Last week in Prague, the famous inventor, sci-fi author and speaker Ramez Naam gave a talk about one of our favorite topics: Exponential Innovation, under the auspices of DirectPeople. Ramez Naam is not only an award winning author, he is also a veteran of Microsoft, the holder of 19 separate patents, and the founder of Apex Nanotechnologies, which is the first company to create software tools for molecular design.

The partnership is fitting, as DirectPeople works with companies and startups to realize innovative new projects, building teams around new ideas and executing them for banks, energy companies, and telcos, as well as smaller firms. (Full disclosure: Petr Sidlo, CEO at DirectPeople, is our mentor, and Phillip Staehelin their Innovation Evangelist is our mentor and investor.)

The talk centered around an introduction to exponential thinking, much like those we have written before. What really stood out from the talk however, was Naam’s particular attention to the reasoning that so often fails to predict radical changes in the business and technology landscapes. As important as the exponential innovation framework is, we must put it in context with the logic that has failed to match it over and over again.

The talk was a long one, so I’ve chosen to focus on 3 key errors that perpetually lead industries as well as individuals to dramatically underestimate the effects of innovation over long periods. As Naam pointed out in his talk,

3 Ways we Fail at Predicting the Future:

First: Disruptive Innovation Comes “Bottom Up”

One of the key analogies Naam used was the broad history of world empires beginning around A.D. 1400-1500, with two key events: the European discovery of America by Columbus’s fleet, and the exploration of South Asia by Zheng He of China.

To scale: the flagship of the Chinese fleet of 1400 (larger), and the Santa Maria (smaller).

Though the two events bear remarkable similarity in their timing and boldness, there the similarities end. Columbus’s voyage was an accidental success. Attempting to circumnavigate the globe with too little food and resources, his fleet was saved by the accidental discovery of Cuba, a fraction of the total distance to his actual target, East Asia and India.

Zheng He, on the other hand, commanded the most technologically advanced navy in the world. His ships made Columbus’s tramp fleet of the Nina, Pinta, and Santa Maria look like bath toys in comparison. His expedition was well planned and financed, took much less risk, and made much more important near term discoveries for China in the form of large potential trading partners.

Yet, Columbus’s voyage set off an imperial age in Europe that lasted for nearly 5 centuries, and saw the colonization and Europeanization of most of the world. Zheng He’s 25 year exploration of South Asia and East Africa ended in the burning of his fleet, and the closing of China from the outside world for another 4 centuries.

Why these two vastly dissimilar outcomes? Moreover, why did the results of a well funded, professional, and government backed expedition (that of Zheng He), result in nothing of significance, when the poorly funded, badly planned expedition of Columbus ended in Spain’s domination of South America for hundreds of years, and Britain’s eventual domination of half the world?

Top Down V. Bottom Up:

There is one key differentiator: China’s exploration of the known world was a top-down affair, instigated by the emperor, and carried out on behalf of China’s leadership. It was essentially a government program/ Meanwhile, Columbus’s voyage was backed by Queen Isabella of Spain (after other monarchs turned it down), but she served as little more than a financier. The project was from the beginning a bottom-up effort by Columbus, and would have been doomed to failure and the death of the entire fleet, if not for America conveniently sitting between Europe and Asia. Columbus lucked into a discovery that eventually made him rich and famous. He took the key risk (with his life), and so he also owned his share of the rewards.

Again, Naam directs our attention to the effects that leadership structure had on the resulting discoveries. While Zheng He returned to China a hero, his discoveries and fleet were the property of the Emperor, who chose simply to dismantle them, rather than to take the risk of engaging with foreign trade and potentially coming into conflict with foreign powers. To China, the fleet represented a cost and risk that was evaluated based on the status quo. Since the voyage proved that China was the most powerful nation on Earth, the Chinese saw no more reason to engage with the outside world and risk their already safe position.

