Why machine learning is the new BI

20.04.2016
Business intelligence has gone from static reports that tell you what happened, to interactive dashboards where you can drill into information to try and understand why it happened. New big data sources, including Internet of Things (IoT) devices, are pushing businesses from those reactive analytics – whether you look back once a month to spot trends or once a day to check for problems – to proactive analytics that give you alerts and real-time dashboards. That makes better use of operational data, which is more useful while it’s still current, before conditions change.

“There’s a demand for real-time dashboards,” says Herain Oberoi from Microsoft’s Cortana Analytics team. “A lot of businesses want to get the pulse of their business. But dashboards show things that have already happened.”

That’s why fastest growing area is predictive and other advanced analytics, according to Gartner. Its latest Magic Quadrant for advanced analytics predicts that by 2018 more than half of all large organizations around the world will use advanced analytics (and algorithms built on them) to compete.

Advanced, predictive analytics are about calculating trends and future possibilities, predicting potential outcomes and making recommendations. That goes beyond the queries and reports in familiar BI tools like SQL Server Reporting Services, Business Objects and Tableau, to more sophisticated methods like statistics, descriptive and predictive data mining, machine learning, simulation and optimization that look for trends and patterns in the data, which is often a mix of structured and unstructured.

IMG1

They’re the kind of tools that are currently used by marketing or risk analysis teams for understanding churn, customer lifetimes, cross-selling opportunities, likelihood of buying, credit scoring and fraud detection. Those users aren’t going away. “Many telcos want to get from being reactive to being proactive,” says Oberoi. “They want a system where they can say ‘tell me which of these customers, based on their customer profile and calling pattern, is going to churn’.”

[Related: How machine learning ate Microsoft]

But Gartner says almost every business unit is going to be interested in these tools and he agrees that matches the customers for Cortana Analytics. “The people we’re talking to are changing. I’m having a lot more dialog with line of business decision makers. We see a lot of budget moving to line of business teams.” He recently spoke with customers at Microsoft’s Convergence conference in Europe. “Of the five customers I met, at least three were line of business and two of those had a charter for driving digital transformation and their company’s innovation agenda.”

Predictive maintenance has got a lot of attention but there are other key uses like predicting demand and finding problems in service or product quality using anomaly detection, as well as decision support systems. Those are questions like “what might happen” and “what should I do” says Oberoi.

“The fundamental customer challenge hasn't changed; how do I go from the data that I have, to getting some insights, to actually enabling some an action or driving something forward” What the tools in Cortana Analytics can do is reduce the number of manual steps you have to take to get to those actions.

“The basic way to go from data to decisions is a static report that says what happened. If I'm in sales and I want to look at what my sales look like by region for the last quarter, that's my static report, and then there's some manual steps I take before I go make a decision,” he explains.

“The next piece of it is not just something happened but why it happened. My sales went down last quarter, but did they go down because my top three deals didn't go through or because my average deal size went down How do I slice and dice the data [to find out] So I have a dashboard with reports I can interact with to understand why something happens, and that typically reduces the numbers of manual steps before I can make a decision and take some actions. Then we get into prediction; not just after-the-fact ‘my sales went down and I know why’ but tell me beforehand, based on my forecast, that I might not make my target for the month so I can respond.”

Business automation would mean the fewest possible manual steps. Oberoi explains what that might look like.

“The final piece of it is recommendation and decision automation. Ultimately, you want to get to a place where the system is proactively informing you, not just what might happen but what you can do about it. ‘It looks like you're going to miss your forecast next week. Based on that, you have two promotions lined up and those are hooked up to your CRM system; do you want to pull those promotions forward by a week, yes or no’ You say yes, and you have the business process and the workflow set up, fully automated, and it helps you pull that forward by a week.”

That kind of advice will often come from an intelligent assistant; in Microsoft’s case, that’s Cortana, who can already take reminders and suggest when it’s time to leave for a meeting, as well as answering questions. Oberoi thinks that’s just as useful for business tasks where a reminder could tell employees they need to send in their expense report and those questions can be about your business; “what were the biggest deals we closed last quarter” or “which of our customers are most likely to churn in the next quarter” or “alert me if this customer ever has a 90 percent chance of churn in the next 30 days.”

“The system knows you're going to miss your forecast next week; it also knows that one of the things that might help you hit the forecast is pulling forward the marketing promotion that's being sent out. There's no reason why Cortana wouldn't proactively reach out to you and say ‘hey George, it looks like you're going to miss your forecast for sales next week and you've got these promotions lined up. Would you like to pull them forward, or talk to the team that owns those promotions to get them to pull them forward It's a more proactive way for the system to interact with you and a more natural way to do it.”

Whether it’s IoT, big data or analytics, companies have a lot more data to base their decisions on, and data-driven decision making sounds obvious. And the next step beyond data-driven decisions is decision support systems and even automation. Are we ready for intelligent assistants with business advice

While a recent study of 50,000 American manufacturing organizations found that the use of data-driven decisions had almost tripled between 2005 and 2010, that was still only 30 percent of plants. And when telecom provider Colt surveyed senior IT leaders in Europe in 2015, 71 percent of them said intuition and personal experience works better for making decisions than using data (even though 76 percent of them say their intuition doesn’t always match the data they get).

More positively, Avanade’s new study of smart technologies says business leaders globally expect to be using digital assistants and automated intelligence for problem solving, analysing data, collaborating and making decisions – and they also expect them to increase revenues by more than a third. With those kinds of expectations, attitudes are also more positive; 54 percent said they’d be happy working with those systems.

[Related: Why Power BI is the future of Excel]

Certainly, early adopters that Accenture has spoken to who are using machine learning to improve the way they manage customer service, financial resources and risk and compliance, in sales and marketing and in developing new areas of business found “significant, even exponential, business gains” in costs, revenue and customer performance, by using a mix of what Oberoi calls “perceptual intelligence” using natural language and voice biometrics, advanced analytics and business decision support .

Those gains included cutting costs up to 70 percent, increasing revenue up to 20 times by tracking buyer behaviour more quickly and getting happier customers by handling call routing more quickly and accurately, and those results might help overcome reluctance.

Getting business users involved in building these systems should also increase adoption. The demand for data scientists is larger than supply, which means companies without their own deep expertise will look to cloud services and marketplaces for analytical solutions and algorithms like the Cortana Analytics gallery, which includes tutorials and experiments as well as APIs and templates you can use to get started and then customize, plus pre-configured services for recommendations and forecasting.

“The way to best succeed in this space in to experiment fast,” says Oberoi; “to go through the search space of ideas and find things that are really interesting.” Marketplaces are a great way to get started on those experiments.

Power users (who Gartner calls “citizen data scientists”) will also pick up these tools to create their own advanced analytics, which means you’ll want a strategy for addressing analytics. You can think of them as far more sophisticated versions of the Excel macros that many business departments rely on, that are going to demand more data literacy and careful thought about the ethics of automated decisions, because of the significant impact this new, more intelligent business intelligence is going to have on your organization.

(www.cio.com)

Mary Branscombe