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.

07/23/2014

Channel NewsAsia Video: Digital Healthcare

On Channel NewsAsia’s Firstlook, FICO’s Saxon Shirley discusses the results of a FICO survey on consumer preferences and tendencies with regards to mobile, online and in-person interactions with their health care providers. More than 2,000 adult smartphone users were surveyed in Australia, Brazil, China, France, Germany, India, Italy, Japan, Korea, Mexico, Russia, Turkey, the U.K. and the U.S.

The survey showed that 56 percent of people trust health care organizations with personal data, and simple innovations around mobile alerts and information services are helping to build the trust necessary for this trend to continue.

 

Click on the image above to view the video, or view it here.

07/15/2014

Infographic: Analytics that Think Like Humans

Check out our latest infographic that shows the various ways that Big Data analytic software mimics processes in the human brain. These applications of artificial intelligence address a range of real-world  challenges, from combatting financial crimes and computer hacking to verifying  identities.

Artificial_Intelligence_4_Ways_Analytics_Think_Like_You_4010IN

07/01/2014

VIDEO: Celebrating 10 Years, a Happy Tour of FICO Bangalore

Who says data scientists can’t let their hair down? In celebrating 10 years in Bangalore, our employees (many of them data scientists) show that they too are “happy.”

The video for Pharrell William's massive hit song has more than 300 million views on YouTube and has spawned thousands of copycat videos all over the world.

FICO Bangalore is no exception and, with its Happy video, is showing that data scientists know how to celebrate.

 

06/24/2014

How Government Agencies are Better Serving Citizens in Collections

Government
By Todd Rollin

“Hey, Rich – it’s Edward Jacobs returning your call.”

“Thanks for calling back, Mr. Jacobs. I had a chance to review your case and have some good news. We can go over the details on this call, or I can send you an e-mail and you can select your preferred option via our web portal.”

“E-mail would be great, Rich; I’m in a lot of meetings this week. Also, could you send me a text reminder like last time? That was great. Glad I signed up for those.”

“Perfect. Look forward to that text and e-mail, and we can go from there. And as always – just pick up the phone if you need to talk, I’ll be your designated point of contact.”

“Thanks. I really appreciate the help. Considering the situation, you’re making my life easier – it’s great working with just one agent and not having to tell my story to multiple people.”

Sound familiar? The answer may surprise you. Yes, it’s a customer service call. But it’s not what you think – Rich is a debt collector for a government agency, while Mr. Jacobs owes back taxes on some of the multiple properties he owns.

And while this specific instance is fictional, the exchange isn’t – FICO is working with several government agencies at the state and local level to improve taxpayer collections, without having to resort to tactics like screaming, threats, or worse.

One county agency connects and collects with its taxpayers every day. The Office of the Shelby County Trustee, home to more than 940,000 residents in the State of Tennessee, is challenged to resolve delinquencies with fewer resources and declining property tax values. Powered by an advanced collection system from FICO, The Office has transformed how its collectors collect – and how debtors such as parcel owners respond.

For example, the system links multiple properties to a single owner, so that one collector can work directly with a taxpayer and total dollars owed, reducing unwanted calls and debtors who have to “keep telling the same story to different people.” Also, data-based rule sets and decision trees improve workflow collection strategies such as account prioritization, segmented responses to different offers, even which resources should work which accounts. The Office has implemented additional communication channels, with full consent from those taxpayers who desire to opt-in, to help resolve debt in a mutually beneficial way – to the tune of a 548% return on investment.

State and local collectors are now equipped to generate better outcomes, while intelligent, flexible, and rules-driven payment options are making taxpayers less hesitant to resolve debts quickly. It’s a win-win for The Office.

The whole story is available at http://www.slideshare.net/FICO/shelby-county-success3080cs.

06/17/2014

Why We Built Our Own Analytic Cloud?

Build
By Benjamin Baer

Not too long ago, deploying advanced analytics required massive investments in IT infrastructure and application software, creating a sort of “analytic elitism.” Over the last year we’ve been focused on democratizing analytics, so that organizations of all sizes can base their operational decisions on data. And to do this we needed to build our own cloud infrastructure.

