Hard Facts: Challenges of a hardware startup

What not to do, and what not to do.


When you pick up a product off of a retail shelf, the beautiful packaging might lure you into the brand, or sell you on the great features that are going to make your life more convenient. But it doesn’t really tell you the story of the development process or about the people who built it, and the art of their craft. The development cycle for a hardware product is arduous, but it’s also the most rewarding.

Now that the Zuli Smartplug is shipping to customers (albeit a bit delayed), we wanted to reflect on our journey post-Kickstarter and give some behind-the-scenes insights into what we’ve been building the last 18 months. We didn’t do everything right the first time, but we definitely figured out a process. This post details some of the things that we wish someone had told us when we first started, so we hope some of the lessons we learned along the way can be helpful to other hardware entrepreneurs.

Phase I: Iterate

3 Months

Spend more time in this phase and you will spend less (time) in China. We started with the experience that we wanted people to have and then worked our way backwards to build the technology. We had already narrowed in on the basic feature set, the high-level architecture of the hardware, and the use cases for the product. A lot of this heavy lifting had been done before we launched our crowdfunding campaign.

Hardware prototyping

In the early prototyping and discovery phase, we were in the mind set to just build shit. We had 2 goals — build a looks-like model and a works-like model.

A looks-like model is great — even though it’s not functional, it helps to understand the color, material, finish, weight, and how the physical product feels in your hand. It’s also the easiest way to convey your vision to investors, potential customers, and other partners.

However, having a final appearance model should only be a guideline for your design process. It should never put shackles on your engineering team. Expect your model to grow, change shape, maybe even color and finish once you get to mass manufacturing. It’s easy to create a sexy 3D model in CAD, but you’ll quickly realize how many factors will limit you from replicating that in your final design of the product.

A works-like model helped us nail down the final functionality that the product would have. It enabled us to figure out what major chips and components we were going to use and helped jump start the firmware development.

As the electrical design evolves, the mechanical fit will inevitably require some rapid changes. 3D printing can be helpful to model these quick mechanical changes. Working with a company like Fictiv can help bring your costs down for high quality parts — delivered in less than 2 days.

Works-like model

Early on in the process, we hacked away at every aspect of the product, trying to improve the functionality every week — the goal was to build proof of concepts for mechanical pieces, electrical pieces, firmware, and the app. The prototyping at this stage was done independently for each of these pieces. It is really hard to iterate on the overall final product, so we segmented everything and worked in parallel — hoping to meet somewhere down the road.

3D printed prototypes

Usability

If you’re a product-oriented person, like us, it’s easy to just jump into the development process and start building things. But it’s also important to take a step back and understand why you’re building your product and not just what you’re building.

Before Kickstarter, we talked with a variety of people — those who used competitive products to those we thought might be potential customers. Additionally, through our Kickstarter backers, we already had an established community that was willing to learn more about our product — we used our crowdfunding backers to dig even deeper: What were their major pain points? How were they currently solving those problems? Which of those problems did our product solve for them?

Don’t just talk to friends and people you know. You need to talk with people who know nothing about you or your product. The best path is to allocate $500 to talk with 10 people (people will flock for a free lunch). Place ads on Craigslist and get a small group together for usability testing over a weekend. You will learn more about your own product in those 2 days than you ever will developing in a vacuum. Test early and test often.

Some other testing we did early on:

  1. We spent time in other people’s homes and offices to understand how they currently interacted with lighting in their homes.
  2. We used our works-like model to test basic functionality. Prototypes don’t have to look pretty.
  3. We tested major app interactions through high-fidelity functional prototypes. Flinto is a great tool.
  4. We conducted multiple in-person usability tests to understand the pain points in the set-up and on-boarding process of the prototypes we had. The tests were recorded using an overhead camera and the footage was reviewed by our entire product team later.

Phase II: Hone

12 Months

We now had a vision for what the product would look like, and the actual physical constraints for what the mechanical and electrical assembly would require. We just had to jam the massive works-like model into our tiny looks-like model. It’s very important to lock down the electrical and mechanical design at this point, so you can progress through the next stages of development.

