4 Key Lessons I Learned After Working in Hardware Startups and Delivering 150,000+ Devices Worldwide
This post is about lessons I learned from nearly a decade of working in hardware startups, developing products in both consumer and B2B industries, and talking to entrepreneurs who succeeded or failed in the last few years.
Who is this post for?
This post is for hardware startups (i.e. relatively small teams with limited cash and meant to scale) working on developing and manufacturing their first product. There is no silver bullet in building great products, but I hope this post will help by bringing you new perspectives.
Here are 4 key lessons I learned working with hardware startups:
1. Keep a very high decision velocity
Having a high decision velocity is key for startups. It is the unfair advantage they have against big corporates which are slow due to their hyper-focus on the quality of their decisions.
However, for hardware startups, having a high decision velocity is even more critical.
The main reason is that there are countless decisions to make in hardware companies when building a product. And these decisions are not only about engineering. There are also decisions to make in manufacturing, supply-chain management, QA, logistics, distribution & fulfilment, e-commerce, etc.
So if you spend weeks trying to be 100% sure that you’re making a correct decision, you are doomed. On the contrary, if you keep a “getting things done” mindset without understanding why you do things in the first place, you’re doomed as well.
Here are a few ways to increase your decision velocity (note that, unlike speed, velocity involves having both speed and direction):
- Identify reversible and irreversible decisions: irreversible decisions such as ordering your injection molds are so costly that if you mess up, you will be in big trouble for many months (and maybe years). So be cautious with these decisions by validating all critical design assumptions. Be as quick as you can for the others.
- Always think in probabilities: think in probabilities about what could go wrong in the long term after taking a decision. Then plan for backups (ideally a backup of backup).
- Be aware of the speed-accuracy tradeoff: if you spend one day to be 70% sure that you’re making the right decision, should you really spend two weeks to be 90% accurate?
- Constantly assess your thinking: be aware of your biases before making a decision and consistently analyze your thoughts to ensure you’re being objective.
- Learn how to run crazy effective meetings: learn when a meeting is necessary, how to collect key information before a meeting, how to write kick-ass agendas, and make follow-ups, etc.
- Become a master of the 80/20 Pareto law: “20% of the work produces 80% of the outcomes” or “80% of your problems come from 20% of key issues”.
- Look forward, reason backwards: start reasoning backwards from a targeted outcome to ensure anything you do follows a logical path. And constantly ask “why” to ensure your task’s outcome has a direct impact on resolving your company’s biggest problem.
- Be biased towards action: being wrong is not as costly as you think if you know how to course-correct quickly. Slow decision-making can be much more expensive than you think.
Note: For low-volume products, you should be able to choose speed over accuracy easily because minor problems can be fixed on the go. However for high-volume products (consumer hardware), you need to be more cautious because you can’t do anything once your products are shipped.
2. Iterate as much as you can to build a product people really want
When you start working on a new product, you’re assuming this product will fix a meaningful problem for a group of people – and that these people will want to pay for it.
To confirm this assumption, you need to experiment. A well-designed product is always developed with an iterative approach (think about James Dyson and his 5,000 vacuum cleaner prototypes).
Surprisingly, most hardware startups fail because they don’t iterate enough. They think they have product-market fit and consequently scale too early. The sad truth is that the lack of customer demand is the 1st reason hardware startups fail (see research paper here).
From discussions I had with many hardware teams, I noticed that they don’t iterate enough because of two reasons:
- They don’t confront their proof-of-concept or prototype enough to their potential customers. As an engineer, I understand this excitement to solely focus on prototyping. But engaging your prototype to the market is a gold mine for your product. You will notice that your potential customers use your product in unintended ways or that your product’s key feature is not as useful as you thought.
- They have a misunderstanding of what is good customer feedback. Customer feedback such as “Great product!” is not valuable feedback because most people prefer to be kind and hide their true feelings about your product. Or, they can be genuinely impressed by your product, but the value proposition is still not good enough for them to make a purchase.
So iterate and experiment as much as you can until your potential customers really need your product to fix a specific pain point and are ready to pay for it.
Note: it is much harder for B2B hardware startups to receive product feedback from potential clients because they deal with established businesses that are overly slow at adopting new kinds of products. So B2B startups need to focus even more on de-risking their product and having a kick-ass value proposition.
