Prescriptive analytics: an adaptive crystal ball
Knowledge is power, according to Francis Bacon, but knowing how to use knowledge to create an improved future is even more powerful. The birth of a sophisticated Internet of Things has catapulted hybrid data collection, which mixes structured and unstructured data, to new heights.
According to Gartner, 80 percent of data available has been collected within the past year. In addition, 80 percent of the world’s data today is unstructured. Even now, most corporations use descriptive or diagnostic analytics. They use existing structured data and correlated events, but usually leave the newer, richer, bigger unstructured data untouched. The analyses are built on partial data and usually produce incomplete takeaways.
Smarter analytics to the rescue
Gaining momentum is a newer type of analytics, called prescriptive analytics, which is about figuring out the future and shaping it using this hybrid data set. Prescriptive analytics is evolving to a stage where business managers – without the need for data scientists – can predict the future and make prescriptions to improve this predicted future.
Prescriptive analytics is working towards that “nirvana” of event prediction and a proposed set of desired actions that can help mitigate an unwanted situation before it happens. If a machine prescribes a solution anticipating a future issue and you ignore it, the machine can think forward and adapt automatically.
Skynet from Terminator is a classic, though evil, fictional example of this concept. In that movie, Skynet initially makes predictions and suggests a course of action about the future to help mankind, which is the purpose the system was created for. But after becoming self-aware, Skynet issues prescriptions to other controlled machines to shape the future, and even goes so far as to send a cyborg into the past to kill John Connor, the one person capable of destroying it.
Of course, unlike the fictional Skynet’s nefarious goal, today’s prescriptive analytics tools are designed to help companies differentiate themselves from their competition. These good machines are not allowed to take control over decision-making, but they are an important part of suggesting decisions to shape the future.
Prescriptive analytics algorithms recalibrate themselves. As the incoming data evolves so do the algorithms – they re-fit, re-predict and re-prescribe. In this case, you can continue to set the business expectations with the machine that may someday do most of the thinking for you.
A useful example of predictive analytics in action would be in the oil and gas industry. Fracking has led to the recent energy boom, but it brings with it not only environmental concerns, but serious inefficiency: 80 percent of the production usually comes from 20 percent of the frack stages. (Drillers spent an estimated $31 billion in 2013 on suboptimal frack stages across 26,100 U.S. wells.) The enormous complexity of the data sources generated during exploration and production processes makes it difficult to make fracking better and safer at the same time.
The petabyte datasets include images (seismic, mud logs, well logs, offset logs, etc.), sounds (of drilling, fracking, completion and production, etc.), videos (cameras monitoring down hole fluid flow; fiber optics monitoring pressure, temperature, strain; etc.), texts (notes taken by drillers and frack pumpers, etc.) and numbers (production data, artificial lift data, etc.). Prescriptive analytics takes this challenge head on by proactively determining where to drill or frack and why as well as how to stimulate and complete wells to maximize production and minimize environmental disruption.
Prescriptive analytics can save millions of dollars
A typical shale well can have a vertical section that ranges from 7000 to 14,000 feet and a horizontal section that can go for 1 to 3 miles, depending on the shale play. Many operators have drilling rights to hundreds of thousands of acres. Determining where to drill in one’s acreage position – for maximum production, of course – is paramount to success. Also, each horizontal well may have 30 to 40 frack stages. We now know from fiber optic sensors, microseismic, etc., that many fracks don’t produce much or even any oil and gas. Prescribing suitable frack locations for maximum output can change the game.
Today, oil and gas industry stimulates and re-stimulates a well multiple times over a well’s life to improve recovery and economics. By synthesizing all the disparate datasets using prescriptive analytics, we can make the first stimulation treatment much more effective thus reducing the need for subsequent treatments. Also, the fewer wells we have to drill to exceed production targets, the lower our environmental footprint will be.
Prescriptive analytics is a domain-agnostic technology; it learns the process and adapts to it automatically and continually using all the data coming at it.
The time has come for machines and humans to work together to make each other smarter. The combination of IoTs, big data, smarter analytics and cognitive computing is transforming the way we see the future — and more importantly, what we do about it.
Andy Thurai (@AndyThurai) is Chief Architect and CTO of Intel App Security and Big Data and blogs regularly at www.thurai.net/blog. Atanu Basu (@atanubasu) is the CEO of Ayata (@AyataAnalytics).
This article was originally published in gigaom on MAR 22, 2014 – https://gigaom.com/2014/03/22/prescriptive-analytics-an-adaptive-crystal-ball/