It Takes Courage to Compete on Analytics

Billy Beane’s “Moneyball” approach to developing and staffing a professional baseball team has come under intense scrutiny as long time major league scouts and analysts take delight in sub-par performances of the Oakland Athletics.  Beane however seems undaunted in using statistical analysis to undercover market place inefficiencies.  Indeed, it takes courage to solve problems in a whole new—perhaps heretical—manner. And while plenty of folks take pleasure in Beane’s recent comeuppance, there’s a good chance he’s already had the last laugh.

By now, most business professionals have either read Michael Lewis’ “Moneyball” or at the very least seen Brad Pitt’s rendition of Billy Beane in Hollywood’s version.  Beane’s claim to fame is that he had the courage to defy decades of “common knowledge” in assembling a winning baseball team.  For many years, winning teams were accustomed to paying top dollar for the best hitters and pitchers evidenced by metrics such as batting average and earned run average respectively.

However, in the small-town Oakland market, Beane didn’t have the luxury of affording a high payroll. And if he wanted to compete, he needed to find a different (and perhaps better) method of keeping up with the likes of the Boston Red Sox and New York Yankees.

Beane knew there were “inefficiencies” in baseball markets that could be exploited; he just had to discover those undervalued metrics.

Enlisting the help of his Harvard friend Paul Depodesta, Beane used statistical analysis to discover metrics that—in his opinion were more indicative of baseball success—such as on base percentage and slugging percentage for batters, or the ability to get “outs” for pitchers.  Then Beane went about acquiring a motley bunch of players that most teams discarded as near worthless, and positioned each player for success based on his analytical models.

And for a while it worked. With one of the league’s lowest payrolls the Oakland A’s competed with teams spending 2-3x in overall salaries. In fact, a Sports Illustrated article mentions; “From 2000 to 2006, (Oakland A’s) averaged 95 wins, captured four AL West titles and made five playoff appearances.”

Alas, all competitive advantage is fleeting. As Paul Depodesta left for greener pastures, and a few of Beane’s disciples took other league jobs, major league baseball teams figured out what made Moneyball work and effectively neutralized the Oakland A’s advantage.  With other teams adapting to Moneyball tactics, from 2007 and beyond, the Oakland A’s were a .500 team at best, and sometimes much worse.

And in the process of adopting Moneyball tactics, Billy Beane made plenty of enemies. The “old guard” definitely had a preferred way of defining baseball success and would often scoff at the “fat bodies” Beane employed, and laugh at the un-orthodox pitchers he would trot out in the ninth inning.  And in regard to the lackadaisical record of the Oakland A’s in recent years, one scout sneered; “So much for the genius … He doesn’t look so smart anymore, does he?”

Competing on analytics takes courage. In response to analyst and pundit criticism leveled at Beane for his Moneyball approach, author Michael Lewis says; “Beane had the nerve to seize upon ideas rejected, or at least not taken too seriously, by his fellow Club members, and put them into practice.”  For a while, Beane had the advantage. It might not have been pretty, but the Oakland A’s were effective at winning as many games as other prestigious teams, for half the cost.

And Beane ultimately may have the last laugh. Today, there are teams like the Boston Red Sox that take the best of Moneyball tactics and overlay them with powerful financial resources to take baseball competitiveness to a whole new level. Baseball has been changed forever from a “gut feel” business to one that’s analytically driven.

Fortunately for today’s analytical professionals, “competing on analytics” is taking hold as more companies understand they’re sitting on literal goldmines of structured and multi-structured data just begging for analysis. But analytics users be forewarned, there’s still plenty of “old guard” that will put up roadblocks, enlist subterfuge and even openly mock your new data driven approaches.

Michael Lewis notes that Beane didn’t invent sophisticated analytical analysis—he just had the courage to use it to create competitive advantage and shake up a traditionally stodgy industry.  There are plenty of industries ripe and ready for the business value that analytics will eventually unlock. Are you ready for the challenge?

