Pitfalls of early-stage deep-tech startups

What are deep-tech startups?

Very broadly, deep-tech startups are startups that build upon deep technical expertise of the founding teams to bring cutting-edge research to the market and creating loads of value at the same time. Examples of deep-tech startups are Blue River technology, Google Deepmind, etc.

Why are deep-tech startups attractive?

Deep-tech startups as a category is attractive for investors and founders for the following reasons:

1. Built-in defensibility. Unlike e-commerce/marketplace startup, very few people in the world can do the same things the deep-tech startups are doing.
2. Unbounded market. Unlike startups like Uber for dental supplies which limits the market to the number of dentists in the world, a technological innovation can be general enough to be applicable to almost everybody in the world. The market for a new piece of tech now might be small now but look at what happened in the PC revolution!
3. It’s the right thing to do. Too many interesting technologies are stuck in the research lab. Everyone should work hard to bring them out to serve the world.

Deep-tech talent such as PhD graduates and engineers with deep industry experience are great raw material for a successful startup. Startups are also a great way to extract value from their knowledge for the good of the society. This is also the model executed by startup accelerator Entrepreneur First with great results.

Pitfalls of early stage deep-tech startups

I’m focusing on some pitfalls of early stage startups deep-tech startups. Early stage here refers to startups within the first one to two years of their inception. These problems are by no means unique to deep tech startups but it’s worth considering how they manifest in deep-tech startups. The background for these problems is that founding teams of deep-tech startups are smart, versatile, and can build complex hard-tech systems. They can always sell something if they want. I hope it will be obvious by the end of this post why this can work to the disadvantage of a deep-tech startup if one is not careful.

1. Doing very similar things

People in deep-tech ends up doing very similar things. You can see this happening in hottest areas of the day such as deep learning, autonomous driving, blockchain. For example, the hottest thing in blockchain technology now is the Initial Coin Offering (ICO) and most ICOs have something to do with cloud storage. Tech investors know the hottest buzzwords. Traditional companies don’t want to be left behind. CS students flock to learn about them and they can easily learn about them from online lectures such as the Coursera machine learning and Udacity self-driving car course. People tend to look at the same hottest areas because they are quickly maturing and beginning to deliver value. In general, the blazing speed of information dessemination is a good thing for the society. Deep-tech founders can quickly follow these buzzwords because of their expertises. But startups have to be very careful and wary of the competition in these areas.

As a technology area becomes hot, the supply for such technologies and the people capable of developing them become abundant. For example, many interesting tech in deep learning and AI often come with good open source code and plenty of community support such that it is literally in the hands of anyone interested in them. With open-source code, any competent engineer can, for example, fine-tune convolution neural network for their own image dataset. This means that a startup trying to sell deep learning as it is, in its technology form (for e.g. Machine learning as a service) or applying to deep learning to obvious problems (for e.g., classify photos) is going to have a hard time because so many people are trying to do the same. (Engineers rather use open source ML library; Google, Apple, Facebook are already applying machine learning to photo albums). It is hard for a starup to create value by building something better when incumbent, other startups, and even the target customers, having access to same set of technology can also do the same.

Successful startups are built upon unique insights of a problem or technology. It is an insight that give startup an unfair advantage that almost feel like cheating the game. The key to develop such an unique insight is to look deeper at both problem and technology. Look for gaps in well accepted assumptions, look for exceptions to common rules. Look for problems that may arise with new technological trends. This is hard work with little results to show before developing the unique insight. EF has a nice way of thinking about ideation that helps uncover unique insights and turn them into breakthrough ideas.

2. Trying to sell too early

For early stage startups, the single most important task is to find a product-market fit. I strongly believe the same applies to deep-tech startups as well. However, there is a strong expectation that most of the technical work are already done in the research lab and the startup job is to just sell it. This might sound ridiculous but there are good reasons for people to push deep-tech startups to sell as soon as possible. Investors are afraid that deep-tech founders might commit the mortal sin of building technology that nobody wants. Many believe that PhD founders have a strong pre-deposition to to build complex things that nobody wants. (Contrary to popular belief, a good PhD education teaches grad students the art of picking the right problem). The surest way to reduce the risk of building something that nobody wants is to try to sell something as soon as possible. If something is sold, at least one customer wants it.

However, trying to sell without a clear understanding of the product’s value hypothesis is just as dangerous as building a complex product that nobody wants. A value hypothesis of a product is an understanding of why customers are likely to buy it. Without clear understanding of value hypothesis, early sales are hard to replicate, signals from market are hard to interpret, which further delay finding product market fit. By trying to sell early, founders may easily end up doing very similar things as many others (pitfall 1). This is a corollary of good product development and customer development takes time. It is not hard for deep-tech founders to build something and sell it especially if professional services are added as a deal sweetener for free (more on professional services later). Sometimes the only way to find out if a business model works is to try and execute it. But this activity becomes meaningless if founders are settling into easy and obvious business model, driving the sales numbers hard and nobody is thinking critically about what is happening.

Resist the temptation or pressure to sell something for the sake of selling. Spend time and focus on customer development.

3. Venture Capital subsidised professional service

It is very easy for deep-tech startups to confuse value of product offering and value of professional services. Professional services are work done for customers to implement or integrate the startup’s product. Professional services are very low margin business. To a certain extent, it is inevitable for startup selling to enterprises to also provide professional services as enterprise products tend to be complex. At an early stage, professional services help to strengthen partnership between startup and its customers. It is also a reliable way for startups to develop deeper understanding of the customers’ needs which drives product development. At later stage, professional services are commonly delegated to channel partners. It is very attempting for early-stage startups to charge lower than market rate for professional services in hope of the customer adopting its product. Early teams might not even realise they are doing professional services at all as they are just trying to make their product work for their customer.

But here’s the danger for early stage startup, if professional services is charged at very low rate or free, the customer might be buying the product only because the professional fees to use a competitor’s product or even open source product is more expensive than the price of the startup’s product. If this is the case, then the customer is using the product only because venture capital is subsidising the professional service, which is not sustainable.

Be extremely clear about the value hypothesis of a product. Actively validate the hypothesis and be honest and sensitive to market signals. Sometimes the easiest person to fool is yourself .There is no right or wrong answer to whether to offer discounted professional services or not. Startups founders need to be aware that doing so muddles up signals from the market.

Conclusion

The above are just some of my thoughts on pitfalls of deep-tech startups. I leave self-introduction to the end because no one would read “startup advice” from someone who has been in the startup ecosystem for less than a year. I joined Entrepenur First last year to start a deep-tech startup. It didn’t work out for me and I joined another startup from the same cohort as first employee. I’m writing this to serve as a record of what I’m currently thinking about. If you happen to be reading this, thank you for reading my first post!