UNC-Charlotte hosted an Analytics Frontiers Conference at the end of March. Over 300 people attended the event which offered great speakers, great breakouts and a great venue. A highlight was data science guru Tom Davenport who shared his thinking on what he called the 4 eras of analytics much of which he has covered in his books and articles. The Eras map well to what I see as the 4 P’s of data – pinpoint, pronounce, predict, and prescribe. The 4 P’s of data can be used by an organization to assess how they are using their data; they can also be used to track the evolution of tools and techniques for managing data with an organization. Recording data –pinpointing -- predates traditional analytics (Davenport’s 1.0 Era) whereas autonomous future action based on data – prescribing -- is only possible in a cognitive technology enabled world (Davenport’s 4.0 Era). A mapping of the 4 P’s of data to Davenport’s 4 Eras follows below.
Pinpoint – to identify or determine. At its core data are information bits -- the translation of an observation into information for storing or sharing. Firms have tools and processes to measure performance or activity. The observation of these measurements by themselves pinpoint what is happening. In a “before analytics” world these data were reacted to in the moment. To use a stop light analogy when the light is green, go. The advent of traditional analytics, Era 1.0, circa 1975, coincident with innovations in computing technology to capture and store data at a reasonable scale, created opportunities for firms to not just pinpoint data but to analyze and pronounce what happened and why. Era 1.0, according to Davenport, is characterized small, structured data used for internal decision support. Many firms today are still operating in the 1.0 era, think Data Warehouse and descriptive analytics.
Pronounce – to pass judgment. Excel and its predecessors, VisiCalc and Lotus 1-2-3, were the killer apps of traditional analytics. Data alone are descriptive and backward looking. Analytical tools allow users to review data on “what happened” and look for patterns to explain why things happened. These pronouncements enable business intelligence. The data are structured but greater volume gives rise to visual tools to make it easier for firms to grasp data that may still be small in today’s world but exceed what the human brain can handle. In a pronounce environment data are important but decisions are still often instinctual. Around 2001, when the term data scientist was coined, tools and techniques evolved so instead of just pronouncing what had happened in the past firms could use data to predict what might happen in the future. Early adopters of analytical tools had a competitive advantage. As the power of analytics grew firms entered Davenport’s Era 2.0 where analytics plays a bigger role and organizations become more data driven.
Predict – to declare in advance. Silicon Valley firms were early adopters of big data tools. The rise of the internet and the dot com bubble introduced firms to the first three V’s of big data: large volumes; high velocity and varieties of structured and unstructured data. Many firms like Google got their start from finding business models that took advantage of data and technology tools. These tools allow firms to analyze past data and predict future activities – think regression on steroids. As predictive tools move beyond Silicon Valley and more firms look toward digitization they are entering Era 3.0. Davenport describes this era, circa 2013, as being characterized by machine learning and the rise of Hadoop. Data is not only big but unstructured data exceeds structured data in the potential for research and learning. This Era is the Digital era.
Prescribe – stipulate action to be taken. Cognitive computing is Era 4.0. With embedded analytics, systems and processes can be automated and can learn how to respond to different data. Autonomous action will be the future but the transition faces many barriers some societal and regulatory but also data science barriers. In his presentation, Davenport used the Internet of Things as an example of interoperability barriers. He shared that a typical automobile has 200 – 300 sensors and 195 different data formats. Until the formats are consistent or can communicate the potential will not be realized. Today data science tools are used to augment decisions or processes for true automation, interoperability and other challenges need to be solved. What do you see as the data challenges of cognitive computing?
If your organization uses data to pinpoint or pronounce beware because prediction techniques are quickly becoming table stakes for most industries. As Davenport said in his talk the “leading edge is sharp.”