Getting Algorithms into Business
The human race is now at an incredible intersection of unprecedented volumes of data, generated by increasingly smaller and powerful hardware, and analyzed by algorithms that this same process helped develop. It’s not simply a matter of volume, which by itself is a difficult challenge.
As formalized by the research company Gartner in 2001 and then reprised and expanded by other companies, such as IBM, big data can be summarized by four Vs representing its key characteristics:
- Volume: The amount of data
- Velocity: The speed of data generation
- Variety: The number and types of data sources
- Veracity: The quality and authoritative voice of the data (quantifying errors, bad data, and noise mixed with signals), a measure of the uncertainty of the data
Each big data characteristic offers a challenge and an opportunity. For instance, volume considers the amount of useful data. What one organization considers big data could be small data for another one. The inability to process the data on a single machine doesn’t make the data big. What differentiates big data from the business-as-usual data is that it forces an organization to revise its prevalent methods and solutions, and pushes present technologies and algorithms to look ahead.
Variety enables the use of big data to challenge the scientific method, as explained by this milestone and much discussed article written by Chris Anderson, Wired‘s editor-in-chief at the time, on how large amounts of data can help scientific discoveries outside the scientific method. The author relies on the example of Google in the advertising and translation business sectors, where the company could achieve prominence without using specific models or theories, but by applying algorithms to learn from data. As in advertising, science (physics, biology) data can support innovation that allows scientists to approach problems without hypotheses but by considering the variations found in large amounts of data and by discovery algorithms.
The veracity characteristic helps the democratization of data itself. In the past, organizations hoarded data because it was precious and difficult to obtain. At this point, various sources create data in such growing amounts that hoarding it is meaningless (90 percent of the world’s data has been created in the last two years), so there is no reason to limit access. Data is turning into such a commodity that there are many open data programs going all around the world. (The United States has a long tradition of open access; the first open data programs date back to the 1970s when the National Oceanic and Atmospheric Administration, NOAA, started releasing weather data freely to the public.) However, because data has become a commodity, the uncertainty of that data has become an issue. You no longer know whether the data is completely true because you may not even know its source.
Data has become so ubiquitous that its value is no longer in the actual information (such as data stored in a firm’s database). The value of data exists in how you use it. Here algorithms come into play and change the game. A company like Google feeds itself from freely available data, such as the content of websites or the text found in publicly available texts and books. Yet, the value Google extracts from the data mostly derives from its algorithms. As an example, data value resides in the PageRank algorithm (illustrated in Chapter 11), which is the very foundation of Google’s business. The value of algorithms is true for other companies as well. Amazon’s recommendation engine contributes a significant part of the company’s revenues. Many financial firms use algorithmic trading and robo-advice, leveraging freely available stock data and economic information for investments.