Giz Explains: How Prescriptive Analytics Could Harness Big Data To See The Future

Giz Explains: How Prescriptive Analytics Could Harness Big Data To See The Future

Big Data — information sets too large to be effectively computed on desktop systems — isn’t just the buzzword du jour. It provides an unprecedented ability for business and industry to precisely model the effects of past managerial decisions on the bottom line. But an emerging analytical process called prescriptive analytics could help companies not only learn from previous decisions, but predict and plan for upcoming issues as well. It’s Big Data evolved, and it could change how the world does business.

Big Data, which relies on structured data sets of crunchable numbers, is just a small portion of the total available data produced daily — about 20 per cent, according to IBM estimates. Text, images, video, social media streams, machine data and audio on the other hand, collectively known as unstructured data, are simply ignored, because they don’t fit neatly on a spreadsheet despite containing a wealth of mineable information. However, by combining structured and unstructured data into a hybrid data set, companies are afforded a complete and total view of the issue and can, in theory at least, make the best possible decision based on it. And it’s these hybrid data sets that prescriptive analytics utilises to predict the future.

Prescriptive analytics is the third phase of business analytics, a decision-modelling system for industry. The first stage is descriptive analytics, which looks at past issues and describes them; it tells you what happened and why after the fact. The second stage is predictive analytics, which combines historical data with predictive algorithms to guess the probability of future events; it tells you what will happen. But prescriptive analytics claims to go even further. It applies a multitude of business rule algorithms, multiple mathematical and computational modelling systems to automatically synthesise hybrid data sets and answer not only what will happen but what also needs to be done about it. Put another way, prescriptive analytics continuously and automatically tries to anticipates the what, when and why of unknown future events. And it has the potential to be scary accurate.

“It answers what will happen, when it will happen and why it will happen,” Atanu Basu, CEO of the Ayata company, one of the leading developers of prescriptive analytics software, told Gizmodo. “And then how to take advantage of this predictive future.”

“I’ll give you a cheesey example,” he continued:

You’re driving, you see a car slowing down ahead. You hit your brakes and you get hit by the car behind you. So you avoided the accident with foresight or prediction and caused another in the process. So foresight is great — provided you can make an effective decision based on the parameters that would affect the decision. Your priorities, buoyancies, capabilities and constraints so far and so forth. So it’s [prescriptive analytics] not only about looking ahead but making the right decision without compromising other things you care about.

Based in Austin, Texas, Ayata began as an independent R&D effort in Toronto, Canada. Since incorporating in 2009, the 35-person company has worked for the likes of Cisco, Dell and Microsoft. Ayata’s software utilises either a private or hybrid cloud model supplied by its customers to perform its continual analysis so as to keep the client’s data, much of which is quite sensitive, secure.

The school of prescriptive analytics has only been in practice since around 2003, so it’s still very much in its genesis stage and, as such, the PA field is littered with failures. That’s understandable given the sheer scope and complex, nuanced interactions that affect each business decision’s outcome. The prescriptive analytic system provides a “better guess than you can make” but not necessarily an “ideal” solution and, like GPS navigation, should be taken into account among a host of other factors rather than blindly accepted as gospel.

Only about three per cent of companies worldwide use the method, according to a recent Gardner Research survey. But the rate of adoption is quickly growing and not just among software companies. As Basu explains:

One massive area where we are working right now is energy independence and over the next few years the US won’t need to import oil and gas because there is so much to be discovered and it will just change the whole equation.

What do oil and gas guys care about? They care about finding oil and gas, producing it, and bringing it in the most efficient and environmentally effective way possible. When you find oil and gas they only get 7-10 per cent of what they find in the ground.

One of the things we are working on are electric submersible pumps is what they use underground to get the oil out from the subsurface to the ground. So when those pumps fail, they can’t get the oil out, production stops, the production forecast affects the market, and stocks go down. What we are doing is predicting those pumps failing and prescribing how to avoid any production loss from anticipated pump failure. When they will fail, why they will fail, which will fail, and what they should do to mitigate the production loss. You are looking at the data from the pump, every component of the pump and you look at the reservoir, you are looking at data of the fluid, water, oil, gas, sand going through the pump. They have to put sensors on everything.

And while virtually all prescriptive analytic applications are geared towards industry rather than consumers (“the value of the decisions are much higher,” says Basu), that won’t be the case for long. Google’s self driving cars, for example, use a form of prescriptive anayltics. The vehicles take all these various data inputs — the LIDAR, the video data, audio data, GPS, everything that the sensor suite is collecting — continually analyses the data stream, and reacts accordingly. The left turn it took to get on this street occurred under completely different circumstances than the right it will have to make two blocks from now but the digital brain that’s driving will have to anticipate and adapt to the upcoming turn.

A self-driving car is a specific example because unlike Ayata, Google controls the hardware as well as the software and data, so the system analyses, prescribes and acts on the data but the prescriptive system has proven essential to the process. Prescriptive analytics could even theoretically grow the basis for future robot learning and even perhaps self-supporting digital life. “That’s the whole idea,” Basu said. “For it to become an adaptive living organism over time.” Skynet will totally be a thing yet. [Ayata, Information Week, Wikipedia, SAS]

Pictures: wavebreakmedia/Shutterstock, Ayata