About the Blog

The FICO Labs blog focuses on innovation, technology and the role of analytic sciences in today’s business world. We’ll share our perspectives, vision, successes and challenges in the areas of predictive analytics, Big Data, decision optimization, cloud computing and Big Marketing.


Video: Building Models that Can Handle Messy Real-World Data

 Data is messy and complex. By volume, 80 percent of the data out there is untapped and much of it is unstructured – too messy and just too complex to crack open. Regression models thrive on numbers and numbers only, so data in the form of customer call logs, online support chats and detailed case records are just irrelevant noise to them. And when the source information doesn’t “fit” neatly into a lot of our tools, we often write it off as unusable or meaningless.

In this video, FICO’s Andy Flint talks about how we build models to handle messy, unstructured, real-world data.


Why Customer Segmentation Still Matters in the Era of Big Data Analytics

In the era of Big Data analytics, it is fairly common to dismiss segmentation as an old methodology with minimal, if any, role to play in customer-centric decisioning. A majority of segmentation techniques are descriptive in nature and hence fail to capture the imagination of those who desire to use the latest techniques in modern crystal-ball gazing. That’s because people still think of relatively simple segmentation methods — in fact, there is a vital role for more advanced segmentation.

Segmentation of Old

The oldest technique is grouping customers based on their demographic traits. Slightly more sophisticated segmentation techniques use value dimensions instead of demographic traits. These involve identification of one or more value dimensions, followed by dividing each of the dimensions into bins, usually of equal volumes. For instance, a retailer could describe its customers by deciling them first on their yearly spend and then deciling them on total number of trips, thus giving 100 micro-segments. These techniques are relatively easy to develop and very useful for descriptive reporting and monitoring, but usually don’t provide insights beyond that point.

An improvement over these approaches is to use behavioral dimensions along with value dimensions. This allows identification of groups of customers who are behaviorally similar but differ in values, or vice versa. The approach necessitates using a data-driven technique to identify segments of customers, usually k-means clustering. The segments are much more nuanced and allow identification of the factors that generate customer value. Armed with this knowledge, sophisticated marketers use external stimuli to influence customer behavior to generate higher values. Still, these segments are not actionable, as they do not differentiate customers based on future outcomes. Another challenge faced is that k-means presupposes the number of segments and this often involves poor guesswork.

Segmentation in the Big Data World

Our approach overcomes these limitations using very powerful innovations. Using a genetic algorithm-based search technique, the optimal number of customer segments is identified. These segments have minimal overlap across segments and minimal separation within the segments. The segments are designed to provide maximal separation on certain predetermined dimensions, called the relevance-drivers. The relevance-drivers are used only during segment identification but are not needed for scoring. Hence, these are ideally suited for defining future performance metrics like future risk or future revenue.

Using our segments, an organization can not only identify behaviorally distinct groups which also differ in their historical value to the organization, but can also predict future desired outcomes like risk and revenue. Appropriate actions can then be identified for each segment based on their historic behavior and value, and expected future outcomes. These segments can be used in lieu of more expensive combinations of predictive models for tailor-made customer-oriented actions when the number of customers or accounts is relatively low, usually under a few million. These segments not only provide a powerful way of describing actionable segments for all sizes of organizations, but also provide a cheaper way to execute 1:1 customer-centric actions for small and mid-size organizations.


Will Wearables be Bigger than the iPad?

By David Ratz

The uptake by consumers of wearable devices, from health fitness bands to Google’s Glass, is seeing continued growth this year. According to CNN Money, analyst firm Canalys estimates 17 million wearable fitness bands will be sold in 2014 and that total will grow to more than 45 million by 2017. Does this mean wearable devices could surpass the total number of iPads sold?

It’s certainly a possibility, especially as large technology companies raise awareness of wearable devices, but Apple has set the bar high. In October 2013, Apple reported it has sold 170 million iPads since the product’s launch in 2010. That’s nearly 57 million iPads sold annually over the last three years.

Plus, experts are still uncertain if mainstream consumers will invest in wearable devices like early adopters have.