Stability Becomes a Liability

Columbus was in a weaker position from the beginning. So why did he win? Naam argues that it is because of the inherently competitive nature of European politics at the time. While China was unified under the Ming Dynasty for 3 centuries (from 1368-1644), Europe’s political landscape was inherently unstable, and composed of hundreds of individual state-like entities, with nobles fighting for position and consolidating control of their regions.

It was that instability which made it difficult for any one government to back Columbus, and it was the need for a competitive advantage that drove Isabella, who had only just united Spain under one ruler, to fund him. The price of not having state control of the expedition was that Columbus himself (along with the explorers and conquistadors who followed), personally profited from their efforts. A traditional of entrepreneurial exploration was established because Europe could not afford to publicly fund it.

Thus, Columbus’s voyages were highly disruptive to the established order because Europeans were desperate to gain competitive advantages over neighbors. Meanwhile China, the strongest nation in the world at the time, had no such need. In fact, quite the opposite, as opening itself to trade only served to destabilize China when it eventually did realize it could no longer afford to remain closed. It was finally Britain that forced the opening of China and established Hong Kong as its trading colony in the late 19th century.

One does not have to look too hard here to recognize the tech startup and the modern corporation. Stability for its own sake can turn to a liability over time, as hungry younger competitors adapt too quickly and are too numerous to be stopped.

Second: Exponential Innovation is Deceptive

Naam spent a lot of time covering the ways in which exponential innovations deceive people into seeing them as always just about to slow down. This is because the fundamental drivers of innovation are often mistaken as being the same as they were before the exponential growth curve took effect. He shared this graph, for example:

The graph shows a classic exponential curve in the electrical capacity of solar power from roughly 1998 to 2014, along with the World Energy Organization predictions of future growth. As is plain to see, every single prediction over 12 years shows static growth, but realigned from the new actual baseline in any given year. Every prediction is not only wrong, it’s wrong in exactly the same way.

Why would that continue to happen for 16+ years in a row? Surely someone at the WEO might notice it happening and try to figure out why?

Exponential Innovations are on a broad front:

Naam provides a partial answer in this blog post on the topic. He points out that predictions typically take a snapshot of the industry at the present moment, and extrapolate expected advances based on what appear to be limitations to growth that are non-surmountable without a step change in the technology or the underlying economy.

This works most of the time, because most of the time the specific conditions for exponential innovation aren’t met. Namely: that advances in different aspects of a technology produce feedback to affect growth in other areas. Innovation is happening on a “broad front.”

In this particular case, the reason the predictions are wrong is not because their underlying assumptions are inaccurate, but because they are about the wrong things.

For example, predictions of the slowed growth of solar energy may point to theoretical limitations in the materials being used to make solar panels. Indeed, solar panel efficiency is not growing as fast as it did in the past. Instead, as Naam points out in his post, the manufacturing methods are improving faster instead, which is compensating for the diminishing returns of material science advances.

This effect repeats itself over and over again. If it isn’t manufacturing, it’s software. If it isn’t software, it’s infrastructure changes. If it isn’t that, it’s changes in the consumer business model that accelerate growth. The growth in total wattage keeps going up exponentially because all these advances are driving each other forward at once.

Third: Exponential Innovation is Democratizing

On the last point about innovation around the economic models of new technologies, Ramez Naam took special care. He argues that exponentiality of innovation depends on the technology’s ability to be democratized over time.

In fact, in reference to solar power, he used Nuclear energy as a counter-example. Despite the enormous efficiency and effectiveness of modern nuclear power, it is failing to keep up with solar energy in the growth of its footprint. In fact it is declining. Predictions by Isaac Asimov and many others about a nuclear future in which atomic cores power watches and handheld computers, or even body implants, are now severely dated.

