Much has already been written about Google, Amazon, and Facebook’s aggressive approach to artificial intelligence and the potential this shift has to leave Apple in the dust. Such is the plight of technology companies that fail to succeed in their quest to find the next technological winner as described in W. Brian Arthur’s seminal work on Increasing Returns. Arthur writes:
Competition is different in knowledge-based industries, because the economics are different. If knowledge-based companies are competing in winner-take-most markets, then managing becomes redefined as a series of quests for the next technological winner — the next cash cow. The goal becomes the search for the Next Big Thing.
However Arthur’s philosophy is backed up by several characteristics of companies in “knowledge-based” markets or what is more popularly described today as “tech companies”. These characteristics are “up-front costs, network effects, and customer groove-in”.
These factors have been the hallmark of a company that raises venture capital money. The idea being to invest large sums of money today to offset near term losses and reap the benefits of potential extremely large profits in the future.
Somewhere along the line though, this changed. As the cost to build a technology product has continued to plummet and the “customer groove-in” becoming a thing of the past (consumer-first and more recently enterprise with companies like Slack and Trello).
As a result, more and more companies were built but despite being technology companies, a good number of these businesses operate in markets that have diminishing returns, competing for customer mindshare with competitors while gaining fewer profits in the process despite leveraging technology to sell their product and starting out as a “startup”.
Since the VC model is still very much predicated on these increasing returns businesses (power law, anyone?), we’re increasingly looking to invest in businesses that have similar characteristics described by Arthur over twenty years ago.
The startups I see today that fit this criteria have combined network effects and customer groove-in to form “data network effects”. These data network effects offer a superior experience for customers or users leveraging data shared from the network. Matt Turck from FirstMark Capital wrote an excellent post on the topic which can be found here.
These companies leverage up-front costs to provide a free or low margin product in order to collect vast amounts of data from a variety of stakeholders that in turn increase the value of the product (and in some cases the product can be the data). Below are some general themes I’m thinking about:
Human Augmented AI: These companies are typically b2b (but not always) and improve productivity and workflow for their users. The big advantage here is that while vast sums of data are still necessary for a great product, the end users are actually both providing the training data as well as the positive and negative reinforcement necessary when the system fails to understand language nuance or fails in other unforeseen ways. With these companies, the product is saving time and increasing efficiency for the end user it isn’t nearly as frustrating for them when they need to step in and correct a mistake or clarify a system’s error as it would be to a user with sky high expectations. We’ve made two investments in this space to date, Digital Genius and Counselytics.
Open Data Hardware: These companies have built hardware that collects public, but difficult to obtain data from a variety of sources whether that’s through their customers or by building a new way to access it. By collecting all of this data, these companies can begin to make inferences relating and prediction correlation. The more data that is collected, the more valuable the data is to end customers. We haven’t made an investment in this space to date but some companies I’m intrigued by include Nexar, Saildrone, Planet Labs, Cty.io, Spire, Swift Navigation, and Understory.
Machine Learning Companies leveraging unique, proprietary or narrow training data: These companies do more than just collect your own data and allow others to interact with it or generate suggestions and insights. Rather, they harness data that is unique to their end customers, create valuable insights that improve the product for all users, and provide a useful tool in the process. They can be software (Clarfai, Flatiron Health) or hardware (Kinsa Health or even Tesla). We’ve made two unannounced investments in this space to date, one hardware, one software.
Data-as-a-Platform: Mary Meeker talked about this in her 2016 Internet Trends report and we’re seeing more and more companies targeting industries that either paid expensive consultants to wrangle their data and make it useful or were in the dark completely. These platforms add a layer of intelligence to company data in a specific vertical and can leverage the combined data of all of their customers to create better insights for everyone involved. Thus far we have one investment in this space, Zodiac but there are a number of other interesting companies not mentioned by Mary Meeker including Panorama Education, Invenia, and GovPredict,
This is far from a comprehensive list of companies that fit into the bill but as the cost of sensors and the cloud continues to plummet, I’m spending a lot of time thinking about companies and industries that can leverage the traits listed above. If you’re working on something like that, I’d love to hear from you!