Product Strategy Consultant Kevin McCullagh, in an article for Fast Company, wrote:

“The biggest threat to the wearable nirvana is the smartphone. The mobiles that we carry around with us are incorporate movement-tracking capabilities. The iPhone 5S’s M7 chip is dedicated to processing motion data from the phone’s accelerometers, gyroscope, and compass sensors. As more health and fitness apps and equipment tap into this functionality, dedicated activity trackers are likely to go the way of the alarm clock, radio, MP3 player, GPS unit and camera, swallowed up by the smartphone.”

As a wearer of a FitBit Flex and user of the RunKeeper app, I can relate to what McCullagh is saying. I use both technologies to track my fitness activities. The RunKeeper app excels at track my running route, mileage, total time, average pace, activities per week, time spent working out per week, plus the app sends me workout reminders.The FitBit is not good at tracking individual workouts, but is great at showing my overall daily activity. Before the FitBit fitness band I knew I wasn’t overly active during the work day, but I didn’t realize how inactive I really was, especially on days I didn’t run.  The value of fitness bands and apps that can track a person’s activity level is definitely in the data collected.

Here’s a snapshot of a week where I only ran twice during the week. The graph visual shows how inactive I was on the days I didn’t workout.

Fitbit Trackere

As we wrote about in the blog post, “Wearable Wellness – The Latest Fashion in Big Data,” the long-term appeal to consumers of fitness bands is turning their personal data into actionable insights. Of course, here at FICO we are big fans of the power of predictive analytics to do just that.

If you use a fitness band or other wearable devices, we’d love to hear what you’ve learned from the data. Let us know in the comments.


Analytics and Intuition: Why Your Expertise Matters

  You Matter

By Tim Young

When making decisions as a business leader, do you ever take time to think from your customers’ point of view? Aren’t you also a customer? Don’t you have similar expectations as your customers? When you’re a customer, how do you like to be treated?

With the growing potential for analytics and Big Data, sometimes we can get caught up in the numbers, and forget to apply our experiences and our intuition to the equation. We can be lead to believe that decisions – like how to price a product, where to add a new retail location and whether to introduce a new product – can only be found though number crunching. We might assume that quantitative analytics can make expert judgment – your judgment – obsolete.

The danger of these assumptions is that they lose sight of that fact that the “end consumer” is a real living breathing person. And people do not always make rationale choices. Emotion plays a big part in the brands that we are loyal to, the decisions that we make and why we make those decisions.

I’m a big believer in using reliable data and fostering a culture that embraces analytics. Making decisions without analytics and data puts you at a decided disadvantage. Surveys of one are inaccurate and not representative of the overall population.

This doesn’t mean I think intuition and experience don’t matter. Intuition and experience can unlock the full potential of quantitative analysis. By using business expertise you can ensure that you’re asking the right questions of the data and appropriately applying the insights gleaned from the data. Expertise can also be gained through observation – getting your hands dirty, and spending time in the field where your customers are.

The trick is to get the balance right, and if you do that, the benefits can be significant. A recent Accenture survey of 600 businesses in the US and UK found that when data and intuition are used together, businesses reported a more than 75 percent return on their analytics investment over a two-year period.

With the right combination of analytics and intuition, your business can make better decisions. Listen to that little voice inside your head – your intuition has value.


Infographic: Helicopter Blades, Airbags and Analytics

People describe lots of things as being “real-time.” But what does that phrase mean? If a task takes 10 seconds, is it real-time? If it takes one second, is it real-time? How about blinking your eyes—can you do that in real time?

Below is an infographic that investigates what constitutes “real-time.” It compares how fast the average person blinks, a car airbag inflates, and a helicopter blade rotates.

When it comes to combating payment fraud, we don’t think any of those speeds are good enough to earn the “real-time” label. In just 40-60 milliseconds, our fraud management software completes 15,000 calculations based on a ton of data associated with each credit card (or debit card) swipe, including the transaction amount, merchant profile, transaction location, point-of-sale device, time of day, and account history. All 5X faster than the blink of an eye!

The infographic can be downloaded at www.fico.com/realtime.