Why is that? Naam argues that it is because Nuclear power is difficult or impossible to democratize. Simply because of how it works, and its ability to be weaponized, atomic power must remain under the control of governments. This means that the core technologies involved with nuclear power have fewer chances to influence other areas of technology growth, and vice versa. Nuclear grows in a kind of parallel, where advances don’t greatly benefit other technology areas, and so other technologies don’t arise to create demand for new innovations.

Nuclear is not doomed because the technology doesn’t work, but rather because it can’t be enmeshed with other businesses and technologies. Like China’s great navy in 1400, nuclear is a state operation: it can’t be democratized because it is too dangerous.

If we go back to the previous example of Columbus and Zheng He, we see that a classical prediction would state that the Chinese technical superiority would allow them to predominate or to catch up quickly with the west. But what actually ended up happening was that the west innovated around the business model of exploration first, and the attractive monetary incentives led a revolution in technology that profited from the scalable finance model that colonization had created.

Consider the fate of nuclear vs. solar power in that context. Nuclear is far and away more complex technology. The only problem is that might not matter. The innovation that advances on a broad front and permeates society will win out.

By the time the British Empire came into direct conflict with China, it was the English who had developed the use of colonies as markets for their own goods, rather than a source of resources for the use of their own governments or ruling classes. Thus the English could afford conflict with China, because colonization was immediately profitable to those involved, rather than to established players only.

Indeed, this innovation is what made the American revolution the opposite of a disaster for England. Because England had repurposed itself as the workshop of the world, the growth of American power and economic strength helped the British empire to finance itself through trade with America. Britain had innovated away from the old definition of an Empire through mercantilism and trade. The state did not have to directly fund exploration and colonization because it shared the profits with the colonies themselves.

The English democratized economics itself. That was the innovation. It was one China still struggles to match.

So too with solar energy: solar power can afford to compete directly with traditional sources because it brings the benefits of the new technology closer to those who will profit from it. Even when solar was not at cost-parity with traditional fossil fuels (which is no longer true), it was cost effective for those who adopted it, because they saw the benefits personally, and over the long term. The risk of adoption had an appropriate reward.

Unlike with nuclear power or coal, which require long-term up-front investments far from view of those who pay for them, solar is a source of energy consumers and businesses can own themselves. This means they’re increasingly likely to invest in it, and to see less risk in doing so. The technology is thus democratized, and suddenly it can be the basis for knew, previously unimagined products or businesses.

The innovation can be in the business model this makes possible: not always in the technologies that underwrite the growth. Exponential Innovation shifts streams and goes around obstacles, rather than charting a straight line in one area. Thus predictions from the WEO may accurately represent technological limitations, but still fail to predict growth because they are using an outdated economic model.

Is There a Better Way to Predict the Future?

On the one hand, if predicting the future were as easy as a new mental framework, then everyone would be doing it already. Yet on the other, we can avoid making the same wrong predictions over and over by paying attention to what causes an exponential growth cycle in a particular industry or technology.

One of the key takeaways for me from Ramez Naam’s talk was that exponential change occurs primarily where there are multiple axes of advance being explored at once. If one hits a wall, another can cause the technology’s adoption to grow any way. What is inherent in an exponential growth model is that growth never depends on one particular characteristic of a technology, but rather on the overwhelming incentives it creates. People want it to grow, and so it does. That was the fundamental insight of Moore’s Law decades ago. Limitations we recognize as crucial today can become meaningless tomorrow, because the incentive structure for innovation is self-feeding.

Singularity University uses a framework called the 6 “D”s  (also used by Ramez Naam), namely: Digitized, Deceptive, Disruptive, Demonetized, Dematerialized, and Democratized. In broad strokes, they argue that an organization or industry becomes exponential when the following conditions occur:

1. Advances are based on an already exponentially scaleable platform (ie: computing, or energy generation)
2. Advances appear to be insignificant or even slow at first, although capabilities are doubling and redoubling on small scales.
3. Advances begin to outperform existing solutions at sub-scale (eg: Solar vs. Coal)
4. Cost falls as demand rises.
5. Advances become commoditized and ubiquitous.
6. The underlying technology becomes commonly achievable and can be applied anywhere.

