Data and analytics are reshaping organizations and business processes, giving organizations the capability to interrogate internal and external data to better understand their customers and drive transformative efficiencies.
Worldwide revenues for big data and business analytics clocked in at nearly $122 billion in 2015 and will grow to $187 billion in 2019, according to a five-year forecast from research firm IDC.
“Organizations able to take advantage of the new generation of business analytics solutions can leverage digital transformation to adapt to disruptive changes and create competitive differentiation in their markets,” said IDC analyst Dan Vesset in a statement issued in conjunction with the release of IDC’s Worldwide Semiannual Big Data and Analytics Spending Guide earlier this year. “These organizations don’t just automate existing processes — they treat data as they would any valued asset by using a focused approach to extracting and developing the value of information.”
Additionally, a recent Forrester Research study, commissioned by the global data and analytics team at KPMG, found that 50 percent of businesses now use data and analytics tools to analyze their existing customers, while 48 percent use them to find new customers and 47 percent use them to develop new products and services.
The picture isn’t entirely rosy, however. That same Forrester study found that many organizations are struggling to adjust their cultures to a world in which data and analytics play a central role, and many business executives mistrust the insights generated by data and analytics.
Other organizations, however, have taken naturally to data and analytics and are using new tools to better understand customers, develop new products and optimize business processes.
To honor those organizations, CIO.com and Drexel University’s LeBow College of Business recently announced the first Analytics 50 awards. The winners represent a broad spectrum of industries, from pharmaceuticals and healthcare to sports and media.
In the profiles that follow, five of the winners explain how their projects are delivering measurable outcomes and offer advice to other IT leaders who are planning analytics initiatives.
Children’s Hospital of Philadelphia: Detecting and preventing venous thromboembolism
Children’s Hospital of Philadelphia (CHOP) has been on a mission to use data and advanced analytics to improve the quality of its care and patient outcomes. To that end, it has launched an initiative to improve detection of venous thromboembolism (VTE) by using text analytics tools to glean insights from unstructured data in physicians’ reports.
“We’ve actually been executing a road map and strategy that we started in 2008,” says John Martin, senior director of enterprise analytics at CHOP. “It started pretty typically with ‘Let’s build a data warehouse based on use cases, with a long-term vision of precision medicine and analytics.’ We started with nothing. We had to build up to it.”
VTE is a condition that involves the formation of blood clots within a deep vein (deep vein thrombosis) that break loose and travel to the lungs (pulmonary embolism).
According to the U.S. Department of Health and Human Services, there are about 350,000 to 600,000 new cases of VTE in the U.S. annually; recurrent cases bring that number up to about 1 million. Nearly two-thirds of the people who experience VTE are hospitalized or were recently hospitalized, and about 300,000 of them die each year.
Children at risk
Martin notes that hospital-acquired VTE is currently the second-most common cause of harm to hospitalized pediatric patients, after central line-associated bloodstream infections. It’s currently the focus of a nationwide prevention campaign. The overall mortality rate associated with pediatric VTE is estimated at 2.2 percent, Martin says. Additionally, pediatric patients diagnosed with hospital-acquired VTE stay in the hospital an average of 8.1 days longer than other children and cost $25,000 more to treat.
As dangerous as VTE is, preventive measures, including early detection, can dramatically reduce the incidence of the malady. And much of the data needed for that sort of prevention can be found in physicians’ notes. The current mechanisms used to identify VTE events depend on manually generated clinical lists and a post-discharge review. Martin says both processes are time-consuming and error-prone and don’t result in immediate detection.
A faster process
To speed up the process, CHOP decided to create a decision support tool for physicians. The hospital applies natural language processing (NLP) to radiologists’ reports, creating a fully automated solution that quickly analyzes complex batches of physician notes and offers a high level of accuracy in identifying and tracking patients with hospital-acquired VTE.
Clinical documentation stored in the electronic health records (EHR) is backed up to a reporting database on a daily basis and then transferred to the CHOP data warehouse.
