Three challenges facing blockchain technology

Nearly five years ago, became the first major retailer to accept bitcoin as a form of payment. It now accepts many top cryptocurrencies. As a member of the senior executive team and board of directors at, I had a front-row seat to those decisions.

It didn’t take long for the Overstock team to realize that bitcoin’s underlying blockchain technology held great promise beyond cryptocurrencies. We also knew that for blockchain technology to reach its full potential, the startup companies advancing its use would need both financial and human capital support.

Overstock set up a venture capital blockchain incubator, Medici Ventures, to do just that.

We believe blockchain technology will eventually impact many industries. We are already involved in promising developments in areas like capital markets, money transmission and banking, voting, supply chain, property and self-sovereign identity. But there is still a long way to go before blockchain technology can realize its true potential.

Here are the three most important challenges facing more widespread adoption of blockchain technology right now.

Finding good enterprise-level blockchain software developers

The world has become so reliant on computers, to the point where virtually every company now has need for software development. In this environment, where demand grows exponentially, good software development talent is hard to find. Game-changing talent is rarer still.

Because blockchain is a new field of technology, there are fewer talented enterprise-level software developers who understand it well. Those who do can practically write their own tickets. While this is an enviable position for them, it limits many companies from developing engaging and transformative blockchain-based applications.

Let’s remember that we are in the early days of blockchain.

At Medici Ventures, we provide regular internal training to help our software developers climb this important learning curve. In this training — which we do in educational presentations which sometimes include accelerated coursework — our teams often present discoveries made when developing on one project, with the hope that the solutions may benefit those working on other projects. This approach lets us cross-pollinate our industries and our disciplines, so creative development and innovation become rising tides rather than isolated spikes.

The time spent learning is well worth it; it is why many of our portfolio companies rely not just on our venture capital, but also our human capital. Until there is a regular pipeline of well-qualified blockchain developers, the shortage of great talent will continue to be a struggle for the advancement of the technology.

Avoiding the temptation of regulation

Like many of their voting constituents, Congress and state legislatures are just becoming aware of blockchain. In some ways, this is good news: Political engagement will increase awareness and interest for utilizing blockchain technology and help drive adoption of these new ideas. Unfortunately, it also brings the temptation of regulation to an emerging market.

I get concerned when regulators and legislators get a whiff of any kind of technological development because they are tempted to regulate it. When U.S. Securities and Exchange Commission (SEC) chair Jay Clayton stated that he considered all initial coin offerings (ICOs) to be securities rather than commodities, and therefore subject to his organization’s regulation, Clayton brought an ICO boom to a screeching halt. While Chairman Clayton and others at the SEC have subsequently modified that stance, this regulatory tendency to fear what is new is dangerous.

The interconnectedness of the world means its adoption will probably take root and bloom quickly.

Technology — and the advancement of blockchain — should not be regulated. In the 1990s, when the internet’s potential was becoming evident, legislators opted not to regulate it. That bipartisan decision led to the open-market creation of the much-lauded “information superhighway” and the power of the internet today.

Certainly, there will be use cases that may require regulation as blockchain applications develop and proliferate. But the growth of blockchain technology will be best nurtured when it is free and unfettered from regulation.

Reaching critical mass

Cryptocurrencies and digital wallets built on blockchain are great uses of the technology. In order for cryptocurrencies to proliferate in use and stabilize in price, and for digital wallets to get widespread adoption, consumers need to spend cryptocurrencies more and merchants need to accept them. A great example of this working the right way is Colu, an exciting new company I recently saw in action when I was in Tel Aviv, Israel. Colu is a digital wallet that uses blockchain technology to create local currencies. People simply download the app, add money and shop locally. The app highlights local establishments and makes shopping convenient. And it is dazzling people in Tel Aviv!

The same can be said of other blockchain-based applications like secure remote digital voting. West Virginia recently became the first state to allow overseas citizens to vote remotely using a blockchain-driven app. The West Virginia program was tested in the May primary and was used in this November’s general election.

We’ll know blockchain technology has become mainstream when we are no longer talking about it.

Some critics have been quick to disparage real efforts to create digital voting with strictly theoretical worries. In reality, the rollout in West Virginia is a very focused solution to a specific issue: low overseas voter participation. The current system is broken. A blockchain-driven digital voting app is a clear solution. Anyone but critics of progress should eagerly support West Virginia’s efforts until there is an actual reason to worry.

Once any blockchain application is embraced in sufficient numbers by both the using and accepting sides, the impressive software will become an invaluable and ubiquitous tool. More widespread adoption of blockchain’s most beneficial use cases will trigger network effects that will multiply the benefits.

Let’s remember that we are in the early days of blockchain. Many industry observers seem to be in a rush to declare blockchain a mainstream technology. As enthusiastic as I am in my support of blockchain, I would not yet call it mainstream. The interconnectedness of the world means its adoption will probably take root and bloom quickly. We’ll know blockchain technology has become mainstream when we are no longer talking about it, but we are simply using it in everyday ways.

I am thrilled to see digital purchases made and remote votes cast in elections with this game-changing technology. As developers, investors and companies continue to focus on using and advancing blockchain, we will see that finding good enterprise-level blockchain software developers, letting blockchain grow free from unnecessary regulation and achieving critical mass use are the next important steps in the growth and adoption of this world-changing technology.

Three ways to avoid bias in machine learning

At this moment in history it’s impossible not to see the problems that arise from human bias. Now magnify that by compute and you start to get a sense for just how dangerous human bias via machine learning can be. The damage can be twofold:

  • Influence. If the AI said so it must be true… people trust outputs of AI, so if human bias is missed in the training it could compound the problem by infecting more people;
  • Automation. Sometimes AI models are plugged into a programmatic function, which could lead to the automation of bias. 

But there is potentially a silver machine-learned lining. Because AI can help expose truth inside messy data sets, it’s possible for algorithms to help us better understand bias we haven’t already isolated, and spot ethically questionable ripples in human data so we can check ourselves. Exposing human data to algorithms exposes bias, and if we are considering the outputs rationally, we can use machine learning’s aptitude for spotting anomalies.

But the machines can’t do it on their own. Even unsupervised learning is semi-supervised, as it requires data scientists to choose the training data that goes into the models. If a human is the chooser, bias can be present. How the heck do we tackle such a bias beast? We will attempt to pick it apart.