Given these conditions, a technology may reach an exponential growth curve, and if we recognize a technology that begins to fall into this particular virtuous cycle, we must closely examine whether or not traditional predictive frameworks of the future remain useful.

StartupYard Batch 9 kicks off in a little over a month. Final selections are happening next week. Time flies! We can’t wait to see who will be joining us for Batch 9.

Every round, we like to share a visualization of the application process. While we haven’t made the final selection, we can share our initial data about the applicant pool. All data is anonymized, but you can get a nice sense of the trends we encounter with each round.

A full overview of the selected teams will be available at least a month after the start of the program (so no earlier than May).

As we did for Batch 6, and then again for Batch 7 and Batch 8 we find it very useful now to look back on the applications, and see what’s changed this time around. Where are people applying? What are startups working on? What are the hottest buzzwords? Previous experience has shown us that StartupYard applications can be revealing about the key trends to watch for in the next 6-18 months.

Here we’ll give you a visual trip through our applications for this round, with our analysis, and comparisons with previous years.

StartupYard Batch 9 : Who’s Applying?

StartupYard Batch 7 was the first time that StartupYard attempted two rounds of acceleration in one year. Batch 8 and 9 also come in close sequence, with applications closing on Batch 9 only 2 months after the end of Batch 8. The pattern in volume of applications has remained the same: we received almost exactly the same number as with Batch 8: 149 in total.

Below is a word cloud of the areas in which Startups identified their top skills and the markets they are working in. Note: we have eliminated the word “software,” due to its ubiquity in the application pool.

As we can see, Finance takes the top spot, along with Transportation, and Artificial Intelligence coming in second. We can hardly be surprised by this: between Bitcoin and Blockchain hype, Uber’s continued fascination for tech people, and the many emerging use cases for AI or IoT, these areas are naturally attracting fortune-seekers with new ideas.

In Batch 7 and 8, we noted that Marketing took the top spot in terms of which domains the startups were interested in. This was followed in the past by Data, and then Marketplace. 

Going even further back, startups from Batches 5 and 6 were still focused on Platforms, Mobile, and Apps, with UX and UI being important skills to mention. Today, these terms barely appear. It seems platforms with good UX and UI are now seen as default requirements for new products.

Marketing is Dead: Long Live Marketing

Interesting to us is the fall from grace of Marketing, which led the board for years. It may well be a reaction to the broader cultural shift toward suspicion of Big Data and Advertising that have dominated the tech conversion in recent years. Facebook and Google have particularly been the subject of much concern over privacy, the spread of propaganda, and the wellbeing of users.

The rising focus on AI along with Blockchain seems likely to be part of a shift in general consumer sentiment against the mass collection of private data, with its inherent loss of privacy and control over our experience of online services. At the same time though, these terms are increasingly being used as instruments of marketing: companies now talk about using Blockchain and AI to do things for which these technologies may or may not be huge game changers.

While data is still the foundation of AI, it may be that startup founders see what they’re now doing as beyond the data mining business, which is already quite crowded. Big Data may have simply become a fact of life. On the other hand, new privacy regulations in Europe, and a cultural backlash against big data collection may be pushing entrepreneurs to avoid the need for massive data collection.

Finance and Distributed De-Uberization

Finally, Finance has made a big break from the crowd, likely because of rapid shifts in the way consumers are treating finance products. PSD2, European open banking regulations that force financial institutions to allow customers to share their financial data across multiple platforms, was meant to drive more innovation around consumer finance products.

We see increased interest not only in personal finance, but in insurance management (both B2B and B2C), payments, and consumer/corporate investing.

In addition, the sudden rise of the ICO, and thousands of emerging technologies based on the Blockchain, have inspired startups to become “finance” oriented companies. Tokenization of services and networks opens the opportunity for financial products to become involved in more areas of the economy than before.