At this stage spend some time carefully planning out your expected expenses. Manufacturing costs adds up quickly and if not managed properly, can be devastating for a young hardware startup. Don’t expect tooling and labor to be your only costs — plan for testing rigs, equipment, tools, components, computers and other miscellaneous manufacturing expenses.

Design for Manufacturing

We arguably spent the most amount of time figuring out our manufacturing processes. Everything we learned and accomplished in Phase I, was put to the test. This is where the rubber meets the road.

Unfortunately, most prototypes are not designed for mass manufacturing from the very beginning. If your goal with the prototyping phase was to work with customers and engineers to figure out the minimum set of experience and features; your goal now with manufacturing is to work with your contract manufacturer, vendors and suppliers to see what is actually, physically possible.

One of the tools for the plastic housing

Inevitably, some of the design decisions you made early on in the prototyping stage will not be feasible at a larger scale. As much as it might hurt your soul, you will have to make some compromises on the final look and finish of your product. Molding a plastic part is a science, but it’s also an imperfect process.

Tools, jigs, and components

Tooling

3D printed parts will only get you so far. The earlier you can start testing with real injected parts, the faster you will be able to achieve the final look and fit of your product. Plastic parts are injected using a ‘tool’ — a mold that is CNC milled out of steel. Depending on the complexity of the product, a tool usually takes 4–6 weeks to perfect and can cost upwards of $50k. Make sure these estimates are accounted for in your budget and timeline. There are no second chances with tooling.

Stick with a single-cavity tool. You might think that a multi-cavity tool is a wise investment — it doesn’t cost that much more than a single-cavity, and it will help ramp things up faster. Bad idea. A multi-cavity tool takes longer to build, it takes longer to make small adjustments (any change times 4), and you will have to deal with 4 different sets of plastic parts.

Components

Don’t be fooled by what’s available on Digikey and Mouser. Once you move into mass manufacturing, the lead-times for components start creeping up into the 16–20 weeks range. The most important resource for a startup is time, and a 4 month gap in procuring components is something that a young hardware company just cannot afford. It’s imperative to start the component procurement process early, which makes locking down the electrical design even more crucial. If done at the right time, the design can be adjusted to accommodate for supply chain issues. Find the cheapest parts, with the shortest lead-time, that are available through multiple suppliers.

If you’re reading this and thinking, I have no idea what I’m getting into, relax. There are experts that can help you get through this.

Working with a firm like Dragon Innovation can help find the right manufacturer, source components, set up the manufacturing line, help with certifications, and define your testing and QA criteria. If you don’t have expertise in house, this would be the no. 1 area to find help — we would not have been able to accomplish what we have without Dragon’s help.

Here are some other decisions that we made during manufacturing, that we would not recommend.

  1. Post-processing
    Our parts are painted, UV coated, and polished after injection molding. This introduces too many new variables into the final plastic piece, increases the per-part cost, and introduces new vendors into the equation. The paint job and color will vary across batches (especially if your product is white). It’s a royal pain in the ass.
  2. Complicated processes
    Our front panel is made using a two-shot injection method — a clear acrylic piece is molded over a white polycarbonate piece in a single machine cycle. Since there are 2 different materials, they cool differently. If you don’t pay close attention to your cooling method, the yield will be significantly lower. Try to stick to tried and tested methods that are less involved.
  3. Single sourced parts
    If one of your components is manufactured by only one young, obscure German manufacturer — probably not a great idea to use that in your design. Try to source parts that have functional alternatives, or work with trusted silicon manufacturers to ensure that the components are not going to be end-of-life soon.
  4. Custom parts
    All of the mechanical metal parts used in the Zuli Smartplug are custom. The metal prongs and heatsink are machined and the receptacles are formed out of brass and then nickel-plated. This gave us a lot of flexibility in our design process, but since most of these parts carry really high current, they had to be individually qualified for safety and reliability. It would have been much easier to use off the shelf parts with the proper safety listings.
  5. Trying to copy Apple, Nest or Jawbone
    Larger companies have the freedom to design features that are hard and/or expensive to manufacture. Avoid trying to emulate these big companies in terms of how they manufacture products and spend that energy trying to emulate how they think about bringing concrete value to the customer.