The landscape of ethical concerns with AI

Bad examples abound. Consider the finding from Carnegie Mellon that showed that women were shown significantly fewer online ads for high-paying jobs than men were. Or recall the sad case of Tay, Microsoft’s teen slang Twitter bot that had to be taken down after producing racist posts.

In the near future, such mistakes could result in hefty fines or compliance investigation, a conversation that’s already occurring in the U.K. parliament. All mathematicians and machine learning engineers should consider bias to some degree, but that degree varies from instance to instance. A small company with limited resources will often be forgiven for accidental bias as long as the algorithmic vulnerability is fixed quickly; a Fortune 500 company, which presumably has the resources to ensure an unbiased algorithm, will be held to a tighter standard.

Of course, an algorithm that recommends novelty T-shirts does not need nearly as much oversight as an algorithm that decides what dose of radiation to give to a cancer patient. It’s these high-stakes decisions that will become the most pronounced when legal liability enters the discussion.

It’s important for builders and business leaders to establish a process for monitoring the ethical behavior of their AI systems.

Three keys to managing bias when building AI

There are signs of existing self-correction in the AI industry: Researchers are looking at ways to reduce bias and strengthen ethics in rule-based artificial systems by taking human biases into account, for example.

These are good practices to follow; it’s important to be thinking proactively about ethics regardless of the regulatory environment. Let’s take a look at several points to keep in mind as you work on your AI.

1. Choose the right learning model for the problem.

There’s a reason all AI models are unique: Each problem requires a different solution and provides varying data resources. There’s no single model to follow that will avoid bias, but there are parameters that can inform your team as it’s building.

For example, supervised and unsupervised learning models have their respective pros and cons. Unsupervised models that cluster or do dimensional reduction can learn bias from their data set. If belonging to group A highly correlates to behavior B, the model can mix up the two. And while supervised models allow for more control over bias in data selection, that control can introduce human bias into the process.

It’s better to find and fix vulnerabilities now than to have regulators find them later on.

Non-bias through ignorance — excluding sensitive information from the model — may seem like a workable solution, but it still has vulnerabilities. In college admissions, sorting applicants by ACT scores is standard, but taking their ZIP code into account might seem discriminatory. But because test scores might be affected by the preparatory resources in a given area, including the ZIP code in the model could actually decrease bias.

You have to require your data scientists to identify the best model for a given situation. Sit down and talk them through the different strategies they can take when building a model. Troubleshoot ideas before committing to them. It’s better to find and fix vulnerabilities now — even if it means taking longer — than to have regulators find them later on.

2. Choose a representative training data set.

Your data scientists may do much of the leg work, but it’s up to everyone participating in an AI project to actively guard against bias in data selection. There’s a fine line you have to walk. Making sure the training data is diverse and includes different groups is essential, but segmentation in the model can be problematic unless the real data is similarly segmented.

It’s inadvisable — both computationally and in terms of public relations — to have different models for different groups. When there is insufficient data for one group, you could possibly use weighting to increase its importance in training, but this should be done with extreme caution. It can lead to unexpected new biases.

For example, if you have only 40 people from Cincinnati in a data set and you try to force the model to consider their trends, you might need to use a large weight multiplier. Your model would then have a higher risk of picking up on random noise as trends — you could end up with results like “people named Brian have criminal histories.” This is why you need to be careful with weights, especially large ones.

3. Monitor performance using real data.

No company is knowingly creating biased AI, of course — all these discriminatory models probably worked as expected in controlled environments. Unfortunately, regulators (and the public) don’t typically take best intentions into account when assigning liability for ethical violations. That’s why you should be simulating real-world applications as much as possible when building algorithms.

It’s unwise, for example, to use test groups on algorithms already in production. Instead, run your statistical methods against real data whenever possible. Ask the data team to check simple test questions like “Do tall people default on AI-approved loans more than short people?” If they do, determine why.

When you’re examining data, you could be looking for two types of equality: equality of outcome and equality of opportunity. If you’re working on AI for approving loans, result equality would mean that people from all cities get loans at the same rates; opportunity equality would mean that people who would have returned the loan if given the chance are given the same rates regardless of city. Without the latter, the former could still hide if one city has a culture that makes defaulting on loans common.

Result equality is easier to prove, but it also means you’ll knowingly accept potentially skewed data. While it’s harder to prove opportunity equality, it is at least valid morally. It’s often practically impossible to ensure both types of equality, but oversight and real-world testing of your models should give you the best shot.

Eventually, these ethical AI principles will be enforced by legal penalties. If New York City’s early attempts at regulating algorithms are any indication, those laws will likely involve government access to the development process, as well as stringent monitoring of the real-world consequences of AI. The good news is that by using proper modeling principles, bias can be greatly reduced or eliminated, and those working on AI can help expose accepted biases, create a more ethical understanding of tricky problems and stay on the right side of the law — whatever it ends up being.

Agrifood — the $8 trillion industry that’s worth your salt

Cannabis-infused drinks. Burgers grown in laboratories. Entire meals in bottles. Consumers, retailers and farmers alike are hungry for the next generation of food, and investors are beginning to acquire the taste, too. Early-stage investment in agrifood tech startups reached $10.1 billion in 2017, a 29 percent increase on the previous year.

Agrifood can be split into two parts. “Agritech” refers to technologies that target farmers. “Foodtech,” by contrast, targets manufacturers, retailers, restaurants and consumers. Jointly, the two have enough reach to impact every part of the production line, from farm to fork.

Recently, foodtech investments have led the charge, with Delivery Hero’s IPO and multi-million rounds in and Instacart. However, agritech deals are catching up: Indigo Agriculture and Ginkgo Bioworks raised $203 million and $275 million, respectively.

There’s also more acquisitions activity in the sector. Some recently baked news suggests that both Uber and Amazon could be in talks with Deliveroo for a potential acquisition. Meanwhile, John Deere put $305 million on the table for the robotics company Blue River Technology, and DuPont acquired farm management software Granular for $300 million.

So why the growing interest in agrifood?

Food is a huge market, and it’s changing fast

Back in 1958, there were 3 billion of us on the planet. Today, population size has reached 7.6 billion, and is due to hit a whooping 11.2 billion in 2100. That is a lot of mouths to feed.