When the backup is done, identifying information is removed from radiology reports and the reports are transferred to the NLP engine via a secure cloud service. The NLP engine produces results in an XML document that includes both a semantic translation of the notes into discrete data and application of a classification model created by CHOP for deep vein thrombosis. The document then goes to the data warehouse and the patient’s identification is restored.
The data is then converted to Hadoop structured data, where the rules engine assigns the VTE label to each study.
“Technology only does one thing,” Martin says. “It only automates and simplifies things that a human could do — but maybe not as quickly or as accurately. It’s a tool. We were able to apply that tool, that technology, to automate a process that wasn’t previously automated, while increasing its accuracy. Then we can get valuable human time focused on the right cases.”
CHOP’s VTE analytics effort has paid dividends, Martin says. The NLP tool identifies patients with VTE with a high degree of sensitivity and specificity, and it has uncovered VTE sufferers who were overlooked by CHOP’s existing VTE screening process.
The NLP engine is now an important component of CHOP’s VTE prevention improvement efforts, according to Martin, who adds that his team is exploring other ways to use NLP applications and hopes to develop methodologies that can be adopted at other healthcare institutions.
Intel: Mastering supply chain analytics
Getting Intel’s chips and other products to market is a highly complex affair. The company’s supply chain is a capital-intensive global network that requires many specialized materials and complex manufacturing processes with long lead times and short product life cycles. The semiconductor giant has developed advanced supply chain analytics and saved millions in the process, says Mani Janakiram, Intel’s director of supply chain strategy and analytics.
“Intel, by nature of being the leading semiconductor-producing firm, is a capital-intensive and high fixed-asset-based company, and our capital expenditures attain a level of approximately $10 billion per year,” Janakiram says. “Critical capital equipment used in our factories may cost anywhere from $30 million to $100 million or more per tool. And a new semiconductor plant can cost upwards of $4 billion.”
Developing and mastering the analytical techniques for forecasting, planning and aligning cross-functional supply chain metrics enabled the company to save millions of dollars by, for example, avoiding purchases of capital equipment, reducing inventory levels and identifying opportunities for systemwide optimization, Janakiram adds.
Advanced analytics tools also helped Intel capture millions (and potentially billions) of dollars of revenue through improved customer satisfaction, increased agility and faster time-to-market, Janakiram says.
In many cases, capital planning and contracting has to happen more than two years before Intel starts producing products — well before those products are finalized. Manufacturing lead time is measured in months, Janakiram says, while customers expect changes in their orders to be accommodated in a matter of days.
Intel is a data-driven decision-making company, and analytics play a role in everything it does, Janakiram says. When it realized that its supply chain metrics weren’t well-aligned with the APICS Supply Chain Council’s Supply Chain Operations Reference (SCOR) model, Janakiram and his team turned to advanced analytics and modeling to solve the problem. They tracked, aligned and improved the key “Tier 1” metrics that steered operational excellence in the core business and provided insight into future lines of business.
“We nurtured highly skilled data scientists with an appropriate blend of business and analytics skills,” Janakiram says. “Our data scientists have expertise in operations research, computer science, mathematics, statistics, data mining, finance and business,” and they drew on their combination of business and technical analytical acumen to identify, solve and align the key metrics.
Through those efforts, he adds, the analytics team showed how it gives Intel a competitive advantage “by providing advanced data models to help our supply chain to make better and more effective decisions.”
“We regularly evaluated and employed advances in technology such as big data, cognitive computing, text mining, agent-based modeling and simulation,” he says. “We also partnered with leading universities to apply advanced analytical techniques to our metrics, as well as other complementary supply chain needs, including advanced production planning, supply chain gaming, inventory strategies, procurement and simulation modeling.”
Janakiram says it wasn’t too hard to convince Intel’s executive leadership team that the project was necessary, but that’s not always the case.
“Sometimes it’s not an easy sell,” he says. “In some cases, where the solution is new or evolving, we have to define what it means for the business. We have to show a future value add. We do a proof of concept. We go through that process to get management buy-in.”
