We use cookies and other similar technologies (Cookies) to enhance your experience and to provide you with relevant content and ads. By using our website, you are agreeing to the use of Cookies. You can change your settings at any time. Cookie Policy.


Predictive Analytics: Providing the Foresight to Act Preemptively

Originally published June 16, 2010

To stay ahead of the pack, a race car driver must focus on what’s coming ahead while being aware of what’s behind him. But if he spends too much time looking in the rear view mirror, he’s likely to drive off a cliff.

The same can be said of a company’s approach to analytics: while traditional analytics provide important knowledge based on the past, predictive analytics gives companies foresight to act preemptively.

With predictive analytics, companies can now analyze their entire data in real time, making the process more efficient than ever. This expansion of available knowledge has the potential to change business strategies in any field, allowing companies to minimize risk, streamline costs and retain customers like never before.

"Predictive analytics is clearly much more than the latest buzzword," says Dr. Raj Nathan, executive vice president and chief marketing officer of Worldwide Marketing and Business Solutions Operations for Sybase, a company focused on in-database predictive analytics. Joydeep Das, a director of Analytics Product Management for Sybase, adds, "We see its potential to become pervasive in most firms aspiring to stay ahead of the curve."

Previously, companies have relied on business intelligence (BI) analytics to provide insights based on recorded data from the past. But unlike these common, traditional analytics models, predictive analytics pervasively allows for foresight by effectively simulating potential outcomes before they happen. This proactive approach, say Nathan and Das, can allow for more strategic freedom than a reactive one.

"Predictive analytics is all about the future – what is next or what might happen," they say, while the traditional BI analytics reports “what happened, what was the most optimal strategy so far.” That means predictive analytics sheds light on the future implications of data while common analytics models can only report what that data meant at a fixed point in time.

That's why companies are increasingly turning to in-database analytics as a way around this logistical hurdle, implementing advanced algorithms and model development within the data repositories themselves. By cutting out this inefficient process, in-database analytics has made the software more practical than ever.

The potential applications of this software are countless, from detecting credit card fraud to test marketing a new product roll-out. Already, say Nathan and Das, American Airlines has used predictive analytics to reduce ticket fraud, detecting irregularities among its countless daily transactions in real time. Playphone, a telecom company based in the UK, recently used predictive analytics to ensure the profitability of value-added offers for key segments of their market.

"Predictive analytics has proven itself in the business world," say Nathan and Das. "Several research studies have indicated that top performing companies that effectively utilize predictive analytics experienced significantly higher profit margins and customer retention than their peers.”

While these applications are significant, there remains significant room for growth within the predictive analytics field. These emerging possibilities for the software, say Nathan and Das, include the possibility for real time analysis of data that may persist on both memory and disk as well as moving decision outcomes and alerts in the backend servers onto mobile devices for prompt response.

Predictive analytics, they say, "…will likely become the next must-have competency for those determined to thrive whatever the business climate."

However, predictive analytics presents a new set of challenges to companies –
chief among them is how to move massive data sets into the traditional analytical systems, which can be a hugely inefficient process. In fact, "moving the data to the logic," say Nathan and Das, "actually wastes a great deal of the analyst’s time, up to 80% by some estimates," limiting both the scope and accuracy of the software's predictions.

Sybase Viewpoints

Cyrus Golkar, the BeyeNETWORK's cloud computing expert, asked Dr. Raj Nathan and Joydeep Das for comments on predictive analytics, in-database analytics and Sybase's capabilities in these areas.The following are their responses in Q&A style.

Raj Nathan Bio Q. What is predictive analytics?

A. Predictive analytics is a technique that gives businesses an edge to understand, explore, validate, prepare, and take the optimal course of action for more favorable outcomes. It involves an analytical approach to examining data assets through a future-facing lens to predict trends and behavior patterns that will determine future business results such as revenue, profitability, customer retention, and market share, to name a few.
 
Technically, for something to be predictive, it has to have the capability of making some assertions about what is likely to happen in the future. This requires some insight into causality that involves mapping of business problems to prediction variables based on appropriate statistical techniques.
 
Q. How does predictive analytics differ from traditional business intelligence (BI) and reporting?

A. At a high level, both predictive analytics and traditional BI derive decision support information from large data sets using analytical techniques. However, the emphasis is different. Predictive analytics is primarily focused on predictions of outcomes of business scenarios and tends to be applied in real time.
 
Predictive analytics is all about the future – what is next or what might happen. In contrast, traditional BI and reporting provides a rear-view mirror analysis of data points – what happened, what was the most optimal strategy so far.
 