3. Keep your product as simple as possible otherwise you won’t ship anything
A mistake that hardware founders make (that I painfully made myself) is to underestimate the cost of a product feature. Making this mistake can lead to destructive consequences – particularly for high-volume products.
Ideally, your first product should only include the simplest feature providing the highest value to your customers. Otherwise, you will likely end up with a product too challenging to produce at scale for a startup with limited cash. Later on you can still add other product features when developing your product’s next generation.
To illustrate my point, let’s take the example of a simple push button to operate your product.
It might look easy to add a push button since this is a common feature that most products have. So why bother experimenting if you really need it, right?
Here is what it might look like during mass production and after shipping your product:
Here is an example of a decision to add a simple press button for a consumer product:
First, to make this button, you will have to pay for an expensive injection mold (for the button itself) and complexify another mold (for the product casing to embed the button).
Then, your Bill Of Materials (BOM) will become more complex: the addition of one electronic switch, a connector that links the switch to your mainboard, perhaps a button gasket, a retaining ring, and a spring to keep the button in place inside your product.
You will also have to implement additional fixtures and jigs at your production line and train a factory worker to assemble this button. Sometimes this worker won’t do the job well, so your quality control team will find out that some products have a defective button, which will decrease your production yield.
But that’s not all.
After shipping your product, a few customers probably won’t use this button as expected or will find it difficult to use (since you didn’t do proper user research). Consequently, these few customers will break it and request returns, which will add more work to your customer support team and contract manufacturer.
Your team might want to remove this button and replace it with a simple touch sensor by that time. But it’s a huge endeavour: you would need to pay a very high cost for a mechanical and electronics redesign (new injection molds, new PCB layout, firmware, etc), update your packaging graphics, product manuals, and all your marketing material (website graphics, ads, etc).
This example might look a little too dramatic, but it is just to give an overview of the potential consequences a simple feature can have. That is why when iterating during your prototyping phase, you should only keep essential design features.
4. Most importantly: don’t underestimate the work to make your product ready for manufacturing
As mentioned in the previous paragraph, it is important to design your product as simple as possible. But it is far from enough: it also needs to be designed for manufacturing.
Sadly, most hardware entrepreneurs underestimate the work behind this critical design-for-manufacturing phase (called DFM). It’s common to see hardware companies not able to ramp up their manufacturing capacity even after receiving tons of funding.
I also underestimated this phase when starting to work in hardware. But my view on it radically changed after years of developing and shipping products in China.
Moreover, I had the chance to build strong relationships with Chinese manufacturers (through numerous feasts) who were thankfully willing to share terrible design stories from past clients and what makes a startup great to work with.
It was hard to choose what takeaways I should share because I have too many of them – but here are a few guiding principles to manage expectations regarding DFM:
- It is not a manufacturer’s responsibility to make your product manufacturable. You can’t just design a prototype, make it “easy” to assemble in-house, and then request a contract manufacturer to mass-produce it.
- DFM is a design process, so it’s the job of the product designers to work very closely with manufacturers to make the product ready for production. If you work with an OEM, the manufacturer will provide design directions according to what works for their production line that you will have to follow.
- It’s not because you partnered with Chinese manufacturers that your product will be cheap to manufacture. Your product needs to be designed to be cheap to manufacture. Hence, you can’t just rely on higher production volumes to lower your product costs, thus increasing your gross margins.
And a few quick points to consider when starting DFM:
- Unlike prototypes made in-house, manufactured products won’t be assembled by engineers but by factory workers who don’t know about engineering and who don’t care as much as you about your product.
- If your prototype used 3D printing to make custom plastic parts, you likely have a lot of redesign work to make these parts manufacturable because injection molding techniques have a lot of design constraints to consider.
- Your contract manufacturer will give you a limited time to produce your batches (and fix eventual defective units). And they won’t want to deal with a product with a high defective rate (for high-volume products you should target 1% defective rate at most). So you will have to design your product to be both quick and efficient to assemble.
To illustrate better how much work is needed to design a product ready for manufacturing, the following drawing can help:
What are the main takeaways from this post?
If I could give only 3 pieces of advice for any hardware entrepreneurs, it would just be:
- Talk to many hardware founders who successfully shipped products
- Talk to many hardware founders who failed at building hardware company
- Hire someone who is experienced in manufacturing and shipping products (and not only prototyping)