by Calvin Chu, Managing Director at R/GA Accelerator, Powered by Techstars
You’ve all heard the old cliché. Hardware is hard. Yes, that’s true. Few challenges worth doing in life are easy.
Internet-of-Things (IoT) is the next hot frontier, so it’s worth a bit of introspection. IoT is the intersection of hardware, data science and cloud architecture, and products typically sport neat human interfaces. Most ordinary startup entrepreneurs pick one of the above, but if you’re the masochist — or you want to be the hero of the startup world — pick all of the above.
At the end of the day, IoT is not hard for any of the obvious reasons. Internet-of-Things is hard because of individual people. What does that mean? It’s about integrating humanity, emotion, beauty and desirability into the end product. We must create devices that have real personal meaning on an individual level, while at the same time divulging information recovered from a population of users.
Let’s take a look at the IoT startup minefield. There are three common pitfalls that trip up IoT startups: critical mass, tool making, and when to roll in the data science. Messing up the first two might propel a product straight into the closet after only a few months (or less) of use and mismanaging the latter can waste a lot of time and money that you can’t afford to burn.
For most IoT startups, critical mass is one of the fundamental aspects of the product. In other words, significant and differentiating functionality arises from having a huge install base. This is all well and good for the Apples and Sonys of the world with the wherewithal to achieve critical mass quickly, but for IoT startups, it is a harrying exercise in timing. Suppose your product is a special automobile tire stem cap that can map the world’s roads down to the pothole level — having only a thousand installed will yield a pretty poor result. The real question is if only one unit of product can be sold, why would that customer buy it and use it every day and integrate your product into their daily life? If a network effect can’t be depended on initially, is your product desirable enough to stand on its own merits?
For those IoT startups that have a maker pedigree, the usual approach to productization is to first build a perfected tool. It’s wonderful to build great tools, but tools end up in the closet most of the time, waiting for the day they’re needed. If this is the intention, best of luck to you.
But if the hope is to build a product that’s a significant part of the 29 billion devices expected to be shipped by the end of this decade, it’s time to think desirability. Consider why and how somebody would use this product every day. To succeed in IoT, it’s better to bake in emotional desirability into a product than to build one that is simply necessary and does its job. It’s why Fitbit is now a fashion company, and it’s why Nest is worth billions.
By and large, IoT startups are — and should be — worried about the first couple of thousand customers. Sometimes this worry can be temporarily bridged using a traction tool like Indiegogo or Kickstarter. But mostly, IoT entrepreneurs are worried about how best to get to the first 100, their next 200, and the next 500 customers after that until they reach 2000 to 5000 install base. If minimum order quantities need to be met, there is little latitude to alter the hardware significantly if the first 100 people hated it.
If you had to prioritize one thing, it should be building standalone superiority: a look, a feel and a brand that emotionally captivates at a daily level. All the special population and data insight features in the world might take a back seat at first. This leads to the next big challenge for IoT startups: the data science catch-22.
Everyone loves infographics, and the very best of these are an epiphany to read. The top IoT products will offer similarly amazing insight about the individual customer. Unfortunately, this level of insight is astonishingly hard, and the fact is there aren’t enough data scientists to go around in the startup world. And for an IoT startup that hasn’t achieved a critical mass of deployed product, there may not be enough data present in the company to afford or entice bringing a data scientist on staff. As Steve Blank defines it, a startup is an organization formed to search for a repeatable and scalable business model. Given that most startups have no real forecasting visibility even a few months ahead, it’s almost an impossible luxury to construct information architecture for a moment that could be several years ahead into the future. It could literally be money wasted.
Whether you’re building into the connected home ecosystem, or designing wearables, or whatever — any part of the IoT ecosystem that instruments sensors into a customer’s life has to start with individual personal meaning. Maybe this comes as a result of delivering insight, or designing beauty and fashion into the industrial design, or defining the product in terms of social signaling. Devices like Google Goggles, Nike Fuelband, and others are forging paths. Still, the ultimate value comes when the device can provide constant insight that the user could not have known. If the device continually teaches me valuable facts about me that I never knew, then this is one of the holy grails of IoT. This could be information derived from constant sensing and deep data analytics of not just individual users but entire installed populations. But what of my recommendation to backseat these capabilities versus standalone functionality? What’s the best way to balance this catch-22?
Remember this: Your product is a layer cake of hardware, cloud, data analytics, and UX/UI. And the different layers have different degrees of malleability. Devices and sensors have to be perfected as if it’s NASA shooting a probe into space — once it’s launched it ain’t coming back. Perfect those features intended to establish product superiority on an individual level, but always have a raw data stream into the cloud so you can be nimble and actuate your product deeper into IoT space. Capture that data, store it, MVP your data architecture and refactor it constantly as the population grows. Stage it out. Bring on the data scientists later, but use your best guesses initially. You can’t easily MVP the hardware, but be agile in all the other layers. In short, initially do the minimum necessary to grow the network effect features, but make your device as human as possible — whether you end up with a large population or not — and then scale those other features as the install base grows.
Calvin Chu is the Managing Director of the first hardware connected devices accelerator in New York: R/GA Accelerator Powered by Techstars. Read his blog at A Gray Twilight or follow him on Twitter @cchu.