And having that buy-in in place is important, he says, because it helps get end users to overcome their resistance to change. With stakeholders and management engaged, key decision-makers and users can participate in the process and get other users on board.
Janakiram has three tips for other executives planning analytics projects:
Engage the right people.
Ask the right questions. You need to learn about users’ pain points and priorities to understand the problem.
Don’t get seduced by the elegance of an analytics system. Instead, focus on how you can improve the experience of your customers and stakeholders with the right analytics and applications.
Do your homework
What Janakiram and his team learned from the project was, first, to “do your homework” so you can understand the problem and, second, to learn from what others have done.
“Look at similar, like-minded companies or groups,” he says, then ask, “What are the things they had to learn that we can fast-track?”
Also, make sure you can build your system piecemeal and earn credibility along the way, he says, adding, “You need to keep feeding the beast to have the feast.”
New Mexico Department of Workforce Solutions: Predicting bogus payments
The New Mexico Department of Workforce Solutions (DWS) has struggled for years with erroneous unemployment insurance (UI) payments. It isn’t alone — government agencies across the country face the same problem. In 2014, more than $4 billion in erroneous payments were made in the United States. The DWS has applied predictive analytics and behavioral science techniques to curb the problem.
In 2014, nearly one dollar out of every eight distributed under UI programs in the U.S. went to someone who was ineligible, says Joy Forehand, deputy cabinet secretary of the DWS. While identity theft and similar criminal schemes have grabbed headlines, they actually account for less than 5 percent of the total cost, Forehand says. In an effort to tackle the other 95 percent of activity that results in improper payments, the DWS set some goals: Enhance program integrity, reduce overpayments without hurting eligible claimants, and increase collection efforts without expanding the collections team.
“We needed to really understand the realities of our improper payments,” says cabinet secretary Celina Bussey. She adds that the department has taken steps to combat criminal fraud schemes, “but, under the surface, there are the core issues that cause the overwhelming majority of improper payments.”
In collaboration with Deloitte Consulting, the DWS found that improper payments are generally the result of claimants doing one or more of the following things: not looking for new jobs, not properly reporting income they earn while collecting benefits and incorrectly reporting the reason for the separation from their employer.
With that data in hand, the agency launched a project that it called the Improper Payment Prevention Initiative (IPPI). Working with Deloitte, the DWS developed a predictive model based on patterns of past overpayments. It identifies individuals at a higher risk for overpayment. Behavioral science and “nudge” techniques are then used to prevent overpayments by reminding claimants to follow the rules.
The department uses messaging, including certification boxes and pop-ups, to remind claimants to review their information for accuracy and completeness at three critical moments: filing the initial application, reporting work and earnings, and making plans to seek new employment.
“We wanted an innovative approach to prevent improper payments from happening in the first place,” Bussey says. “Individuals have to submit required information on a weekly basis in order receive unemployment benefits. We were able to determine who is at a higher risk for reporting inaccurate information. The predictive algorithms were developed and tuned to historical cases of overpayment to isolate situations at the highest risk of overpayment. As a team, we knew that we could possibly prevent improper payments if we nudged the individual to change behavior and provide accurate information upfront.”
Moreover, Bussey adds, “we needed to not only understand the analytics, but then also understand why our customers make certain decisions.” Armed with that data, the agency turned to “the science of behavioral nudges” to encourage claimants to make the right decisions, she says. “We chose to test three types of behavioral nudge techniques: certification boxes, enhanced screens and pop-up messaging.”
To ensure that the combination of predictive analytics and behavioral science would be effective, Bussey says the state set up a randomized trial to test hundreds of combinations of message layouts, wording and more.
The IPPI project launched smoothly in May 2015, and Bussey says claimants who see the reminders are 40 percent less likely to file improper claims. The tools have helped state investigators find 28 percent more overpayments with the same level of staffing. They also detect overpayments an average of eight weeks faster. Agency officials say the approach is expected to reduce earnings fraud by 35 percent, amounting to $1.9 million in savings for New Mexico annually.