But the appeal of the food market doesn’t stop at volume. Indeed, following Bennett’s Law, as people’s income increases their diet becomes more diverse. This economic compulsion to seek variety is being complemented by a rise in ethical consumers voting with their forks. Many have grown aware of the link between food and ecology, health and animal welfare. The number of vegans in the U.S. has increased six-fold in the last three years, and more than tripled in the U.K. over the past decade.

This is not simply a case of having our cake and eating it.

These dual trends have led to supermarket shelves and restaurant menus evolving at pace. Consumers are keen as mustard to find new and healthy “superfoods” such as insects — Eat Grub and Cricke — and new consumption-forms, from meal replacement options like Huel to vegan meal boxes such as allplants.

When it comes to agritech, alternative production models have also arisen to cater to consumers’ preferences. Vertical farms such as GrowUp and LettUs Grow, for example, could dramatically reduce the environmental impact of farming.

Combining the above two ingredients — a growing population and a diversification in diet — cooks up quite the appetizing dish for investors: The global food and agriculture industry is estimated to be worth at least $8 trillion.

New technologies are creating big opportunities

The food and agriculture value-chain is full of bottlenecks and inefficiencies. Some of them could be solved with the intelligent application of well-known technologies.

The humble online marketplace, for example. Marketplaces, including Yagro, Hectare Agritech and Farm-r, let farmers transact machinery and goods, while peer-to-peer platforms like WeFarm enable knowledge sharing. Food procurement marketplaces have cropped up too, such as COLLECTIVfoodPesky Fish and COGZ — as have direct-to-consumer services, such as Farmdrop and Oddbox.

Some tech solutions are far more complex.

Genetic engineering, for one, is providing plenty of food for thought. Indeed, the UN suggests that food production must increase by 70 percent by 2050 to feed the world’s population growth. Genetic engineering could increase crop yields by 22 percent globally, as well as help pre-empt pre-harvest losses.

To this end, CRISPR is revolutionizing how food is grown. CRISPR technology helps producers optimize photosynthesis and the vitamin content of crops. Since it was first tested on tobacco production in 2013, CRISPR has been used on a range of crops, from wheat and rice to oranges and tomatoes; and for a whole spectrum of applications — from boosting crop resistance to pests, to improving nutritional contents. CRISPR is also being applied to livestock. At the Roslin Institute in Scotland, researchers have successfully used CRISPR to develop virus-resistant pigs.

In the same vein, there have been major advances in cellular agriculture. Cellular agriculture combines biotechnology with food and tissue engineering to produce agricultural products like meat or leather from cells cultured in a lab.

It is easy to see how cellular agriculture and the companies applying it, such as Meatable and Higher Steaks, could dramatically change farming and food production.

Therefore, investors have thus been tempted to take a bite at the “clean meat” industry. In Europe, Mosa Meat just raised $8.8 million, while U.S.-based Memphis Meats raised $17 million in 2017.

Even though products are yet to hit the shelves, the appeal is clear: The meat market will be worth $7.3 trillion by 2025, with a 73 percent increase in demand by 2050. And clean meat technology could allow for the production of meat at virtually infinite scale: In just two months, 50,000 tons of pork cells could be grown per bioreactor by using starter cells from 10 pigs. This could dramatically reduce the production cost of meat, and also its environmental cost: 6x less water is needed and 4x less greenhouse gas is emitted per pound of clean meat compared to “traditional” meat.

Artificial intelligence and machine learning is also impacting agriculture. One of the main opportunities, amongst many, is in precision agriculture.

Farmers now receive better information on crop status due to advances in image recognition, sensors, robotics and, of course, machine learning. Startups such as Hummingbird Technologies and Kisan Hub have developed solutions that outperform human “crop walks.” Similarly, Observe Technologies provides fish farmers with AI-powered insights to optimize feeding.

Consumer acceptance is also likely to be shaped by the economic, environmental and ethical implications of agrifood technologies.

Moving indoors, Xihelm (full disclosure, Oxford Capital is an investor) is developing a machine vision algorithm that enables roboticized indoor harvesting. Such technologies could help solve the labor crisis in agriculture: The 2017 labor shortage saw labor costs rise by between 9-12 percent in the U.K.

When the food moves from farm to retailers, supply chains can become unwieldy and difficult to manage. As a result, there is a $40 billion fraud problem in food. Blockchain technologies are being applied to solve this problem, powered by companies such as Provenance. Walmart recently announced that their leafy-green vegetable suppliers must upload their data to the blockchain, allowing them to trace food back to the source in 2.2 seconds instead of a week.

Agrifood tech is still an acquired taste

Although the agrifood market is huge and presents many opportunities for investment, it still isn’t quite the tech investor’s favorite dish. Yes, investment increased to $10.1 billion in 2017; however, fintech hit $39.4 billion in the same year.

There are several reasons. Digitization is growing, but it is slow. Farmers are understandably risk-averse. Their aversion is strengthened by the seasonality and fallibility of their activity. Most crops deliver produce once a year, so any missed harvest can have dramatic and long-lasting consequences. Implementing any large-scale technological solution represents a risk; therefore veering away from the status quo is a decision that cannot be taken lightly.

Regulation is a huge consideration for the sector. The Court of Justice of the European Union recently ruled that plants created with CRISPR must go through the same lengthy approval process as GMOs. In France in 2018, a law banned the use of terms like “meat” and “dairy” for vegetarian and vegan products — although it is not clear how this law will apply to cultured meat products in the future, nomenclature is a fight clean meat startups will want to win for the sake of consumer acceptance.

Consumer acceptance is also likely to be shaped by the economic, environmental and ethical implications of agrifood technologies. It is chastening to remember that agriculture employs one in four people in the global workforce, a large proportion of which are women.

The future of food could see unemployment issues in farming; large changes in livestock and feedstock production; and significant alterations in land management. Furthermore, gene editing is likely to benefit large corporations more than independent farmers — who could be put at risk.

This is not simply a case of having our cake and eating it. Instead, the ingredients need to be chosen with great care, or the “future of food” risks leaving a very bitter taste.

We’re addressing gender disparity in engineering way too late

STEM innovations, especially those in engineering, are an essential part of our modern-day lives. These innovations impact us all, and cut across social, economic and geographical boundaries. Yet, at a time when engineers must meet the needs of a vast population of users with diverse opinions and backgrounds, the engineering workforce continues to suffer from gender disparity.

The U.S. Department of Commerce reported that women accounted for 47 percent of all U.S. jobs in 2015. However, women only account for 24 percent of STEM jobs. And the percentage of women in STEM fields continues to be the lowest in engineering, with women representing just 15 percent of the workforce (NSF, 2018).