Analytics, Optimization and the Asia Travel Boom

By Amit Parekh

Over the last few weeks, the world has been transfixed by the baffling mystery of the Malaysia Airlines jet. One particular article on news.com.au on March 21 caught my atttention, Missing Malaysian Airlines flight MH370 ‘won’t stop Asia travel boom’. The article took an interesting look at the airline industry itself and the transformation occurring in Asia. The backdrop is compelling even if far removed from the headlines, so I thought I would share with you some of the most interesting points from the article.

Air travel in Asia is surging as the middle class gets bigger, discount airlines proliferate and business ties with the rest of the world deepen.  According to the Association of Asia Pacific Airlines, the biggest demographic shift for Asia in the past 20 years has been the sheer number of people who have been lifted out of poverty into that middle income segment of $10-$100 of disposable income a day.

The International Air Transport Association has forecast airline passengers to grow by 31 percent worldwide between 2012 and 2017. For Asia, that will mean the number of passengers increases an average of 6.3 percent each year, nearly three times as fast as the US.

Routes within or connected to China will be the single largest driver of growth, accounting for nearly a quarter of the additional 300 million passengers during those six years.

Given the demographics, it is actually the low-cost carriers who are the hungriest players. In Asia alone, Airbus has 1375 unfilled airplane orders or about a quarter of its worldwide order book. Malaysia-based AirAsia and its affiliate AirAsia X together have orders for 385 new planes. Indonesia’s Lion Air has an order for 234 jets from Airbus and another 301 from Boeing.

Further north in China, not only are new low-cost carriers about to take flight, but airport construction is most rampant in China, with authorities in the world’s second-biggest economy authorizing the construction of dozens of new airports and the expansion of others.

To keep up, Asian governments are scrambling to build new terminals and runways. Singapore expects additions to its airport will within a decade more than double the number of passengers it can handle yearly to 135 million, while several airports in the region are already operating near or at capacity, including Hong Kong, Bangkok’s Suvarnabhumi Airport, Manila, Jakarta and Beijing.

This travel boom is contributing to a need to model and solve complex problems with analytics and optimization, in order to improve operations and deliver safe and reliable transportation to millions of passengers. At FICO, our optimization tools are used to help airlines make critical decisions around fleet planning and scheduling, tail assignment, MRO (maintenance, repair and overhaul) planning and scheduling, capacity planning, flight planning, crew allocation, yield and revenue management, and operational controls.

So as Asia’s skies get busier, the airline industry continues to grow intensely competitive. Sometimes the difference between a flight being profitable or losing money comes down to just a couple of unsold seats. That's why airlines are increasingly turning to analytics to build efficiency into every part of their operation.

To keep up with this boom, modern optimization tools are continuously  improving performance and computational speed.  In fact, based on recent calculations on our airline test set, we observed an improvement in performance of nearly 300-400 percent in the last decade. If you factor in the increased hardware speed, this means that for a medium-sized airline, a typical crew planning optimization run takes only a few minutes compared to an hour or more just ten years ago.

While this may not seem like a big deal to outsiders, the ability to optimize key operational decisions on the fly (no pun intended) has become critical for the Asia travel industry. It is one small way the airline industry continues to invest to improve operations and keep pace with the travel boom.


Mythbusters: Do We Need More Customer Data?

More data
By Feather Hickox

With our ongoing series of mythbusters, inspired by the Discovery Channel’s television show MythBusters, we’ve been tackling hotly debated beliefs related to Big Data, analytics, customer engagement and mobile technology to determine whether the belief can be confirmed, is plausible or is busted (not true). Today's myth that will be put under the microscope: Do we need more customer data?

The “do we need more customer data” debate has been making the rounds for the last year. All Things D ran a series last year asking the question “More Data or Better Algorithms?” It ended with the punch line that better data wins. This goes in stark contrast to the conclusion of Kenneth Cukier and Viktor Mayer-Schönberger.  In their book Big Data: A Revolution That Will Transform How We Live, Work, and Think, they describe Big Data as unbounded and unstructured; imprecise but predictive; and not causal, but correlative. Big Data by its nature is messy data, it doesn’t fit in neat rows and columns, and by definition it begins when current data collection and analytical paradigms fail to fit your needs. Their premise is that more messy data, not better data, wins.