Whereas in years past, we saw many applications interested in the “uberization” of many services, today startup founders are seeking to “blockchain” everything, from in-person payments to insurance contracts, to day-labor.

This cannot be too surprising, given that the sharing economy Uber has represented for many years has really failed to deliver the benefits it has long promised. Indeed, Uber itself has still failed to become a sustainable business, and has suffered a severe backlash from consumers – once the name on the lips of every founder, Uber is no longer in the conversation.

Whereas “uberization” leverages a common platform like a mobile app to create a marketplace, and extracts rent from the different participants, blockchain startups forgo the platform in favor of a protocol that can be remixed with other services, and which is not centrally controlled.

This is doubtless appealing to startup founders, because it distances them from the deeply unpopular practices and cultures of companies like Uber. Uber still takes a large share of every transaction on their platform, but it is increasingly clear that their centralized model does not deliver significant benefits to the market (which has been largely covered up by Uber’s massive subsidies for drivers and riders).

This is not even to mention that Uber itself is not a sustainable business after nearly a decade in operation. It lost nearly $4 billion in 2017, leading Forbes to label its business model “fundamentally broken,” something many critics had been pointing out for years already.

Startup founders are realizing that services like Uber are certainly attractive to consumers, but they are also asking: “why do you need Uber the company to have Uber the marketplace?” Most of what Uber can deliver to customers and drivers can be done more fairly, more reliably, and more cheaply using a distributed service model.

Where are Startups Applying From?

Let’s start with a view on where applications came from for Batch 8, and compare that with Batch 9. Here they are in sequence:

Startup Applicants to Batch 8 – the Previous Batch

Applications to Batch 9 – Current Batch : Blue = low volume, Red = High Volume

I’ve changed the color scheme here slightly to emphasize the volume of applications from 2nd tier countries. As you can see, Czechia is still the leader, but it leads with only a third the number of applications as even last year. Batch 8 had about 45 applications from Czech companies, Batch 7 had even more. This year, there were only 16 Czech companies in the pool.

At the same time, 0ur geographic footprint widened considerably. This year, we attracted applicants from 44 countries: more than ever before. Batch 8 saw only around 32 countries. New countries included Sri Lanka, Gibraltar, Taiwan, and Lebanon- with one application each.

Which countries saw an uptick in applications? Surprisingly the biggest jump was from the UK, where 13 companies applied. That’s more than any previous year. Perhaps Brexit nerves or political tensions are pushing UK entrepreneurs to seek growth in Europe before it’s too late.

Detailed breakdown of application numbers for Batch 9

Anecdotally, we noticed that a significant portion of co-founders in UK based teams were in fact southern Europeans or other ethnicities. Are startup founders looking to abandon the UK and return to mainland Europe before the Brexit hammer drops?

Applications from Ukraine and India also rose this year. Ukraine’s jump was expected because Ukrainians have recently been granted the ability to visit Europe without need of a visa in advance of travel. This is good news for startups, as it decreases a huge barrier to Ukrainian companies seeking to do ongoing business in the EU. It may also be a matter of concern for Ukrainian startups, many of whom have personally confirmed to me during visits that more young people are leaving for the west than in the past.

Full Employment Has Huge Impact

We had believed in the past that the number of applications from Vysegrad countries would continue to increase at a faster pace. This has not happened over the past two cycles. Instead  Slovakia, Hungary and Poland contributed fewer applications in total than came from Ukraine alone.

There may be a key to explaining this in the employment statistics across the region. Slovakia’s rate of unemployment dropped like a rock in 2017, from almost 9% to less than 6% today. Anything under 5 is considered “full employment,” meaning that jobs are available and workers in demand across the income spectrum.

In fact, most of the Vysegrad group are suffering from a lack of workers for the first time since the fall of communism. Poland, Hungary, and Czechia are all below 5%, whereas the European average is about 8%. Wage inflation has increased accordingly, rising in Czechia to nearly the 2007 pre-crisis level of 6% per year.