These are startling numbers — made even more striking given the range of STEM advocacy groups that making concerted efforts to increase female representation in engineering through programs that encourage women to enroll in engineering courses in high school, major in engineering in college and then go into the profession.

The problem, it seems, is that girls self-select out of engineering before these efforts even have a chance to be effective.

At a young age, girls internalize long-lasting stereotypes that tell them that boys are better at engineering and computer science, and that girls simply aren’t engineers. And during these formative years, they never have an opportunity to imagine themselves as engineers.

By the time we try to get young women involved in high school, their minds are already made up that engineering is not for them. Young women do not enroll in engineering-related secondary school courses at the same rates as young men, according to the 2018 National Science Foundation Science and Engineering Indicators Report: About two and a half times (21 percent) as many male students earned engineering and technology credits in high school as compared to female (eight percent). This gender disparity is also apparent in AP courses. In computer science, 77 percent of exam-takers are male.

Then, when women go on to college, they do not select STEM majors at the rate of men: 44 percent of men elect a STEM major compared to 24 percent of women, and only 19.3 percent of engineering degrees are awarded to women.

If we are going to bring more women into engineering, we must start to reach out to them when they’re still young girls.

We know from our work in creating the Museum of Science’s Engineering Is Elementary curriculum, which has been used by more than 15 million elementary students and 190,000 educators across the country, that when given the opportunity and when exposed to engineering concepts, girls are just as successful as boys at understanding the engineering design process. Additionally, a five-year of study of those curricula funded by the National Science Foundation found that girls perform just as well as boys on engineering outcome measures. (Exploring the Efficacy of Engineering is Elementary (E4) NSF No. 1220305)

This data is reinforced by what we see everyday within the halls of the Museum of Science: Girls like engineering if they get a chance to learn it.

More than one million kids have participated in Engineering Design Challenges at the Museum. Our research has shown that when girls immerse themselves in our exhibits, they demonstrate confidence and sustained interest in solving engineering problems and express an interest in future engineering activities (Auster & Lindgren-Streicher, 2013).

If young girls have the aptitude for and interest in engineering when they are able to experience it, and yet they are still not pursuing it as they get into high school and beyond, it means we are simply missing them.

It’s incumbent upon all of us to introduce girls to engineering, in both informal and formal educational settings, during the very earliest years of schooling. We can’t wait until high school and hope to sway them. Rather, it is time we expand engineering education to all children, starting as early as preschool — and then support educators in doing so — so we can build a learning environment in which engineering is part of girls’ daily conversations. When we start young, we never allow the stereotypes to take root in girls. They learn that all students are natural problem solvers and that all students are engineers — especially girls.

Smart marketplaces bridge the implicit and complex

Marketplace businesses are intrinsically linked to the technologies that enable them. There would be no Craigslist without email/SMTP, no eBay without the graphical browser and no Uber without location services and ubiquitous smartphones. As enabling technologies have evolved, marketplaces have grown to facilitate increasingly complex exchanges in new environments and industries. The next level? They will get smart: goodbye marketplace 1.0, hello smarketplace.

The theory is that dense edge computing and machine learning will enable marketplaces to understand more complex demands, and thereby facilitate transactions currently impossible using the present models. Before moving into a discussion on the “next era” of the marketplace, though, let’s review the history, to plot the course of marketplace evolution to date.

In the beginning, there was Craigslist. The progenitor of all digital marketplaces led the way, and its enabling technology was dial-up internet and email! It was the digital reproduction of the printed classified ads that early internet users were used to IRL. Craigslist grew rapidly and organically, but was hard to trust. There was no payments mechanism baked in, meaning users had to set up deals and payments “off marketplace.”

Things evolved. “Web 1.5” brought online payment technologies and better user interaction thanks to glossier programming frameworks such as CSS and HTML+. There were also simply many more users, increasingly ready to pay for goods online. Hence the rise of eBay, which later acquired PayPal (and later still, Amazon).

Alongside eBay rose the generation of the “vertical marketplace.” These had clear limitations: marketplaces were commonly restricted to a niche product, and (outside of eBay motors) users were generally unwilling to pay for high-value items online.

The co-emergence of social media and the camera-phone dramatically altered marketplace models. Social media removed trust barriers between users, and the ease of taking high-quality photographs made selling goods easier, and increased trust on the part of the buyer. This movement saw the emergence of marketplaces such as Etsy,* where users purchased specialist items from a vast range of sellers able to showcase both craft and personal stories.

As transacting became easier, users were further encouraged to sell and buy, and the marketplace platform became the (often vertically focused) clearinghouse that sat between them. Instances include, on the consumer side, Airbnb for travel and Habito for mortgages and Convoy, for trucking, on the b2b side.*

The latest species is the mobile, on-demand marketplace. Uber is the archetype — enabled by location technologies, increased mobile user density, flexible labor and behind-the-scenes supply-facing tools that bridge atoms to bits: scheduling, direct deposits and earnings dashboards.

The next evolution in marketplace businesses will be substantial. Enabled by sensor technologies and AI, marketplaces will be able to understand and satisfy complex multivariate needs.

The next generation is intelligent

Until this point, marketplaces relied on small numbers of variables to inform matches and transactions. For example: I tell Airbnb I would like to rent an apartment in Paris for a week in July and let it know my budget, and it shows me the options. While the pricing in the background may be extremely sophisticated, the customer’s explicit “expression of need” is a fairly straightforward equation.

Now, a new set of technologies will enable a next generation of marketplaces. AI will allow marketplaces to process richer data, and to therefore understand complex multivariate needs. So in the Airbnb example, combining previous trips to infer taste + available flights to balance costs + local events of interest to surface a packaged journey that is more likely to convert.

Smart marketplaces directly link complex demand to supply by understanding multiple needs.

Or take a common manufacturing process such as laser cutting — we have been meeting several online entrants into the space this year. Historically, this has been a non-trivial quoting process which can require expert CAD engineers, and a lot of back and forth between customer and machinist. To determine the right price, these experts consider the (available) laser-cutting machine, the desired material and specifics of the work. This is understandable for an offline over-the-counter transaction.

This is where machine learning could enter the conversation. With a rich history of orders and their resulting price, you could train a neural net that would ingest the 3D file and spit back out an accurate cost, without the customer explicitly defining the parameters. This would leave the factory to set a target margin, and by extension, bring the supply-side online to a manufacturing platform where the Boschs of the world could source parts.