Our take is that we DO need more customer data—but that voluminous data alone won’t lead to success. What will is the ability to pull more meaningful insights both from the data you have, as well as from new data sources.

In the era of Big Data, you need to be able to add new data sources and analytics to your decision processes with ease and without having to queue up for IT time. Business rules management systems (BRMS) and applications that incorporate them enable nontechnical users to inject these powerful new elements into existing processes. Very quickly, decisions get smarter.

Many companies have piles of data they’ve collected but are not yet leveraging for better decisions. Unstructured data, for instance, from customer service chats and phone conversations, as well as online customer product reviews, are full of potentially valuable insights. Analytic techniques, such as FICO’s Semantic Scorecard technology, can extract the most predictive features from this data, and combine it with traditional structured data analytics to improve the accuracy of customer behavioral predictions.

In other cases, where data is already being analyzed, there may still be potential to draw out more value by looking at it in a new way. Many companies have collected transaction log (TLOG) data, for example. But few are currently analyzing it with time-to-event models in order to predict with precision when a customer behavior is most likely to occur (this week? next week?).

  Mythbusters figure 6

Figure 1: Time-to-event models pinpoint the best time to make an offer

There are also analytic techniques today that can go beyond finding the usual correlative data relationships (when A occurs B also occurs). They can scour existing data to find more powerful, previously undiscovered causal relationships (A affects B in this specific way). With these and other advanced techniques, companies can answer complex questions such as: “Which customers will buy only with this particular coupon, which will buy with any one of three possible coupons, and which will buy whether or not they receive a discount?”

Using mathematical optimization, we can find the best decision strategy, given all business objectives and constraints, for maximizing a particular goal, such as profit. We can also simulate the impact on key performance indicators of making adjustments to one or more constraints.

FICO has helped a large US retailer use techniques like these to achieve response rates of as high as 30% from its loyalty club members. We’ve also helped a leading Canadian bank adopt the analytic techniques to develop an “optimization culture” for driving performance improvements ever higher. In credit line management, as shown in Figure 2, this bank’s latest optimized strategy (purple V3 line) has already lifted incremental profit by $5 per account at the 4-month mark, set to outpace its previous 2 successful optimized strategies (blue V2 and green V1 lines).

Mythbusters figure 7

Figure 2: Using optimization to drive incremental profit gains

The reality is that more data sits on a continuum between where it can offer real insight and where it will be a costly distraction. The better question to ask is: What value can you get from more customer data? Collecting data is easy and relatively inexpensive; making the data useful is what’s challenging – whether you’re dealing with more data or not.

Did we convince you? For more mythbusting, check out our Insights paper titled: Marketing Mythbusting—Six Maxims Get put to the Test (No. 73; log-in required)


How Analytics is Driving Success in the Age of the Customer – Live Webcast

Update: If you missed the live webcast, you can now access the recording (registration required).

This Thursday, FICO’s David Ross and Forrester’s John Rymer will be hosting a live webcast about the next generation of analytics and cloud technology in the age of the customer. Rymer will discuss his research on the age of the customer, where the most successful businesses will reinvent themselves to systematically extend and serve increasingly powerful customers. And Ross will give you an exclusive preview of some new decision management technology from FICO. Rymer and Ross will also be taking questions and comments via Twitter at #FICOLiveCast. 

To tune in and join us for the live webcast on Thursday, March 27 at 2 pm EST, you can register to attend here. See you then!


It’s 2014. So Why Are You Still Wrestling with Old Collection Software?

by Morgan Nagle

Ten years is a long time by any measure. Back in 2004, people were talking about:

  • Women’s wrestling debuted as an Olympic sport at the 2004 Olympic Games in Athens, Greece. Irini Merleni of Ukraine won the sport’s first gold medal.
  • Martha Stewart was in the big house.
  • George W. Bush was re-elected as President of the United States, defeating John Kerry.

Like myriad other businesses, the debt collection industry is using software that’s a decade old, and even older. These firms are literally wrestling with outdated technology, trying to make it work in a debt collection environment that has also changed over the last 10 years. In the race to collect more, with fewer resources, and remain compliant, old technology just won’t cut it.

