabilities.
But playing devils advocate, it may be tough for startups to find the resources to have a hybrid hardware/software company from the beginning. It just seems like a daunting idea to start a hardware company. BUT, with crowdfunding having become so popular, startups may be able to raise enough money through initial sales from their hardware that they are able to put that money towards creating unique software.
Sorry for all the back and forth, just was thinking/writing out loud!
For example, I design and build radio systems in my day job. You'd think an 800MHz 100W radio transceiver would be "hardware". In actual fact, even that sort of system is largely driven by software. Most of the signal processing is done on a DSP. The "hardware" is essentially a DAC/ADC, a mixer, a gain element and a filter, and even then the control is handled by a computer. For VHF frequencies and below, even the mixer is disappearing.
These days, nearly every electronic device fits the following model:
Transducer > sampling > SOFTWARE > reconstruction > transducer.
An exception with respect to some communication systems would be ones operating at extremely high frequencies, where it would be impractical to process signals in software due to memory constraints (higher operating frequencies -> more samples per second). In these cases, the mixing/modulation functions get implemented in hardware.
I wouldn't say "complement hardware with software," (although in many cases, it is a good idea - but let's be honest software, like hardware, can be copied and sold at a lower margin) - but instead "complement your hardware with a thorough understanding of where you want to be in your market." You may find that there's less money trying to capture and keep customers who want the cheapest product than to find a good spot somewhere in the middle or top-end of a market.
I see a lot of poor thinking in some products, where cost is nearly the primary focus - where the hardware product is something to be thrown away when the next model comes out, because it is stripped to its barest essentials. That's a difficult position to hold - it's much better to think about where your product will need to be in a few years, and make sure that it can grow (add-ons, upgrades, etc.) without needing to be replaced. Not much different than software in that sense.
I'm not sure how one starts a hardware company these days without being a software developer (there's usually more firmware work than electronics...) - but I see it happen all of the time, people don't really care how good the firmware/software is, as long as it works, it would seem.
I can see it now where start up founders with successful exits from todays start ups who grew up assembling machines and writing software will want to build something that could improve the lives of many people outside of open gaming consoles (which is still pretty cool).
One thing that I think still needs to come a long way is the amount of shared knowledge on hardware design. My co-founder and I both went to school for Mechatronics Engineering, and one of the biggest obstacles we have run into so far is that there just isn't as much shared knowledge on the internet when we run into problems - unlike software.
> My co-founder and I both went to school for Mechatronics Engineering, and one of
> the biggest obstacles we have run into so far is that there just isn't as much
> shared knowledge on the internet when we run into problems - unlike software.
Well, at least there's cnczone. And mechatronics journals. There are a bunch of online marketplaces (including alibaba, mfg, or protomold or something), but that's not the same thing as institutional knowledge..
The problem we have is IP... IP management is such a burden for a small start-up. I'm working with a University technology office. While they are great people (really), it still adds a layer of complexity to an already tough situation (IMO, start-ups do not benefit from having more people at the table). Add in lawyers, and it becomes an order of magnitude more expensive than your typical software start-up. And the IP you get from a University is so raw... the cost to develop it into a consumer ready product can be immense.
On the other hand, it has been an education.
From the Dragon Innovation home page : "As a matter of corporate policy, Dragon Innovation will not accept any job that involves transferring existing in-house manufacturing from the USA to China."
"There is no AWS-equivalent for hardware."
I usually don't like to comment on rumors. I heard that some people from Amazon are already working on a AWS-equivalent for hardware too based on their experiences from manufacturing the Kindle. If you know anyone from Lab126, ask them about it.
But accelerators are generally reserved for people with little to no experience in the craft, bare-bones resources and simply just ideas... They help a lot and bring great resources (connections, labs, some get little funding) but I've personally never been a big accelerator supporter... Except YC, which seems to have a 90%+ funding rate once you pass... But this is a subjective position here.
It would be nice to have more transparency during the process.
Developing, prototyping, and manufacturing hardware (and generally dealing with China) can be a painful process even if you have great contacts and know what you're doing. Companies like PCH International alleviate quite a few of these problems when dealing with hardware and maybe we need to see more of these sprout up. The middleman is huge in the China issue.
