News / Features



Four ways big data can benefit your business

Three industries, three customers, four big uses of high-performance analytics

By Gary Spakes, SAS

At the simplest level, advanced analytics allows you to develop models and then use them to ask "what if?" questions about your data. For example, developing a statistical model that associates buying behavior with customer profiles can then be applied to future behavior of customers. The application of that model is referred to as "scoring" and is the basis for predictive analytics.

That type of analysis is worlds away from traditional business intelligence, which is more about asking simple questions about data in one or two dimensions (e.g., How many shoes of Brand X do we have in stock?). That kind of analysis is fairly straightforward using a traditional database, needing only a small pipe to get the data in and out and a software component on the client to manage the interface.

Combining big data with predictive analytics can be a challenge for many industries, but high-performance analytics, which speeds the process of scoring and reporting, is helping SAS customers in many areas. Here, we explore four of those areas:

  1. Detect, prevent and remediate financial fraud.
  2. Calculate risk on large portfolios.
  3. Execute high-value marketing campaigns.
  4. Improve delinquent collections.

Detect, prevent and remediate financial fraud
Across consumer and B2B industries, it's an unpleasant fact of life: Every day around the world, criminals are busily at work trying to defraud companies through a constantly evolving portfolio of schemes and strategies. As the volume and sophistication of these schemes continues to increase, many organizations are turning to powerful analytics to sift through massive data volumes and uncover hidden patterns, trends and suspicious events that can indicate criminal fraud.

In most instances, fraud detection involves analyzing the various attributes of transactions and making a determination about whether those orders should be flagged for further review. But as volumes increase, the thresholds for intervention or review increase, meaning there's a greater likelihood of fraudulent transactions eluding scrutiny.

For one SAS customer – a major global manufacturer of consumer electronics –- the challenge of stemming losses attributable to fraud had overwhelmed its previous infrastructure, and a new approach was needed to scale up to the increasing demands of fighting fraud. Although the company had numerous server instances of SAS for a wide variety of reporting and analysis, there was only one server dedicated to fraud analytics. As a result, the fraud-fighting initiative was hampered by limits in compute resources, storage, and overall performance as well as a limited ability to refine the predictive abilities of their fraud models.

The project team laid out a plan and process for transitioning from a series of disconnected departmental data marts to a true enterprise-scale analytic data warehouse. Thanks to SAS software's in-database processing capabilities, instead of bringing the data to the application, the IT organization can bring the application to the database and perform the analytics and statistical functions within the database itself. This means the models run operationally across all of the company's customers, as opposed to extracting data from the environment, processing the data, and importing the results back into the warehouse.

Using SAS Grid Computing, the manufacturer's fraud and IT teams were able to more efficiently collaborate on their efforts to improve their fraud-detection models. That enables them to keep up with fraudsters and their evolving schemes, achieve a better "model lift" – the ability to trap more fraudulent transactions correctly – and scan more transactions, which ultimately leads to preventing more instances of fraud before they can have a financial impact on the company.

Calculate risk on a large portfolio of loans
In the past few years, it's been anything but smooth sailing for financial services firms that have struggled to effectively manage their extensive consumer homeloan portfolios. An industry-wide failure to properly assess the latent risks lurking in thousands of substandard loans led to billions of dollars of losses, thanks to record-setting writedowns.

For one major SAS customer in the financial services market, one of the root causes of its unacceptable risk exposure was simply an inability to efficiently create models and run those models against its growing data volumes. The institution was capturing data at a rate that was far faster than its ability to compute that data in a timely fashion. Literally, the company's modeling team couldn't work fast enough to meet the demand for new and refined models.

The company faced unacceptable risks and unacceptable processing times in trying to create and run models to minimize that risk. Seeking a better way, the firm pursued and deployed a new paradigm for its analytical processing: High-performance analytics.

The performance improvements were extraordinary. Instead of waiting a week to execute a new model and assess the risk contained in the portfolio of consumer mortgages, the firm can now generate the same results in only 84 seconds. This gives time and motivation to analysts to iterate their models many more times than before. Since the firm is managing a loan portfolio of billions of dollars, even a modest improvement translates into savings in the tens of millions of dollars.