Meanwhile, unemployment remains higher in Eastern Europe, with Ukraine close to 10%.

For the first time in many people’s lives in Vysegrad countries, governments are taking measures to improve workplace participation among young people, encouraging them to join the workforce earlier.

Because unemployment is very low, wage inflation and working conditions are improving quickly, possibly encouraging more young people to seek jobs in existing companies, rather than starting their own businesses. Increased career security may also give employed people less motivation to leave a comfortable salaried lifestyle and found a startup.

As living standards rise, the risks of starting a business are more obvious: loss of opportunity and income can be more acute when incomes are rising, as people tend to fear losing what they already have. In times of high unemployment, some will take more risks, having less to lose.

In Which Domains are Startups Working?

One of the key items every year is the breakdown of “domains,” or areas of tech innovation where the startups are working. This is often somewhat different from the market focus, as startups seek to apply new technologies to old problems, or bring existing technologies into new markets for the first time.

Batch 9 Domains excluding “other.”

The core domains we focus on represent more than half of all applications. Of these domains, Machine Learning remains the strong leader it was last year.

However, IoT and Blockchain have strongly edge out Analytics and Security, which have been dominant in previous rounds.

It’s important to note that the domains startups identify themselves as belonging to may be a function of fashion as much as the reality behind the words. Whereas in previous years, we would get many applicants who called themselves “security companies,” today those same companies might emphasize AI, or IoT as their domains, with security being downplayed.

In the previous round for example, Steel Mountain, which makes a security device that employs machine learning to hacker-proof connected homes and devices on Wifi, called itself a Security Company. Today, it might rather call itself a Machine Learning company, which does security, or an IoT device company that uses Machine Learning.

Often the product can be the same, but the emphasis in the application may change to fit the new trends. The security topic has somewhat cooled for startups, as blockchain has taken over many of the same applications, and as security companies become more common.

Batch 9 Domains Including “other”

As always, the largest single group of applicants represents none of our areas of focus. Some of these companies do in fact work in IoT, Blockchain, or AI, but they choose for whatever reason to emphasize their domains differently.

This has a lot to do again with how founders see themselves and their future in the industry they’ve targeted.

We sometimes ask applicants: “Do you want to be a technology company or an ‘X’ company?” X being insurance, music, entertainment, finance, or 101 other possible domains. The distinction at an early stage is not always completely clear, especially because founders are not always from technical backgrounds, and don’t always see themselves as primarily tech companies.

One conversation we had with an applicant working on Machine Learning revealed that he was focused on one particular industry, and would not be interested in shifting focus. Of course that’s a valid choice, but it’s also something you can’t always tell from the way an application is written. Sometimes applicants begin with one idea, and end up working on something entirely different, aside from the core technology involved.

What It All Means

This data set is too small to draw a definitive picture of the startup landscape in 2018. However, as we have been consistently monitoring trends in our application pool for the past 3 years, we can outline some clear trends.

The state of employment has had a clear impact on the volume and sources of applications. We also believe that the state of the economy has affected the types of companies applying to StartupYard. The ubiquitous “me-too” startups and the local clones of American products and services of 5 years ago are all but gone today. Perhaps many of these ideas have already been successfully executed in Europe, but the state of the economy has also changed the landscape considerably.

In 5 years, networks around high-tech startups have matured, and the communities around Central Europe have gained new heroes in the process (including many of our alumni). The first generation of online entrepreneurs, and successful online businesses, are now engaging very actively with the younger generation of startups, as we have observed within our own mentor community. Banks, Telcos, Insurance companies, Manufacturers, and Tech giants feel the increasing need to engage with early stage companies, and younger tech entrepreneurs now have a menu of options for growing a business with corporate partners, early stage investors, and active Business Angels.

It is a good time for startups.