“The best advice I could offer for other organizations, particularly government agencies, is to not feel overwhelmed by the concepts of predictive analytics and behavioral science,” Bussey says. “While they will challenge you to rethink many internal processes, procedures and current ways of thinking, the potential benefits of projects such as this are worth the effort.”
Philadelphia 76ers: Winning fans without winning
In the 2015-16 NBA season, the Philadelphia 76ers earned the dubious distinction of having one of the worst seasons in NBA history, with a 10-72 record. The franchise also set the record for the longest losing streak in professional sports, at 28 games. And all that followed two other very poor seasons.
Despite the team’s struggles, fans have remained loyal. The Sixers earned a No. 5 ranking in NBA season ticket sales for the 2014-15 and 2015-16 seasons, and they’re currently No. 2 in the NBA for new season ticket sales.
But the organization was concerned that season ticket holders who had already spent three years waiting for “next year” would begin to lose patience. And by sports industry standards, the Sixers have a relatively small service and retention team, with only six account executives responsible for more than 8,000 season tickets. During the renewal period, it took the six-person team more than four weeks to work through their accounts and contact all the fans on their lists individually. Hoping to make the process more efficient, the organization charged its analytics team with finding a way to use data to help account executives prioritize their time so they could maximize the renewal rate.
Fill those seats
“Season ticket members are the lifeblood of our organization,” says Braden Moore, the team’s director of analytics and insights. “We want each seat to be filled with a passionate season ticket holder for all 41 games. And it’s even more important to make sure the seats are filled for seasons to come.”
To start, Moore, who previously worked in quantitative risk management at the Federal Reserve, and the analytics team gathered all the demographic and psychographic information they could get their hands on — tenure, location, purchase and attendance histories, demographic data in the team’s Acxiom system, CRM touch points, email marketing behavior and more. They then ran the data through machine learning processes (including logistic regression, support vector machines, random forests and decision trees) and developed a two-pronged model that incorporated the following:
Logistic regression to predict each prospect’s likelihood of renewal. This was used to set a base forecast and to determine overall priorities.
A decision tree to gain insights on breaking points of consumer behavior. This was used to tell the story to the account executives in a digestible way. It also identified which types of interventions and levers yielded the most success.
“The Philadelphia 76ers service and retention team is the best in the business — they are ranked No. 2 in the NBA in customer service — and they have been my greatest resource in determining where the information gaps were that would help the team hit its goals for the season,” Moore says. “I definitely wanted to make a model that was useful and delivered insights.”
“We didn’t necessarily have any metrics or KPIs specific to the model,” he adds. “Instead, we had the organizational revenue and retention targets. One of the organization’s core values is ‘Collaboration Wins.’ Therefore, it wasn’t about the success of this analytics project as much as it was a piece of the overall picture.”
With the full support of the executive leadership team, the project included an individualized attack plan for each account executive based on the value and tenure of their accounts. This, Moore says, enabled the salespeople to better understand the intricacies of the retention process, their client value and chances of renewals so they could better focus their time.
Moore says the changes instantly increased the speed and impact of initial sales. In the first week, accounts renewed, seats renewed and overall revenue improved by 3 percent to 4 percent. The service and retention team exceeded the NBA’s projections by 8 percent, and the current renewal rate is second among all non-playoff teams (19 percentage points ahead of the next non-playoff team).
“The team was excited for the results, but as with any new process, it took a little time to put into perspective why the new process was important,” Moore says. “On the surface, listing off coefficients and regression statistics doesn’t seem to help service season ticket members more effectively, but taking time to explain the information allowed the team to utilize key takeaways from the model to add an extra level of strategy when organizing their time in the hectic renewal season.”
Don’t give up
Moore’s advice to executives planning an analytics project is simple (and applies to the Sixers on the court as well): Don’t be discouraged by failure.
“Keep trying,” he says. “Not every project will lead to a robust model with clear takeaways, but you’ll learn something from each iteration.”
“Take time and do your homework,” Moore adds. “I’ve stumbled across numerous new methods or algorithms that I’ve used in subsequent projects just from continuing to research and learn. The field is continuously evolving, so we as professionals have to as well.”
