An interesting article on data economics by Martin Casado and Peter Lauten on Andreessen Horowitz’s site, entitled “The empty promise of data moats”, raises some interesting ideas about the supposed competitive advantage of holding large amounts of data.
The authors argue that data in itself provides no defense against competitors. Sure, having quality data about its customers and operations and having established a solid analytical culture can help a company create a knowledge base of its sector and its key variables, while its customers may allow it to create differential value proposals, but in practice, data does not respond to economies of scale or network effects as such: accumulating more data, over time, does not improve costs or result in better analytical systems, in fact it tends to lead to the opposite: higher storage, processing and analysis costs.
Having more data can help with using more powerful analytical methodologies to draw conclusions and create business rules. However, the effect of the amount of data falls when these conclusions are obtained with reasonable quality and are evaluated positively.
From then on, growth in data volume really contributes little or nothing, and doesn’t do much to build entry barriers for potential competitors thinking about a similar business model. As a service provider, moving from a few clients to a million will allow a company to learn much more about the business and derive rules, recommendation systems, etc. that potential competitors may not have, giving them competitive advantages. However, when a company already has enough data to facilitate this process, continuing to accumulate more not only doesn’t improve conclusions but simply generates more costs, more noise and more complexity.
Over time, your competitors will no longer have to worry about accumulating more data than you, but instead how to obtain the minimum amount to draw meaningful conclusions and analyses; in other words, moving away from a linear conception of data economies and instead evaluating a series of more practical concepts such as being able to evaluate the size of a minimally efficient set of data, as well as understanding the costs of data acquisition, its incremental value or its newness. Generating and storing new data can help maintain up-to-date business rules and advanced automation processes that really do make sense or even lead to new effects, but we should avoid becoming data hoarders, because this contributes very little to creating a sustainable competitive advantage.
Understanding the rules of the data economy is increasingly important for all types of companies. In every analytical, machine learning or advanced automation project, the truly significant costs are those that we have to incur so as to define our objectives, to collect and transform data and to propose the analysis, cleaning and structuring of data fields that separate noise from signal. All these factors require, of course, large amounts of data, but the effect is neither linear or cumulative and proposing appropriate policies in this regard can help develop models that not only make better economic sense, but are also more agile and dynamic. It is increasingly simple to apply models on the basis of properly structured data, as opposed to being obsessed with quantity. And as time moves on, this will be even more so.
(En español, aquí)