The end of platforms as we have known them seems to be around the corner. This blog is the second chapter in our blog series “Rise and fall of the platform business”.
All the major platform companies base their entire strategy on getting access to more data in order to teach their machine learning systems. The end goal for them, of course, is the same as everyone else’s as these companies are looking to use data for analytics to create insights. Yet, they might be mistaken in their assumptions that big data drives insights, and the same could not be done more effectively from first choosing what insights are needed for behavior change and then picking the right data sources to use. There are four reasons why their big data-focused investments may be misdirected, which are introduced below.
There are many (1) legislative actions ongoing regarding data protection and rights of use. Open data is increasing. There will be more flexibility in the fair use of partly open public data. There is a legislative push for myData, especially in the EU. In addition, in 2018 the General Data Protection Regulation (GDPR) in the EU will set stricter barriers of use for different datasets (forbidding secondary use).
(2) The barriers in data sharing so far have been the fact that it is very difficult to share data in a way that prevents the other party from accessing the actual data. There are two approaches that might help circumvent this issue. First, it is possible to build blockchain-based systems, which allow for complex analytics without accessing the unencrypted data. Secondly, especially with regards to large public datasets, it is possible to simulate large datasets in order to anonymize them. This simulated dataset would not have any actual entities in it, but it could still be used mathematically to compare individuals with the whole set or a certain subset of the population.
Despite the current hype, (3) machine learning does not develop linearly towards general AI. By increasing the layers in deep learning and the size of matrices, some more complex problems may be solved with better accuracy. However, for each problem, there is an optimal size for a matrix and an optimal number of layers, and these are not, respectively, the biggest size and number possible. During the next decade, the current deep learning paradigm begins to yield diminishing returns.
At the same time, there may be interesting developments in semi-supervised learning. This refers to the (4) ability to use much smaller sample sizes for teaching the machine learning system and then using unlabeled clustering to get results that are as good as the ones yielded by the very large annotated datasets of supervised learning.
If any of the four aforementioned developments significantly help companies compete using smaller datasets, the competitive barrier built by the platform economy companies starts to crumble. The question that follows is: Is it possible for data-poor companies to skip the value creation paradigm based on machine learning, big data and devices and move directly into the new and upcoming paradigm where machine learning is used to produce actionable insights and initiate behavior change?
Platform companies are aware that their competitive advantage is diminishing. Exactly this is the reason they are using their large resources to desperately trying to find new protected sources of monopoly. Because many platform companies managed to collect immense cash reserves, they are now investing these cash reserves to tangible assets that might offer them monopoly protections.
From the perspective of the platform company, they have made a great switcheroo: they were able to benefit greatly from an anomaly in the business environment where they were as first movers able to capture immense temporal monopoly rents from their services. They are now in a very good position to make another such move, but this time their history might blind them to be too brave: they should not try to invest hugely to physical assets to win the temporal monopoly, but instead they should aim to identify next critical turning points in demand chains and invest in abilities to control the market after that change. It’s going to be a struggle, as they might not have the patience and humbleness to do this.
What to read next?
Continue to Chapter 3: The fall of platform business model
Go back to Chapter 1: The secret ingredient of the first generation platform business
Read the intro: Rise and fall of the platform business