In this way, smart marketplaces directly link complex demand to supply by understanding multiple needs.

The double cold start problem

Building a marketplace is not without challenges, notably achieving liquidity. Starting a (non-smart) marketplace business is difficult if you need to begin the supply and/or demand from a “cold start.” For example, consider the problem of starting a Deliveroo competitor with no customers and no restaurants on the platform.

Smart marketplaces are even more complex to set up. In addition to needing demand and supply-side engagement; the matching algorithm needs to be trained. To use a real example, Uber Pool could not exist without Uber providing the training data: It needed the pickup and routes data from the solo rides to optimize the communal rides.

Therefore, smart marketplaces face a “double cold start.” Not only do they have to populate supply and demand, but they also have to train their matching algos before they can be effective.

Alternatively, as platforms scale, concomitantly increasing the volume of training data, does the platform choose to eat the prediction errors (and consequently their profit margin), the better to train the model? If so, when does “break even” happen? It is worth it?

In the short term, this all prompts the question: Are horizontal smart marketplaces feasible, given the quantity and quality of training data required? If so, which categories could currently support it? The model may be tested in hyper-specific and data-rich markets such as manufacturing before it becomes more widespread.

These are not the only considerations for and limitations to the model. Builders of smart marketplaces need to ask themselves: Is AI a core competency “baked into” the platform, or is AI a “partner” technology that is brought in? If the latter, what does this imply for defensibility?

What next?

Smart marketplaces are coming: They will first emerge in analogue industries producing complex, high-dimensional data that humans are not good at describing: industry (customer part orders), content (stock footage exchanges), healthcare (MRI/CT related) and finance (commodities trading).

The current crop of marketplaces are likely to layer AI on top of their current offerings to reduce friction and increase efficiencies, but the truly smart marketplaces will be those reliant on AI to broker previously impossible transactions.

*Full disclosure: Mosaic Ventures is currently invested in Convoy and Habito. One of my partners, Simon Levene, was previously an early investor in Etsy.

Disruptive technology and organized religion

More or less since Nietzsche declared God “dead” nearly 140 years ago, popular wisdom has held that science and religion are irreparably misaligned. However, at a recent conference hosted by the Vatican, I learned that even in the era of artificial intelligence and gene splicing, religious institutions and leaders still have much to contribute to society as both moral compass and source of meaning.

In April this year, the Vatican launched Unite to Cure: A Global Health Care Initiative at the Fourth International Vatican Conference. This international event gathered some of the world’s leading scientists, physicians and ethicists — along with leaders of faith, government officials, businesspeople and philanthropists. The goal was to engage about the cultural, religious and societal implications of breakthrough technologies that improve human health, prevent disease and protect the environment. I had the privilege of participating as a board member of the XPRIZE Foundation.

We are living at a phenomenal point in human history. It’s a moment when our machines are flirting with godlike powers. AI and ever-accelerating innovations in medical technology are enabling humans to live longer than ever. Yet with increased machine capabilities and human longevity come heavy questions of morality and spirituality.

When bodies live longer, so do the souls inside of them. What are the spiritual implications for people who are given an additional 30 or even 50 years of life? Is enhanced longevity meddling with creation, or a complement to it?

As technology disrupts the way we relate to the few remaining physical and spiritual mysteries of humanity, it also disrupts the way we embrace religion.

It is here, at this nexus of technology and spirituality, that the Vatican wisely decided to bring together thinkers from both science and faith.

It was humbling to sit inside the tiny and unconventional country that we call Vatican City, surrounded by the world’s leading scientists, ethicists, venture capitalists and faith leaders. We talked about regenerative medicine, aging reversal, gene editing and cell therapy. We discussed how humanity is shifting from medicine that repairs and remediates toward a system that overtly changes our physical composition. We discussed the incredible augmentations available to the disabled — for example 3D-printed prosthetic limbs. How long before the able-bodied begin to exploit these enhancements to augment their own competitive advantage in an increasingly crowded world? To what extent, if any, should society attempt to control this paradigm shift?

One of the more interesting discussions surrounded how to ensure that humans don’t just live longer, but also better.

What exactly does “living better” entail? Does it imply physical comfort, spiritual well-being, financial security? At this moment in history, we have more instant and unlimited information than the kings and queens of ancient Greece or the Middle Ages could have ever imagined. That technological power is allowing more and more people to become enormously wealthy, at a speed and magnitude that would have been unthinkable for anyone other than a monarch just a century ago.

But are these people living “better”?

In as much as longer-living humans use their accrued wealth to support and encourage the creation of projects as audacious and ambitious as — for example — the Coliseum, I believe the answer is yes. If longevity and riches encourage the average human being to create change on a scale that matches the enormous potential of our exponential times — all the more so.

Yet, others in the room had a different take. For many religious leaders, “better” meant a more sharply defined relationship with God. For some scientists, “better” meant a life that creates fewer emissions and embraces better and smarter technology.

It was astounding, really. In one of the most hallowed spots on earth for the Catholic Church, sharing oxygen and ideas with cardinals and future saints, stood the world’s leading researchers, scientists and corporate leaders, who hold in their hands the technology to extend human life. Together with the clergy of the world’s great monotheistic religions, we held an open dialogue about how to improve the heart and soul of human life while the technology we create continues to advance beyond our ancestors’ wildest imaginations.

As technology disrupts the way we relate to the few remaining physical and spiritual mysteries of humanity, it also disrupts the way we embrace religion. In this conference, the Vatican very correctly leveraged the opportunity for organized religions to disrupt themselves by thinking about how they can be meaningful contributors to the conversation on spiritual, physical and mental well-being in the future.

Product Hunt Radio: Gen Z, what ‘the kids these days’ are using, and the future of social apps

In this episode of Product Hunt Radio, I’m recording from my home in San Francisco to talk to two young entrepreneurs.

Tiffany Zhong interned at Product Hunt while she was still in high school. After she finished school, she worked in venture capital before starting Zebra Intelligence, a startup helping brands and old people like myself better understand Gen Z. She’s also an investor with her fund, Pineapple Capital.

Drake Rehfeld is CEO of Splish, a Y Combinator-backed company that’s building social apps to make the internet more fun. He formerly worked at Snap, where he was one of the youngest hires, as well as at Team 10. Drake’s been a tech entrepreneur since high school, when he created a product for school events that made real money.