The North Face: Customers for all seasons
California-based apparel company The North Face has built a highly recognizable global brand focused primarily on cold-weather gear — winter coats, ski jackets and warm fleeces. But that strong association has had a downside: Customers primarily purchase once a year and don’t buy much in spring or summer.
Moreover, though loyal, its customers don’t necessarily come back every year to buy new products.
“Customers were not returning; not due to dissatisfaction, but because the quality of the brand’s products was too high,” says Ian Dewar, senior manager of The North Face’s Consumer Lifecycle unit. “The level of ongoing engagement with customers was not strong beyond the first major purchase.”
Focus on activities
The company realized that to build repeat business, it needed to push beyond the winter jacket and fleece market. To do that, it had to identify other activities its customers enjoyed and other brands of products they used.
Whereas traditional segmentation focuses on finding the products people buy the most and then marketing additional options, The North Face needed to find the category of products its customers use the most, not just those they purchase the most.
“We began working on big data in 2013 with a pilot project proposed as an innovation experiment,” Dewar says. “We had great results, so we launched a second-phase pilot in 2014. Those two sets of results formed the basis for our recommendation to incorporate advanced analytics into our 2016 plan.”
Both pilots focused on using transactional data, social data and data on spending behavior to predict future purchases. “We have incorporated that learning into our current program in partnership with Tibco and SAS,” Dewar says.
From there, the company had a consulting firm pull together a collection of teams at The North Face to identify opportunities that could arise as the company gleaned insights from its analytics initiative.
Test and learn
“We identified over 25 unique opportunities across ecommerce, direct-to-consumer retail, brand marketing, sources, procurement and product development,” Dewar says. “For 2016, we established a short list of six key use cases we wanted to test and incorporate into our plans. As we test and learn from each use case, we know we have more to go back to.”
In this case, the company focused on enhancing direct customer engagement via a loyalty program, hoping to translate that into a higher level of engagement and increased sales across all retail channels over time.
Its loyalty program, VIPeak Rewards, allows members to earn redeemable “PeakPoints” for every dollar spent and for participating in local activities — endurance challenges, mountain athletics training sessions, skiing and snowboarding competitions and even lectures by athletes. Data from sales, web searches, event registrations, competitions, surveys and other sources is analyzed using platforms such as Tibco’s Spotfire and SAS and IBM analytics tools. The company examines that data to understand the sporting categories customers show the most interest in.
A wealth of data
Standard RFM (recency, frequency and monetary) analysis of past transactions is applied to identify top potential customers, while predictive analytics take into account the company’s model for selling high-quality, long-lasting outdoor products.
“There is so much data available,” Dewar says. “We initially thought we would be spending a lot of time looking for additional data sources — a.k.a. the big data question — but we have been pleasantly surprised at how much transaction and behavioral data we already have. A key lesson for us has been to maximize use of what we already have, data- and customer-wise, before chasing too much external data or expanding to a broad customer prospecting initiative.”
The North Face’s efforts resulted in a dramatic increase in cross-category sales, with the same customers making purchases more than once, Dewar says. The VIPeak program gives the company the ability to build a 360-degree view of its customers while also strengthening customer and brand engagement and increasing online shopping activity.
“By identifying the key product categories customers are most likely to buy next, The North Face has been able to increase both the annual frequency of purchases and the year-over-year return purchase behavior of the VIPeak customers,” Dewar says. “In addition, the lessons learned with the top loyalty members are now being applied to nonmembers to identify top prospects across the whole direct-to-consumer base.”
Asked about advice he might have for other IT leaders planning analytics initiatives, Dewar offers these tips, drawn from the three keys to the success of the North Face project:
Get executive and cross-functional buy-in prior to committing to the project.
Use your own data first; maximize the opportunity to get more from your existing customers.
Make sure data analytics projects have the same KPIs as the overall business, so key wins can be celebrated across departments and key results from a test and learn protocol can be integrated immediately.