But why? Why did we decide to build our own infrastructure rather than leveraging a hosted provider like Amazon Web Services (AWS)? This was hotly debated within our company and essentially it came down to three very important reasons:

  1. Cost: While AWS, Microsoft Azure, the Google cloud and other cloud hosting providers garner significant attention and mindshare, first and foremost their value is speed of resource delivery. When and if a customer needs server and computing resources, the network for these services can be deployed and accessed in very little time. Speed is a critical value for hosting service customers. By partitioning and deploying compute resources in minutes or an hour customers can quickly get access and scale applications on AWS an order of magnitude faster than physically acquiring and deploying servers themselves.

    All of this comes at a significant cost. Amazon, Microsoft, Google and other cloud providers pass off the cost of buying, deploying and managing the datacenter resources to you, the customer. And depending on the type of deployment you are considering, this model might make sense. But the FICO Analytic Cloud is a significant endeavor, and we wanted to make sure that it was cost-effective for our clients. By building our own, we can better control the costs of our cloud solutions.
  2. Security: Customer data, financial data and other potential sensitive data are at the heart of making decision management and analytic solutions effective and useful. The diversity of our customers means that we are required to maintain the highest levels of data security and integrity. Even if we were willing to pay an outsourced provider for hardware PCI compliance, there would be additional compliance issues (for example HIPAA or DODIS) that would not be supported by the cloud providers. As such, we'd either not be able to support the largest FICO customers or would have to build our own data portal or unique data security solution.
  3. Expansion and agility: By creating, deploying and managing our own cloud infrastructure, FICO has control over its destiny and infrastructure direction. We can quickly add, complement and expand the cloud infrastructure capabilities to suit customers’ needs and requirements. This will increasingly be apparent as we add new features, capabilities and configurations. The ability to quickly expand and harden our cloud offerings, be agile in our development and roll out new capabilities is a feature of the control we've built into the infrastructure. We can only do this because we're making the investments from the ground up.

Today, we introduced free 30-day trials for our best-in class-tools in the FICO® Analytic Cloud, our new cloud infrastructure for creating, customizing and deploying powerful, analytics-driven applications and services. As we rapidly debut new solutions, and our first customers take our cloud for a test drive, we are confident that we made the right decision to build, rather than rent or buy.

06/11/2014

Play More Games to Get Your Ph.D.: Fun with Spurious Correlations

Correlation doesn’t equal causation. This point sometimes gets lost in the Big Data discussion.

The argument goes that if you have enough data, correlation is good enough. Even our friend Kenneth Cukier wrote that the focus of analyzing Big Data will shift from causation to correlation. “This represents a move away from always trying to understand the deeper reasons behind how the world works to simply learning about an association among phenomena and using that to get things done.” And current TED curator and former WIRED editor Chris Anderson proclaimed in his 2007 essay that the data deluge will make the scientific method obsolete. “Petabytes allow us to say: ‘Correlation is enough.’”

Even our Chief Analytics Officer will admit that for some things correlation may be OK. However, for many things it is important to demonstrate the impact of an action in a cause/effect fashion. For example with loyalty programs, members in the program may spend more than those not in the program. But, is it just a correlation? Would they have spent more anyway because they are the best customers, or did the program drive the higher performance? It is important to determine causation, or you may be leaving money on the table, or throwing good money after bad.

We also don’t believe that you can use Big Data to solve problems without understanding the problems. Big Data can tell you nothing, if you don’t know what questions to ask it.

This is where the fun with spurious correlations comes in. Tyler Vigen has put together a great collection of spurious correlations. Here’s one that we appreciated:

Correlation

So where do you stand on the correlation/causation debate? Let us know.

06/03/2014

Your Washing Machine Knows What You Did with Your Darks

Washing Machines
The Internet of Things is expanding beyond smartphones, game consoles, wearable wellness and the ilk to include your appliances. In April, Berg, a design and technology development startup in London, debuted Cloudwash, a prototype for a connected washing machine.

Unlike the complex and often over-engineered internet appliances of old, this washing machine is a refinement on the 106 year old innovation. So why should you connect a washing machine?

The Berg team believes the biggest innovation is the ability for the machine to tell time. Cloudwash can tell you when a load will be complete, eliminating the guesswork typically involved in transferring things to the dryer. Other capabilities include smartphone controlled pre-sets for communicating directly with your washer.