Certifications

Like we said before, test early and test often. Every hardware company will (or should) have their own internal benchmarks for safety and performance. But if you want to sell your product through any meaningful channels, obtaining safety certifications through a reputable third-party is very important. If you’re developing a wireless product, obtaining FCC/CE certifications is mandatory.

Certifications are tricky since only final manufactured parts can be qualified. If you have any missteps, you will essentially be set back 2–3 months in your development process. Preliminary testing at every step of the development process is the only way to guarantee a high probability of success. Although we had ‘passed’ all of the preliminary tests, we still had issues with the final product listing. We would recommend engaging a third-party test house at the prototyping stage, and continuing the process as you receive your very first manufactured parts from the manufacturing line.

Testing and QA

It’s never too early to plan for your testing and QA. This is your most important process in manufacturing, make sure it is thorough and controlled. While in development, always be thinking about how to test each part of your product (hardware, firmware, mechanical, software). If unprepared, these will cause either big delays, or poor production quality.

App design

Hardware companies don’t often talk about (and often don’t pay enough attention) to the development of their mobile apps. The app is often viewed as an appendage to the product. We consider the app experience to be the most important aspect of a successful app-enabled hardware product. The primary way for a consumer to interact with the product (and our brand) is through the app. In our case, the hardware is meant to essentially blend in with the home and disappear.

Designing the user experience

At the time, the app landscape for the connected home was filled with cluttered interfaces, and overly complex experiences. It was easy for us to figure out what we didn’t want, but it took a fair amount of prototyping to nail down what we did want.

We can’t stress enough the importance of whiteboards and printed flows/wireframes. This was integral in getting the team to brainstorm and solve problems together.

The entire flow mapped out for the Zuli app

Here are some things that we found to be vital in our design process:

  1. Keep a living interaction document
    As the app development progressed, we began to make changes to its structure and flow. The one thing that fell by the wayside while we made these changes, was the interaction documentation and it’s associated wireframes. As the team and product matures, the product will also evolve. Your living interaction document can be the north star that you can refer to when you get lost.
  2. Write down all decisions, and how you arrived there
    This might seem obvious to some, and a waste of time to others — creating a product specifications document will pay dividends in the long run, no matter how small of a team. Brainstorming sessions, especially at a hardware startup, involve many different stakeholders and disciplines — electrical, firmware, service, app, etc. Having a common specification to always refer to will help keep your team organized and focused.
  3. Prototype everything and test often
    When you are designing a completely custom UI experience for iOS, make sure you’ve thought through all of the screen transitions and animations completely. Anything custom in iOS takes lots of time and research to accomplish. It’s easier to iterate at the design stage than it is after your developer has already implemented a feature.

Phase III: Polish

Forever ongoing

How does an artist know when the painting is finished? This is arguably the hardest part of the process. It’s very important to draw a line in the sand that prohibits any further tweaking and prevents feature creep. For now, we will ship v1.0 of the Zuli Smartplug, but we view this as just the starting line. We have learned a lot along the way and cannot wait to use that knowledge to continue to improve the product experience and our skills.

There is no better feeling, nothing more satisfying than creating a physical embodiment of your ideas. Building a physical product and going through the development process of a hardware product is expensive, time consuming, and painfully challenging. We’re tired of hearing hardware is really hard. It’s no different than what any other startup has to go through — a monumental challenge, little or no cash, and a group of very passionate problem solvers.

Different animal, same beast.

This is the first of our series on Medium. We will continue sharing our thoughts on how we view building products, creating experiences, and growing a hardware company.

Show your support

Clapping shows how much you appreciated Zuli’s story.