In this episode we talk about:

  • “What the kids are using these days” and all things Generation Z, including what they’re looking for in products and some of the common misconceptions about this younger demographic.
  • The projects that Tiffany, Drake and I started while still in high school, including the story of, a site I created with the goal of earning $100,000 that netted $70 before I shut it down. (Tiffany and Drake had more success with their high school ventures.)
  • “Digital influencers” on Instagram, what Gen Z thinks of them, and why you would start your own. Also — why any of this has anything to do with fake plants.
  • The phenomenon of a “finsta,” the ways that “the kids these days” are reshaping how identity works on the web and some of the experimental social apps that don’t have any of the typical social features like comments, followers or likes.

We of course also talk about some of their favorite products, including the HQ Trivia of music, a tool for creating your very own “digital influencer” and an anonymous app that (surprisingly) brings positive vibes.

We’ll be back next week, so be sure to subscribe on Apple Podcasts, Google Podcasts, Spotify, Breaker, Overcast or wherever you listen to your favorite podcasts.

10 lessons from Marketo’s growth to a multi-billion-dollar exit

With Adobe’s acquisition of Marketo, I have been reflecting on what an amazing and pioneering company Marketo has been since it was founded in 2006. There are very few tech companies that have defined a new category, executed a successful IPO, been acquired by a private equity firm for more than four times the company’s initial IPO market value and now, at a price of $4.75 billion, become the largest acquisition of a world-class company like Adobe.

The credit for this dream-come-true Silicon Valley company goes to the co-founding team of Phil Fernandez, Jon Miller and David Morandi, who together built an amazing customer-first product, defined a breakthrough category and launched a marketing automation company that continues to delight and amaze partners and customers alike.

I had the unique pleasure of meeting the founding team in 2006 when they shared their vision and passion for marketing automation. At the time, all they had was a PowerPoint deck. But it was clear then that they had a special idea and the unique capability to build a breakthrough product to deliver on their vision.

In all honesty, I couldn’t know how truly extraordinary the company would become. Thankfully, I was lucky enough that the team chose me and my former partner Bruce Cleveland as their first investor and also was fortunate to serve on the board for 10 years. Most recently, I was thrilled that Phil joined me at Shasta. One of the qualities I admire most about Phil — which was apparent all those years ago and continues to this day — is that he never stops iterating to do things better or faster or more efficiently or more thoughtfully. Phil always carried a notebook that said “THINK” on the cover, which epitomizes how he approaches his work.

Phil recently shared his “10 Things I’d Do Even Better If I Did It Again” presentation with our team and our founder/CEO community. We believe his insights are “10 Must-Dos” for today’s software entrepreneurs. It’s hard for entrepreneurs to know the trade-offs required when making the tough decisions — especially early on ­– but what follows is what I learned from Phil, and the key takeaways from his talk that I believe can help more founders create iconic companies with lasting value. (Note: Click here to view excerpts of Phil’s talk.)

Have one person own revenue

If your company is like every other company, there are two executives — vice president of Sales and the chief marketing officer — who are regularly locking horns because they are each tasked with taking different approaches to the same goal of increasing revenue. How do you solve this?

Hire a chief revenue officer (CRO) who can see both perspectives, plus give the context that sales and marketing are missing. This seat understands the big picture and doesn’t belong in marketing or sales. The CRO needs to talk strategically about life cycle revenue — across the customer journey. She or he should be a storyteller who can look at the numbers and the models and explain it all in plain English to the executive team so that everyone understands. Like a chief people officer, you’re going to have to spend on a CRO — but it’s worth it in the long run.

Hire a chief people officer (CPO) ASAP

Your company needs a leader of “all things people” who can make sure your workplace is welcoming, diverse and responsive to employee needs. For the staff to have trust, this person needs to be in a role that is empowered by the organization and not just by the CEO. Hire the most senior, overqualified HR executive into your business as early as possible — Series A level — and have him or her report directly to the CEO. By constantly listening to people — which is really hard when you’re working really hard — the CPO will help build your culture and be the eyes and ears for the CEO. Investing early in HR will come back to you tenfold through employee retention, team morale and an enviable culture.

Give back when it makes no sense

The day you think you’ve got to get a product release out the door and there’s no time to do anything else is the day you get out and give back in whatever way makes sense for your company and your community. Give employees time off to volunteer. Pick a cause for your company to support. Or, consider starting a charitable foundation with pre-public stock. It will create a spirit and energy that will give back to your team five or 10 times whatever it is costing you.

Charge your first customer

Phil personally wrote a stupid thing on their website that said, “At Marketo, your success doesn’t have a price.” That copy stayed up for years as a testament to how customer-centric they were. They were proud that they weren’t charging for services. But as Phil said, that was a big mistake; they should have been charging from day one.

When you’re a startup, short-range thinking is seductive, but long-range thinking is powerful.

There really isn’t any friction about asking customers to pay for services. If you say, “Look, this product is great. It’s going to transform your business but it’s not easy and it will cost money,” they will spend it. Feature-level sales is a great way to justify why you are charging what you are charging, and it keeps customers renewing services and adding more features as their business grows and changes. To make this strategy work, gear your sales metrics toward incremental increases over time­ instead of pushing sales reps to sell as much as they can all at once. Customers will pay for quality products that meet their needs.

Build a world-class Rev Ops/Sales Enablement team

You need a VP-level Rev Ops/Sales Enablement executive by the time your company reaches $2-3 million in revenue. That individual must think holistically about how revenue is happening, from the early lead in the door and the sale to renewal and the up-sale; understanding full lifetime value and thinking about it in a modeling sense. She or he needs to be a storyteller — one who can look at the numbers, look at the models and then explain it in plain English to the executive team. That’s gold.

Focus on continuous ARPU expansion

Today, to increase ARPU (average revenue per user), you need to design feature-level packaging every bit as much as how you design product functionally. The same people on product management ought to be thinking together with Rev Ops and Sales about how you dish out the product, how you launch the pieces, how you turn on pieces and how you enable pieces. It becomes a part of the art of product design as much as the art of revenue design — and that’s where these two rules of thought really come together. Basically, you need to design an expansion pass.