While those are good for a privately owned washing machine, the benefits for public or shared washing machines could be even more significant. Connected washing machines will have the ability to accept PayPal or a credit card, just like parking meters do today. You no longer have to track down those pesky quarters or prepay those annoying cards. They have the ability to text you, or send you an alert, when the washer is available, or when your load is ready to move to the dryer. Thus removing the awkward Saturday afternoon wash days when you sit around watching the machine spin.

The Internet of People and Things is ever-expanding. Giving breath to things as mundane as your washing machine.

05/20/2014

The Fool’s Errand: A Segment for One

For one
By Mike Farrar

Back in the day, one of the dumbest activities I ever engaged in was to develop a program we called "The Segment of One."

The big idea was to develop a business-rules-based process to create a decision tree to assign a cascading set of offers so that every customer would get what we thought was the best offer uniquely targeted to that customer. Our philosophy was that there ultimately might only be a single customer at one of the terminal nodes of the tree, but that customer would be getting the very best set of offers possible.

Oh so young and oh so foolish. Notice that we weren't using optimization. We were just declaring our decisions "best" and calling it a day. Fair enough. When all you have is marketing judgment to go on, you go with it.

Unfortunately, fully ramified, the tree would have given us something like 2^15 permutations to try to manage. The program never quite lived up to its billing.

The big point we ignored was the incremental value of such a fine segmentation. If 1,024 segments delivered value, then great. But was that 1,024th segment delivering that much more financial value than 1,023 segments? Would the extra targeting give you more revenue? Would it make a customer all that much more loyal? Would it give you better logistics and inventory control? Would it improve efficiencies in your marketing? Would it give  you enough to cover the extra costs and headaches of having that extra segment?

Could we have gotten away with 512 segments, or 256, or 128, or even eight?

It's not all that bad to have eight segments. Lots of firms get away with it and do just fine.

I’ve seen several recent presentations, white papers and articles arguing for the death of segmentation, which is equivalent to saying a Segment of One. Maybe that is the way the world is headed, but I doubt it. People are complicated, but we're not all that dissimilar.

 

05/15/2014

Video: Unleashing Text Analytics

Emails, text messages, documents and other unstructured human-generated data represent the next major frontier for data scientists. Being able to mine text with text analytics tools could help businesses make better predictions and power better decisions – but it is easier said than done. In this video, FICO’s Andy Flint talks about both the promise and the challenges of putting text analytics into practice.

 

05/13/2014

The Algorithm that Runs the World

Map
By Qi Huangfu

“The algorithm that runs the world” – says the cover story of a 2012 issue of the NewScientist Magazine. And yes, it is talking about the simplex method.

While the theoretical idea behind the simplex algorithm is relatively simple, how to efficiently implement it in practice is more complex, and has been a major topic for research. Looking back at its history of how the theory evolved, ever since its introduction by George Dantzig in the 1940s,  there have been new ground-breaking research published roughly every decade, with the latest one being Hall and McKinnon’s hyper-sparsity algorithm around the 2000s. However, the simplex world has been relatively quiet since that publication, as if hyper-sparsity is the final brick to the well-polished simplex building.

Inspecting the situation more carefully, the simplex research has never stopped. A highly active direction is parallelization. In the last 20 years, there have been lots of simplex parallelization attempts, from massively distributed machines to modern GPUs, using either dense or sparse matrix computations. However, good speedup ratios are related to less efficient sequential implementations of the simplex method. As a result, the overall performance of these parallelization efforts is still inferior when compared to world-leading sequential simplex solvers.

A meaningful simplex parallelization has to be designed and developed based on a state-of-the-art sequential simplex solver and still allow for significant parallelization. This has been achieved in the most recent version of our optimization suite. As the first commercial parallel simplex solver, it achieves speedup to a factor of two on various large-scale problems.

The key idea behind this implementation is to create more parallelizable scope by combining multiple simplex iterations into one meta simplex iteration. In order to implement this efficiently, several innovative techniques had to be developed. A detailed introduction to this approach is available here. The implementation in our optimization suite is an adaption and further development of this key idea.

After 70 years, we are finally realizing the promise of the algorithm that runs the world.

Search Site


FICOLabsBlog.fico.com

Subscribe

Enter your Email

Preview | Powered by FeedBlitz