Predictive analytics enables a proactive course of actions that can lead to the most optimal outcomes. Traditional BI and reporting, on the other hand, enable a reactive course of actions that do improve outcomes, but at a later point in time and not so optimally.
 
Predictive analytics relies on advanced algorithms based on descriptive statistics and machine learning to explore the data for predictions. Traditional BI and reporting rely on a different set of algorithms that yield aggregates and summarizations along key attributes.Joydeep Das Bio
 
In general, predictive analytics is used as a strategic competitive weapon to stay ahead, whereas BI and reporting are used to improve efficiencies.
 
Q. Please explain the foresight enabled by predictive analytics and how an organization can benefit from predictive analytics.

A. Predictive analytics by definition is about gaining foresight about the future, but the future can be very near or quite distant. Usually foresights gained around predictions that are near term are quite reliable (and valuable) in making decisions and acting on them. For example, abnormal credit card purchase behavior may alert the card issuers of potential fraud that may be immediately investigated and/or prevented. Foresights for longer term predictions have significant value as well, but they are less reliable because the underlying data and facts tend to change. For example, forecasts around multi-year economic trends are useful for planning but are not necessarily accurate for precise decision making.
 
Overall, though, predictive analytics has proven itself in the business world. Recent research from several research studies has indicated that top performing companies that effectively utilize predictive analytics experienced significantly higher profit margins and customer retention metrics relative to their industry peers. The practical benefit of predictive analytics is that it makes businesses more competitive in the market through incremental improvement over time—such as iterative improvements to success rates of fraud prevention and accuracy of risk management models, to name a few.
 
Q. Can you provide a few industry examples of how Sybase customers are benefiting from predictive analytics?

A. Sybase customers have enjoyed significant competitive advantages by employing predictive analytics best practices. A few industry examples would be:

Media & Telecom: Firms in this industry often employ predictive analytics to determine value-added offers to extend to targeted customers to ensure profitability. PlayPhone, a media company in the UK, uses predictive analytics techniques on Sybase IQ, the world’s leading column-based DBMS platform, to obtain deeper understanding of customers’ interests, preferences, and behaviors to tailor marketing campaigns that consistently produce high quality results.

Transportation: Uncovering fraud and ticketing errors can save millions in this highly transaction intensive industry. American Airlines uses predictive techniques combined with deep and powerful ad hoc inquiry capabilities of Sybase IQ to uncover and trap fraudulent ticket processing on a continuous basis – saving the firm millions of dollars.
 
What’s more, Sybase technology is being employed successfully for predictive analytics in hundreds of organizations in a wide variety of verticals such as financial institutions, information providers, public sector, and insurance.
 
Q. What is Sybase’s vision for predictive analytics and how will Sybase IQ evolve with predictive analytics?

A. Predictive analytics is clearly much more than the latest buzzword. We see its potential to become pervasive in most firms aspiring to stay ahead of the curve, and it will likely become the next must-have competency for those determined to thrive whatever the business climate. Nevertheless, pervasiveness entails embedding of predictive analytics at various data management points: real-time streaming data, near-real-time transactional data, historical data, and unstructured data such as data in textual format.
 
But in a world awash with data, the hard questions that businesses first need to understand are the types of business problems that may benefit from a predictive analytics approach. Firms must review the applicability of predictive analytics to specific problems in their enterprises, and evaluate the technologies they will need to roll out an effective and reliable predictive analytics framework.

Sybase has been at the forefront of providing high performance analytics platforms on which innovative predictive analytics applications have been built. The future trends clearly point to a need for more accuracy, lower latency, large and multi-format data sets, complex, and ad hoc queries. Sybase views predictive analytics technologies that focus on real-time analysis of data that may persist on both memory and disk along with complex event processing (CEP) for streaming data as most valuable. Finally, we also see decision outcomes and alerts from predictive analytics in the backend servers being pushed to mobile devices for prompt response.

We are putting in significant investments in R&D and partnership fronts to ensure that Sybase analytical technologies are able to provide businesses the most reliable, high performance, and cost-effective platform for predictive analytics.

Q. What is in-database analytics and why is it important?

A. Advanced analytics techniques such as predictive analytics involve complex, resource intensive computations. Invariably, this leads to more practical questions such as:
  • How to deal with large data sets
  • How to improve the accuracy of predictive models
  • How to achieve the performance required for timely predictions
Confronted with these questions, the attention turns to the data management repository for the data set(s) used to build and score the predictive analytics models. These tend to be high performance database servers.