Incubate new product initiatives

Marketo failed in defining a multi-product company, from when it was $30 million a year to when it was $300 million a year. If you’re going to bring a second product line into the company — whether it’s organic or inorganic — it needs to be incubated. It needs to have its own dedicated sales team and its own separate quotas. If you’re thinking about becoming a multi-product company, do not pass Go, do not collect $200; go read Geoffrey Moores’ Zone to Win, the only business book Phil has ever recommended.

Pursue constant technology renewal

The pace at which tech is moving and the competitive advantage that new tech is providing over old tech has never been like this during the past 35 years. Today, you need someone that’s charged with thinking not about product but about the future. You need to value technical currency. If you’re three years old on your technology and a new company enters your market — the degree of agility, pace and performance the new entrant has in running circles around your company will win over a five-year cycle. Every time.

Always be seeking more TAM

No matter how good your initial tenure is, no matter how good it feels, no matter how amazing you see your company, as the CEO, as a leader, have a Plan B. Know what’s next, know where you’re going next and make sure you’re always talking about it. Be absolutely zealous about ensuring you know the next piece of TAM you’re going to go after. Think about what’s going to happen if you have more money; what would you do next? Give yourself that opportunity to dream, but make it real, make it defensible.

Watch the clock during scale up

When you’re a startup, short-range thinking is seductive, but long-range thinking is powerful. Always be watching the time. The tension between operating leverage and scale-up investment is really dangerous. At Marketo, they got to it late and their growth slowed a little too much. Live in the real world and focus on cash and on making the investments so you have the capacity when you need it. Have a long-range planning process and understand the day when you’ll need $2 million of ramp capacity. Don’t let the tyranny of a seductive short-range model triumph over what the real world is telling you about the dynamics of growing the business. Understand what it takes to really scale.

How this Kazakhstan internet giant built success on ideas from Russia and China

The advantage of entering an emerging market is that the market still has a lot of empty space to fill, and as a startup you can be the first player. Kazakhstan might not be the first country that comes to mind when you think of overseas expansion. However, it is the world’s largest landlocked country, and shares borders with Russia and China, which are important consumer markets as well as technology hubs.

In fact, companies in Russia and China provided good benchmarks for Chocofamily, now the biggest e-commerce holding in Kazakhstan. The 2011-founded startup’s current capitalization is $50 million, and they’ve hired 350 employees in their office in Almaty, the country’s largest city and previous capital.

The company claims it has 2 million registered users on its platform, and expects $170 million gross billings in 2018 with 7,000 purchases per day. Chocofamily launched their payment app, Rakmet, in 2017, following in the steps of WeChat Pay.

2011: Copying from Russia 

Looking at how Groupon was exploding in Russia, and how Delivery Club, a Russia-based food delivery service, was growing at a fast pace, the founder of Chocofamily, Ramil Mukhoryapov, decided the success could be replicated in Kazakhstan. So he quit his studies in Russia and went to Kazakhstan.

“Russia is three years above Kazakhstan. Check out what is happening in Russia and do the same in Kazakhstan, it is going to work in three years. That’s what we did, how we started the Chocofamily itself,” Nikolay Shcherbak, CEO at Chocofood says. “We just copied. If this works in Russia, it will work in Kazakhstan as well, because the markets are really close to each other.”

Ramil started with a daily deal service, Chocodaily, in Kazakhstan. After his first attempt was successful, he later started Chocofood, a food delivery service; Chocotravel, an online travel service; Lensmark, an online shop for contact lenses; and iDoctor, a platform with all the doctors in Kazakhstan where patients can find the doctor that they need and check doctor and hospital information and reviews. Now all these services are affiliates of Chocofamily Holding. 

How this company consolidated the market

In Kazakhstan, Chocofamily had competitors, but they either defeated them, purchased them or merged with them.

Chocofamily’s food delivery service, Chocofood, faced stiff competition in the market. Its rival was Foodpanda, which also started in Kazakhstan in 2013. After a four-year war of attrition, Foodpanda remained as No. 2 and wanted to get out of the market — so Chocofood acquired Foodpanda and took over their customer base and the legal entity. The company got additional growth uplift after integrating with Foodpanda.

“The best we got from the deal was the team, the people. They joined the company and just doubled our orders,” Nikolay says. 

Now Chocofood has an 80 percent market share in the food delivery market with 34,000 orders per month, working with 350 restaurants.

Online travel services Chocotravel merged with its rival Aviata in 2017. Now the two companies take a 67 percent combined market share. They’ve been profitable since the second half of 2018, with 80,000 air tickets and 50,000 railway tickets sold per month. 

In the future, Chocotravel plans to enter the South-Eastern and CIS (Commonwealth of Independent States, namely northwest Russia, Eastern Europe and the Baltic states) travel markets, and they are looking for additional funds for expansion.

2017: Copying from China

In 2017, the company launched a payment app, Rakmet, which means “thank you” in Kazakhstan. It allows users to pay for purchases by simply scanning QR codes. For users, the advantage is that every merchant gives them cash back.

The idea for the app came about in early 2017, as they were looking at other companies in different countries. 

“We were looking at WeChat, and it had a good system of using QR codes for payments. We thought it was a good idea. QR code technology is really old, but it only comes to our everyday life now,” Nikolay said.

Like other payment apps, users can use the Rakmet app by connecting their bank card to the app as a payment option. With a population of 18 million in Kazakhstan, bank card penetration is quite good; 19 million bank cards have been issued in total, and there are 10 million active bank cards.

For businesses to join the Rakmet app, they must give a certain percentage of cash back to customers. The approach has been especially popular with cafes and restaurants that have been using loyalty cards to attract consumers. Nikolay says it is a marketing strategy for merchants, because they’re paying the commission to Rakmet only for those transactions. To date, 300+ predominantly small businesses in Almaty have posted their stores via the Rakmet app. 

“Rakmet app will be on top of the ecosystem of all Chocofamily affiliates. We also plan to add different services to Rakmet app, such as allowing users to pay traffic fines to the government on the app and pay the parking fee using the app,” Nikolay says. 

Women are typically responsible for the household in Central Asia. Thus, women make up 60 percent of their users, making transactions on mobile such as booking flight tickets, ordering food, and making doctor appointments. The biggest growth is among their users in the age group between 25 to 35.

They also are working on big data. The team is now building the infrastructure for big data analysis, such as data warehousing and the support. Then they plan to build the mechanism for data processing. In September, they signed a contract with one of the universities in Kazakhstan so they can attract students who are experts on big data analysis.

