For a long time, the use of data was essentially a matter of recording the past.
We had 40 years of the database management system and 20 years of data warehousing and business intelligence.
In the last 10 years, the development of event processing has enabled businesses to focus on the present, reacting to information immediately. In the financial markets, for example, it is possible to spot inappropriate trading as it happens.
From dealing with the present, we moved on to facing the future, analyzing historical data to reveal significant patterns. Now, we have reached the point at which we can advance to the next level of capability in analytics: that is, to combine the power of prediction with our immediate understanding of what is happening this instant.
Take the example of a telecommunications company with 30 million subscribers. It has built a predictive model that uses information on how users are accessing the network right now, whether it is making a phone call, accessing a website, or using particular apps.
The company combines the power of prediction and event processing to send offers at the most appropriate time. The model it has built will notice, for example, that a customer regularly makes international calls and is close to upgrading to a more suitable plan. Instead of sending that customer an offer when he or she is likely to be busy with other matters, the company sends the offer just after the customer has ended an overseas call, making the offer far more relevant and timely.
Sending the offer at the right time can be very powerful. Take, for example, a business traveler who is regularly at Heathrow Terminal 5. In this example, the traveler may have been shopping online with a retailer or has an account with the retailer that includes handing over a mobile phone number. The traveler also may have given the retailer permission to share her location with trusted partners.
Then as she walks past the store in the airport mall, she receives an offer for a product she was looking at before she left the office, informing her that it is in-store and that if she buys it within the next 20 minutes, she will receive a 10 percent discount. A slick operation of this nature is about the ability to interact immediately with customers.
The same kind of data processing and analysis that is behind this interaction already is deployed in manufacturing. Where companies rely on their own generators, for instance, maintenance is highly disruptive, taking as much as three months for the overhaul of each engine.
However, by using data from sensors monitoring cylinder pressure, temperature, or fuel consumption, it is possible to indicate when individual parts are at the point of failure. Using streaming analytics, the model will pick up a drop in pressure and, by correlating with records, predict that the current rate of wear will lead to malfunction within, say, four weeks.
Equally, in capital markets, the capacity to analyze vast streams of data from trading operations and other unstructured, external sources such as news feeds, chat rooms, and employment records will expose inappropriate activity when matched against norms established from historical records. The same techniques also will detect when automated trading algorithms deviate without cause, permitting interventions to prevent events such as the infamous Knight Capital loss of $440 million in 30 minutes, when its algorithms bought at ask prices and sold at bid prices.
In consumer finance, predictive analytics is allowing banks to detect credit card fraud as it happens and to take immediate action by blocking transactions rather than becoming aware of the crime after the event.
Train companies can use predictive and streaming analytics to optimize use of their track infrastructure. They can ensure that passengers are directed to the most suitable station platforms so that delays are minimized.
Further applications are set to occur in utility companies, where predictive analytics will allow them to operate their grids more efficiently in the age of renewable energy. As more residences and businesses generate electricity from solar installations, it remains difficult to match production and demand. Using predictive analytics in combination with live data, it is possible to smooth out the peaks and troughs, reducing waste.
The organization that knows most about us is our bank. If banks can persuade us to let them use our data, they could predict what we might want to buy, the services they could provide, and much more.
But perhaps some of the most significant impacts will be in healthcare and medicine, where the advanced use of predictive analytics is set to change many practices. For example, in cases of patients with chronic conditions who wish to maintain independence by living in their homes, their movements can be tracked via a smart badge or wristband so that if they depart from their normal pattern, remote caretakers can be notified to check on them.
In an acute-care hospital ward, sensors that measure a patient’s vital signs, such as heartbeat and blood pressure, can feed data into a predictive model that will alert staff to the imminent danger of a bad event such as a heart attack. The model achieves its insights by matching current information against the corpus of data about patient reactions in the past.
So while the deployment of predictive analytics is already having a major impact, its effect is only just beginning.
For the consumer, handing over data in this way may require a constant process of negotiation with brands and third parties. It will be a question of exchanging data in return for better service. As this data market develops, it may be the kind of blockchain technology behind Bitcoin that emerges as a convincing protection mechanism.
However, provided that organizations put effort into building trust and security, they should be able to accumulate ever greater volumes of consumer, instrument, or trading data, with which they can build the predictive models that bring major gains in revenue and efficiency.
Giles Nelson is Senior Vice President, Product Strategy and Marketing, at Software AG.
Subscribe to Data Informed for the latest information and news on big data and analytics for the enterprise.