Startups exist to make money.
From the outside, it doesn’t always appear to be the case. We hear stories of insane perks, we watch TV shows about odd marketing decisions and we laugh at business models that don’t have a sniff of revenue now or in the future.
But head over to their investor page or look them up on AngelList, you can immediately see a public list of individuals and groups that guarantees money is somewhere in the equation. The investors took the risk to toss the company some cash and expect to be paid back in many multiples of that first amount.
In the software world, this payback time can stretch out over a longer time period. Assuming the perks and salaries are nominal, overhead costs can remain low enough that a small band of intrepid software engineers can take startup capital and subsist for months or years. Turning a prototype around and shopping it to investors can result in higher perceived valuations, which brings in more money and allows for an even longer timeline (runway). Nothing about this process is unknown or surprising, especially to readers who have frequented sites that enjoy covering VC activities and the funding of early stage companies.
There are different boundary conditions for hardware startups. The most glaring one is overhead. Whereas software companies pay for the office setups and salaries of employees, hardware companies need to pay the same office/salary costs AND pay for the cost of prototype and production hardware. This includes PCBs, components, assembly costs, packaging, testing fees, certifications and test equipment to qualify products. Prototypes shown to VCs and other money holding institutions may delight and amaze, but the perceived market value is often lower, based on simple logistics: Distribution does not come from the Android or iTunes marketplace but from FedEx and UPS (only after the units make it out of the factory).
The net result of higher overhead for hardware startups is a shorter amount of time until the money runs out. Controlling overhead means having fewer employees, people paid at lower than market rates (making it hard to retain employees) or trading in a higher percentage of the company in order to secure more capital at the beginning of the life of a hardware startup. These added weights, coupled with the aforementioned insistence on early investors making a return in 5ish years makes for one final tradeoff: complexity of product.
Can a hardware startup develop a complex product?
Over the past 5 years, there has been a sharp rise in hardware startups in the market. This trend is supported by plummeting hardware costs due to higher silicon integration on microcontrollers, the burgeoning ecosystem in tech centers like Shenzhen and the rise in smartphones. As a result, there has not only been an explosion in available product, but also the acquisition of some leading products. A short list from Cyril Ebersweiler shows a summary of recent aquistitions:
Looking at the “Unicorns” that Cyril writes about, they are far from simple products. However, these are consumer level products with a common thread among them: A hardware front end for collecting data. To use server parlance, these are akin to “thin clients” whereas the system design still requires a back end server. The end user and interface is what ultimately determines the complexity: in the case of something like the Nest, the designers decided to have the UI and networking contained on the device, instead of interacting through the cell phone. Also note that many of these companies are consumer facing companies. This allows for scale and profit, even at lower margin.
However, let’s zoom out from the chart above and realize that these companies represent the best case scenario. These products were commercially successful and were either purchased by an outside company or IPO’d for a bunch of money. What is at the lower end of the spectrum?
The effect of developing consumer products
Cost constraints are a difficult problem in any system. A designer working without any semblance of cost constraints is likely working on a project with false assumptions, on a project that has limited growth potential due to few, insanely well funded customers(ie. military) or a project that is doomed to failure. Every project eventually runs up against the realities of cost.
This affects the consumer level products most. Margins being what they are, a 4x cost adder to any final product price is reasonable. This means that a $100 MSRP means no more than a $25 BOM. Take off packaging/shipping materials, enclosures and assembly costs and a sub-$10 BOM is not unheard of. This begins to peak into the world of complexity.
Tradeoffs begin when budgets are thin; if you were going to have a chip that can measure a signal (ADC converter, for instance) but you instead can lower the resolution of your measurement and measure it with your existing microcontroller, you might just need to choose that option. The same goes for the range of radios you had planned and hoped to have on board–you might be constrained to a single Bluetooth chip to now get your signal where it needs to go. Now that your processor is busy reading signals and taking care of communication, perhaps you decide to bundle up some data and send it over the bluetooth link to a linked in smartphone. Maybe even that is too much and you bounce those packets directly to the cloud.
The point is that the consumer focus and the resulting costs of many hardware startups begins to affect the nature of the product itself. The formerly complex stereo system with the fancy class A amplifiers is now replaced with a range of bluetooth, battery power speakers using Class D amplifiers. These are placed around the house and now can only function through your smartphone.
Does it matter if the hardware is complex?
Given the above example, many of the tradeoffs don’t seem like tradeoffs. Why would you want an expensive, heavy, immobile amplifier when you can get a convenient set of Sonos speakers? Why would you want a self contained step-tracking product that requires a screen and a range of input buttons when you can sync up with your FitBit or Jawbone? Many of these products are entirely new categories, spawning new industries. They are thin client pieces of hardware connected to large software infrastructures that will allow interesting data harvesting and profitable add-on services.
It is the overall complexity, the synchronicity between hardware, firmware, software and data tracking that ultimately determines the complexity of the system. The best companies need to master each of these pieces; the ones that do realize large valuations and acquisition offers.
So what are the depths of a hardware product?
Now we (finally) reach the question: Given that there are commercially successful products, will there be complex and interesting hardware developed in the future?
The answer: only if the split between hardware, firmware, software and data requires that the complexity live in the hardware portion of the product. If the product could be a single resistor with an app that somehow measures that resistor (and there was a large market for the entire ecosystem), it would still be a commercially successful hardware startup. If the application requires that it be a connected device that cannot be connected through a smartphone (which normally lends the screen and the internet connectivity), the product will need other pieces in order to make it viable.
In future pieces, we will discuss the rise of robots (some of which were included in the chart above), which present hardware that are both complex, expensive and potentially game changing to the hardware ecosystem.