Execute high-value marketing campaigns
The same financial services company faced similar big data challenges in its marketing operations as well. Financial services segments are increasingly competitive as institutions seek to offset the loss of fee-income and minimize their own churn. This company operates a sophisticated marketing operation, running campaigns to millions of targets. However, as the data volumes grew and the campaigns began to target 10-15 million recipients, it couldn't physically process the data, preventing them from maximizing their customer lifetime value and executing more efficient and effective cross-sell/up-sell campaigns.

Using high-performance analytics, the institution has achieved tremendous gains in the throughput of its database marketing – as much as 215 times faster – dramatically compressing the modeldevelopment lifecycle and enabling their teams to test and validate additional variables for greater reliability in their models. High-performance analytics removes the limits on observations and variables that the company can process. That opened up the scope of questions to ask and avenues to pursue.

The result is that the team's productivity in executing the models for the campaigns was dramatically improved. More importantly, the effectiveness and predictive reliability of the marketing models improved as well. Given the volume of data, even small incremental improvements create significant savings. For instance, a typical direct mail campaign usually generates a 1 percent response rate. When sending to 15 million prospects with a lifetime customer value of $500, a fractional improvement in response rates for cross-sell or up-sell offers quickly translates into tens of millions of dollars annually in top-line revenue.

Improve delinquent collections
Although prepaid phone services continue to gain in popularity, in the US mobile telecom market, services are largely delivered on a post-pay basis, which essentially means that providers are extending credit to tens of millions of customers with widely varying credit histories and propensities to pay. Not surprisingly, that leads to significant problems with delinquencies that result in tens of millions of dollars of uncollectible bad debt – a significant drag on the industry's financial performance and a problem that only increases as higherdollar smartphone plans expand their share of the market.

For one major carrier, high-performance analytics from SAS has transformed the company's delinquent collections process to increase collections rates, decrease churn, and increase lift from "next-best offers" presented in real time to at-risk customers.

Previously, the company employed a traditional application built with SAS to load and score a subset of delinquent customers carrying overdue balances of more than $500. This process involved aggregating records for more than 40 million customers locked in more than 300 separate data sources, applying some rudimentary segmentation to trim the candidates to 350,000 names, and then scoring and ranking those 350,000 at-risk customers based on a SAS model that predicted the delinquent customer's propensity to pay.

Without question, the process was effective, generating $7 million to $10 million each month in recouped revenue. However, it was time-intensive, requiring three hours each night to run the aggregation and another 30 minutes to score and rank/sort the data. Operating in an industry where every financial advantage has an outsized impact, the company wanted ways to optimize the process even further.

That's when the carrier turned to highperformance analytics to bring the application to the data instead of the other way around. The goals were to predict probable non-payment and delinquency by getting new customer models into production in seconds or minutes, provide call-center reps with queues and prioritized alerts based on customer scores that incorporate multiple metrics (credit history, employment status, payment history, and other factors) across all 40 million accounts, and enable those agents to offer better rates and repayment packages in real time, while the customer is still on the phone or on the website.

The IT strategy called for running SAS high-performance analytics doing realtime loads into a single Teradata enterprise data warehouse to support more frequent and faster modeling and scoring so that the best candidates are called at the right time and receive the best offers. The SAS team worked with the customer to create an entirely new analytical model in less than four hours. And by "taking the analytics to the data," the company eliminated hundreds of ETL steps, creating additional performance improvements.

That new model delivered a 13 percent incremental lift over the company's previous model, translating into an additional $900,000 to $1.4 million in recouped bad debt each month on a comparable basis. The ability to analyze the company's entire 40 million customer records (in less than four minutes) creates substantial added lift because the company can assess customers faster to prevent wasted efforts on high-risk customers who are unlikely to pay.

Armed with this information, call center agents are able to focus on delinquent accounts with the greatest propensity to pay. Furthermore, they can access a customer-specific list of "next-best offers" to propose to the customer to prevent churn and lost revenue.

High-performance analytics makes the difference
How do you want to solve your biggest big data problems? If you want to focus on the analytic basis, like these customers, as opposed to a transactional basis, you need high-performance analytics. What could you do differently if your processing times for fraud, risk, marketing and collections were reduced by 90 percent? Those are big questions worth exploring, and SAS has the architecture to support the answers.

Bio: Gary Spakes focuses on both the technical aspects and business implications involving SAS High Performance Analytics. An IT veteran, Gary leverages his technology background to help organizations alleviate immediate pains, but not compromise long-term initiatives.

Gary Spakes, SAS

Read More