But the fact remains, the data to be processed for building and scoring models, need to be shipped out of these repositories into specialized tools for assembling the data so that the algorithms for the predictive models available in these tools can be applied – in essence, moving the data to the logic. This process actually wastes a great amount of the analyst’s time, up to 80% by some estimates. Clearly, in the context of the three questions raised above, this archaic approach is a major hurdle to effective predictive analytics deployment and consumption.

In contrast, in-database analytics, a relatively recent innovation, embeds the predictive analytics algorithms and model development and scoring within the analytics data repositories itself to increase performance, latency, and security. This completely bypasses the problematic data movement approach outlined above by moving logic to data instead. The following figure contrasts the two approaches.    


Q. What is Sybase doing with in-database analytics?

A. Today, Sybase delivers in-database analytics in several ways. Built into Sybase IQ-based platforms are the following:
  • An extensive library of built-in numerical and analytical functions that greatly expand the analytical capability resident in Sybase IQ.
  • Latest built-in ANSI SQL Standards-based OLAP extensions allow aggregation analysis on large data sets, yielding quick results for iterative computations such as correlation and covariance over shifting data sets. 
  • Partner-friendly API against which certified, pluggable high performance analytical algorithms from statistical and data mining software partners (e.g., Fuzzy Logix DBLytix) that bring in the processing of highly sophisticated computations such as Monte Carlo simulations, K-means clustering, naïve Bayes classification, to name a few, into the Sybase IQ kernel.
  • Analytical library of advanced time series functions ranging from weighted moving averages to auto regressive moving averages (AUTO_ARIMA), outlier detection, etc.
Built on Sybase IQ’s industry leading and patented column store technology, these capabilities enable a new generation of analytical processing that is especially well-suited for predictive analytics scenarios.
 
In summary, Sybase in-database analytics delivers immediate performance and scalability improvements. Using such a model, data never leaves the database until results are materialized. The analytics code and models are shareable across an organization and allow ad hoc analysis, plus they are applicable to the most current data sets. By keeping the data in-database, a higher level of data security is ensured. Moreover, access to the analytics is standards-based, and the in-database logic is extensible by anyone with a good working knowledge of SQL.

Given the onus on organizations to show value through analytics on large data sets, we are getting a lot of interest from firms in our in-database analytics, offering to help them overcome the various hurdles discussed earlier.
 
Q. What is Sybase’s vision with analytics in the cloud? What are the implications and requirements?

A. Sybase views analytics in the cloud from a very practical angle – the cloud-based technology should be as robust and valuable as the on-premises alternatives. With that in perspective, we are observing significant traction in the following areas:
  • Sybase serves a large segment of the information provider industry where the provider firms offer various forms of value-added analytics services on industry aggregated data. Sybase IQ’s high performance column store architecture combined with its grid-based deployment option enables information providers to offer multi-tenancy based analytics via a cloud interface to a large base of clients who can concurrently conduct self-service ad hoc reporting and analytics. This cloud-based analytics interface powered by Sybase IQ is enabling a paradigm shift since it allows clients access to valuable analytics 24x7 without the overhead of investment in an in-house infrastructure.
  • Sybase has also announced a hosted/cloud based analytics service for mobile messaging traffic analysis called Sybase Operator Analytics 365, based on Sybase IQ. Operator Analytics 365 provides a wide range of reports including message traffic, network utilization, performance indicators, quality of service testing and customer analysis. Sybase Operator Analytics 365 puts business intelligence capabilities into the hands of telco operators, giving them insight into their network data and subscriber behavior, something that has never been possible before. This turnkey solution gives mobile operators unparalleled features and performance at dramatically lower costs than on-premises solutions.

  • Cyrus GolkarCyrus Golkar
    Cyrus Golkar is an information technology executive with a unique combination of entrepreneurial, business and technical expertise, and more than 20 years of experience in enterprise software, database, analytics, web services and cloud computing.

    Previously, Cyrus led Sun Microsystems (now Oracle) $2 billion database, business intelligence and data warehousing business for 9 years, recommended the acquisition of and investment in database companies, and led the development of Sun's data warehouse appliance.
    He is also the co-founder of several companies, including 3Bubbles
    (acquired) with the creators of Jabber, the open standard for web-based instant messaging.

    At Siemens/Pyramid, Cyrus was recognized by the CEO for the creation of "Smart Warehouse" an industry leading data warehousing business that generated $200 million in annual revenue. He also held leadership and technical positions at Open Text and GE.




Recent articles by Cyrus Golkar

 

Comments

Want to post a comment? Login or become a member today!

Be the first to comment!