The 3D printing "revolution" is still in its infancy, it's tough to print high-res models for prototyping certain items and you end up deferring to styrofoam and other methods of prototyping before you get to the tooling stage. But it will improve.
It's so refreshing to read articles like this and to see a hardware renaissance in the Valley and throughout the industry in general. I'm getting so sick and tired of "Hey, check out my new app, it's like Instagram and it's social" Enough of this crap.
People need to make things, it's tough, but as someone who's a few months into a hardware start up, I'll say this: it's a great feeling to simply hold something in your hands. Let's make stuff!
"so you guys are like real engineers" [in reference to us making hardware]
"Oh, so you're not just another mobile app"
I love reading about hardware startups - I'm really excited about the next generation of things to come!
That may be true, it may not. It's a topic that's actually the subject of intense debate right his minute. Many people contend that the US is in the midst of a "manufacturing renaissance" as manufacturing moves back here from China. Not everybody agrees that it's a meaningful trend, but it's certainly not a given that US manufacturing is dead.
??? What does this mean? The US is making more stuff today that it ever has at any time in its past. We may not make a lot of super-cheap consumer gadgets, but there's more to the manufacturing world than disposable crap.
And if you use arduino , how do you manage the need to release object files (according to the arduino license) ?
> There is no AWS-equivalent for hardware.
Well, yes and no. Of course, it doesn't make sense to talk about AWS for manufacturing. Making physical items means that your supply chain, manufacturing process and channel of distribution have different time constants when compared to pure software-and-servers businesses.
By this I mean that step changes in supply, production or demand happen at far slower rated than in pure software products.
If by "AWS" we mean a manufacturing system that allows the entrepreneur to turn on and off chunks of manufacturing capacity with some immediacy and granularity, no that does not exist. And, that is so because it is just-about impossible for most consumer products.
The supply chain for most consumer goods is such that it has to start being stuffed with components and assemblies weeks in advance of manufacturing. Some OEM components have lead times measured in months. Others have MOQ's (Minimum Order Quantities) that can force commitment levels both to product volume and manufacturing effort/timelines.
As an example, I am currently working on a project --to be featured on Kickstarter-- that requires a custom aluminum extrusion in order to be able to manufacture it economically. Aside from tooling and setup NRE's most extruders have minimum buys that can range up to 2,000lbs of aluminum. In other words, once you play that game you are all in and committed to manufacturing at a certain level. Once these processes start you simply can't logon and stop them from your browser (or spin-up new ones) as you might do with AWS.
> Proper planning is essential because mistakes can be unrecoverable. For example, you might create a design that fails environmental tests but only discover this years later when you are about to go to market.
Right. And this is why elsewhere in the article he mentions that a lot of hardware entrepreneurs are older and more experienced engineers. When it comes to physical product manufacturing, experience can be invaluable. There's a lot that simply isn't learned in school. You can't recompile your way out of a design error.
> The build-test-iterate model that is popular in software startups doesn’t translate well to hardware startups.
Well, it does in certain domains. For example, every FPGA-based design I've done goes through extensive simulator testing before actually being committed to real hardware. The same is true of such things as high speed signal and power integrity (using field solvers and other tools).
When it comes to mechanical and thermal design a lot can be done with modern CAD systems through simulation. As an example, we devoted no less than six months to the design and simulation of an advanced liquid/forced air thermal management system for a 1,000W (yes, one thousand watt) LED array. The simulations were good enough that we could go to metal knowing we were going to be reasonably close.
The same is true with mechanical and plastics simulation.
Now, if what we mean by built-test-iterate means the inclusion of the actual end-user in a real market situation I have to say, yes, it is much harder. This is where prototypes are designed to be easy to change and adjust in order to experiment with form, fit and function.
In the end hardware startups are very different animals. I am not sure they can be compared through the same metrics.
Large warehouses full of 3D printers where you can scale up/down your production would be a fairly reasonable analog to AWS and what it provides.
3D printing is neat but it is very slow, expensive, limited in materials and costly. Surely this will change with time. Still, it will be hard to compete with optimized processes where, as I said, you can produce parts faster than clapping your hands.
Now, for micro markets where you need just a few of something? Yup, that probably works today.