New debt collection software: That’s so 2014

Modern software offers a bunch of advantages for organizations that upgrade. We’ve found that our clients who upgrade often see:

More revenues:

  • 10 percent reduction in write-off roll rates
  • 18 percent increase in average remittance
  • $20,000 increase in revenue per agent

More efficiency:

  • 40 percent increase in right party contacts
  • 30 percent uplift in self service
  • 15–20 percent agent productivity improvement
  • 40 percent increase in daily processing volume
  • 30 percent decrease in average account process time
  • 40 percent reduction in average customer contact time
  • 5 percent increase in immediate payments

Less compliance risk:

Speech analytics puts an audit trail in place to ensure that agents use language that not only complies with a multitude of regulations, but also achieves better business outcomes.

For  example: After implementing an advanced speech analytics solution, a leading collection agency identified more than $200,000 in potential violations. The solution also identified the agency's top collectors, which helped it implement a best-practice training program.

It’s easy to get started

Today’s debt collection software is easy to deploy, with cost-effective options ranging from subscription-based cloud solutions to traditional on-premise licensing.

Wrestling is an exciting sport, but wrestling with outdated collection software is inefficient, frustrating and detrimental to your business. So why not upgrade your system and step out of the ring?

Feel free to comment or share my blog post on social media. Thanks!


The Modern Day Sports Schedule Makers: The Science and Art of Solving Massive Analytic Problems

By Horia Tipi

For 23 seasons, starting in 1982, Holly and Henry Stephenson, a husband and wife duo, were the masterminds behind the Major League Baseball (MLB) schedule. The Stephensons were profiled in an ESPN 30 for 30 Shorts film, which describes how the couple, with a computer, a pencil, and a whole lot of permutations and variables, took on the daunting and thankless job of MLB scheduling from an upstairs bedroom in their Staten Island home.

In 2005, Holly and Henry were replaced by analysts specializing in computational methods in optimization. These modern day schedule makers with “random” computational optimization scheduling software were blamed for issues in subsequent MLB schedules. For example, the film points out that the 2013 Yankees season ended at an away game in Houston, which also happened to be the last game played by Yankees great Mariano Rivera. Of course, Rivera announced his retirement after the 2013 season schedule was set, he could have retired anytime during the season, and the Yankees didn’t even make the postseason! Hardly a snafu by modern day sports schedule makers – but it speaks more to the inability of the league and the teams to know in advance or clearly articulate their requirements. 

Sports scheduling is complex business involving variables such as venue availability, fairness in balancing travel times and distance, league rules, sponsorships and advertising, TV schedules, and fan attendance. Scheduling games for a single season can involve literally trillions of scheduling permutations with thousands of variables and tens of thousands of constraints. Each year, team management, sportswriters and broadcasters, athletes, and fans ridicule the scheduling logic and blame the schedule makers.

Schedule making is as much science as it is art. Modern schedule makers – like our client Bortz Media, which produces season schedules for professional, collegiate and recreational baseball, football, basketball, hockey and tennis leagues – develop mathematical models that can find the best scheduling permutations.

Coming up with the best schedules that are fair and most closely reflect the goals of the league (e.g., marketing, competitive balance) and the sensitivities of individual teams (e.g., travel, attendance, etc.) also takes a fair amount of domain expertise. A sports season needs to tell a story, and television ratings and fan attendance depend on the drama created by great match ups and rivalries. For example, the NFL often opens the season with a Super Bowl rematch, the NBA plays on Christmas day, and many leagues end their seasons with battles between divisional rivals.

This domain expertise is a critical asset that helps the modern day schedule maker develop the intuition to create the best mathematical models. The scheduler needs to elicit any and all requirements from all stakeholders, translate these requirements into mathematical equations (objectives and constraints), and use an optimization engine to rapidly identify schedules that meet most or all of the constraints. Mathematical optimization software delivers the automation and the execution speed that allows the schedulers to do their work of understanding and extracting value. But it’s the schedulers and their expertise that bring the real alchemy in this scheduling equation.

Search Site



Enter your Email

Preview | Powered by FeedBlitz