In 2011, Chocofamily started with their own money. In 2013, they received $50,000 from two angel investors, then another $150,000. Then they attained Series A funding of $1 million from Murat Abdrakhmanov, an experienced entrepreneur in Kazakhstan, and later received $2 million in Series B.

Brazil’s healthtech sector is new hot spot

Solving big problems for many people is the kind of opportunity that both entrepreneurs and investors love. Like recent Brazilian investment booms focused on fintech innovation and new on-demand business models, there’s been a recent explosion in healthtech startups in Brazil. With tens of millions of the country’s people impacted by gigantic inequities in access to health services, some serious quality problems, burdensome costs and inefficiencies on all sides, entrepreneurs’ plates are full in bringing healthtech innovations to the market.

In a recent study by Liga Ventures, there are now more than 250 health-focused startups in Brazil, the world’s seventh-largest health market with more than $42 billion spent annually on private healthcare. Yet, with more than $18 billion wasted due to inefficiencies, and health-related costs doubling in the country during the last five years (with accumulated inflation at 38 percent), Brazilian healthcare is ripe for disruption. Healthtech startups are one of the five featured verticals at Cubo Itaú, one of world’s largest entrepreneurial hubs based in Vila Olímpia, in the southern zone of São Paulo.

Last year, healthtech was the second-fastest growing tech sector in Latin America, according to “Inside Latin America’s Breakout Year in Tech” published by LAVCA. There was a 250 percent increase in the number of healthtech deals compared to 2016. A $50 million investment in Dr. Consulta, a network of brick-and-mortar clinics in Brazil offering top-quality healthcare at an affordable price, was among the top venture capital deals for 2017.

The healthcare sector is a complex market that connects people, processes and products between patients, intermediaries, care providers, distributors and suppliers. Based on tech innovation in Brazil that’s having the biggest impact, here are some of the key categories and players bringing new business models to market.

Healthcare on demand

About 75 percent of Brazil’s population (approximately 150 million people) only have access to the public healthcare system, which is poorly managed and inefficient. Often times, to schedule a single consultation or exam, a patient needs to wait weeks or even months to see a care provider. Technology-driven startups are springing up to address better, more efficient access to healthcare for a large and aging population.

For example, Dr. Consulta’s chain of low-cost medical clinics have expanded in three years from one to 51 branches and now claim to have the country’s largest clinical data set drawn from more than one million patients. In comparison to other private-sector clinics that cost at least $90, consultations with doctors at Dr. Consulta cost $25. Others offering similar clinical services on demand in Brazil today include Clínica SimDr. Sem FilasDocway and GlobalMed.

Telehealth and mobile health apps

To help make healthcare advice, diagnosis and monitoring more accessible, telehealth services in Brazil are expanding. Brazil Telemedicine (Brasil Telemedicina), for example, provides a variety of services around the clock that include medical exams and doctor consultations, a remote monitoring system and psychological counseling.

Startups with B2B telehealth services to improve patient care include Telelaudo, which provides 24/7 radiology imaging analysis, and Ventrix, which provides specialty devices to monitor heart health, treat vacuum wounds and monitor babies’ breathing and well-being. Another São Paulo-based startup called NEO MED has launched a marketplace to make it easier and faster to generate medical reports for ECG and EEG exams, facilitate improved collaboration between clinics, laboratories and hospitals and support physicians seeking more income and flexibility in where they choose to work.

The key ingredients to create another boom sector like fintech in the region are abundant.

Mobile health apps have grown in popularity in Brazil, in part due to a high prevalence of diseases like diabetes and hypertension and a large number of internet users in the country. For example, a mobile app and online program called Diet and Health (Dieta e Saude) has helped more than 1,600,000 users make better nutrition choices and motivate them to exercise regularly. Youper, founded in Brazil and now based in San Francisco, is a virtual emotional health assistant that helps overcome social anxiety. It helps its users re-formulate thought patterns and arrive at healthier states of mind.

AI and data analytics

Like many industries, AI and data analytics are transforming healthcare in Brazil and beyond from improving the speed of patient diagnoses to managing healthcare costs.

Gesto is one such emerging innovator that’s using machine learning to sift through and make sense of a lot of data on more than 4.5 million patients in its database to help select better insurance plans for corporations that optimize patient care while controlling costs. Intensicare, the largest specialist in intensive care unit management in Brazil, uses AI to speed diagnosis and reduce patient stay time and mortality rates. Epitrack is a Recife-based startup that uses crowdsourced data, AI and predictive analysis to combat outbreaks and epidemics through computational epidemiology.

Electronic medical records

Last year, the Brazilian government launched a project to modernize patient records for more than 42,000 public health clinics across the country by the end of 2018. This digitization of records is estimated to save the federal government about $6.8 billion according to The World Bank. As of late last year, only 30 million Brazilians (out of 208 million) had electronic medical records (EMR), and nearly two-thirds of the family clinics in Brazil didn’t have any way of recording digital information about their patients.

iClinic, a SaaS EMR platform, is one of the top Brazilian startups that has made a big impact on modernizing healthcare. It helps health professionals organize patient records electronically, store all that data in the cloud and retrieve it from any device. iClinic provides an extremely easy-to-use system to make healthcare more efficient, reduce costs and improve the quality of patient care. It’s now used in many parts of Brazil and has begun to spread its usage outside Brazil in more than 20 countries.

Digitizing prescriptions

Another major issue caused by a lack of digitization is that close to 70 percent of medical prescriptions in Brazil have potential for errors, according to the World Health Organization. As a result, Brazil has thousands of deaths per year linked to medication errors. A good number of them could be avoided by scanning. In the U.S., more than 77 percent of prescriptions are already done digitally.

To address this life-and-death issue, Memed has emerged as a key player for managing e-prescriptions in Brazil. Its platform, now used by more than 55,000 doctors from all medical specialties in the country, helps cross-check for allergies and drug interactions, makes treatment adherence easier and improves health outcomes. It’s developed the most complete, reliable and updated drug database in Brazil.

Certainly, healthtech startups in Brazil have emerged as a sector to watch, and we’re only at the tip of the iceberg in terms of problems in the country to be addressed by healthtech innovation. The key ingredients to create another boom sector like fintech in the region are abundant. Healthtech in Brazil will surely remain a hot spot for entrepreneurs, and the investors who believe in them, for many years to come.

Disclosure: Redpoint eventures is an investor in Memed.