Cruise Ship Captains and Normal Accidents

With the official death toll at eleven and plenty more persons missing in the Costa Concordia cruise ship disaster, reports have trickled in that this accident was mostly “operator error” induced on part of the ship’s captain. However, as we will see in any system –whether it is tightly or loosely coupled, there is always probability of accident, leaving us to calculate if we should build more slack or buffers into a given system, or accept risks of leaving things just as they are.

As the Costa Concordia departed from the Italian port of Civitavecchia, no one predicted it would crash and sink.  According to news reports, the ship’s captain wanted to “salute” the island residents of Giglio and took the ship near shallow waters. However, through mis-calculation, the cruise ship hit rocks, took on water and eventually capsized causing passenger injuries and some deaths.

The captain of the Costa Concordia has admitted to navigational error, even though on previous journeys he performed similar maneuvers taking the cruise ship too close to the rocky island.  And while it’s easy to pin blame solely on operator error, or in this instance bold incompetence, we should not forget that accidents such as these should be expected in moderately coupled systems.

We’re all risk takers to some degree. Implicitly, we understand whether we fly commercial airlines, drive a car, or take our family vacation on a cruise there is risk of injury or worse involved.  However with probabilities of “accident” so small, we judge risk vs. reward and usually settle on taking our chances with an understanding that accidents cannot be altogether avoided.

Author Charles Perrow reminds us that in systems—whether they are tightly coupled (little to no slack) or loosely coupled (linear interactions with possible delays) there “will be failures”.  In fact, in the maritime industry, which Perrow calls “moderately coupled”, it’s surprising there are not more boating accidents.

In his book, “Normal Accidents” Perrow mentions how the maritime industry is full of risky behavior where captains make poor judgments in “playing chicken” with other boats, or take sightseeing tours that deviate from planned navigational routes.  In addition, on the sea, operator error abounds where captains “zig when they should have zagged” or make mistakes because they often work 14-hour work days. Aboard a ship, Perrow says, the captain is supreme and with so much riding on one person, it’s not surprising many of the worse maritime disasters are due to poor decisions of the captain.

So we then realize—whatever the system—whether it’s a transformation industry like nuclear power, or one that’s less complex like maritime, there will be failures and there will be accidents. There’s no such thing as a non-risky system.

In fact, since risk in systems cannot be eliminated, we can either take drastic steps such as shutting down such systems (i.e. Germany’s pledge to eliminate nuclear power by 2022) or learn to live with risks by inserting more buffers and slack to stop chain reactions—if possible. This last strategy of course is more costly to society.

Perrow reminds us that when it comes to systems, “nothing is perfect, neither designs, equipment, procedures, operators, supplies or the environment.” Indeed, we know risk is all around us. What we can do however, as business managers and/or members of society, is to think more carefully about risks we’re willing to accept vs. plausible rewards. In addition, where possible we should strongly consider building in redundancies and planning for disaster so that when the inconceivable strikes, we’re much more prepared to deal with potential outcomes.

Questions:

• It is alleged the Costa Concordia captain took the ship close to shore so that families on the island could see the ship passing. Might system automation such as “auto-pilot” prevented the ship’s captain from deviating course?
• In the Costa Concordia disaster there is potential for fuel leakage polluting one of the “most picturesque stretches of the Italian coast”.  Perrow asks; “What risks should we promote to enable the private profit of a few?” Thoughts?

What’s Next – Predictive “Scores” for Health?

In the United States health information privacy is protected by the Health Information Portability and Accountability (HIPAA) act.  However, new gene sequencing technologies are now available making it feasible to read an individual’s DNA for as little as $1,000 USD.  If there is predictive value in reading a person’s gene sequence, what are implications of this advancement? And will healthcare data privacy laws be enough to protect employees from discrimination?

The Financial Times reports a breakthrough in technology for gene sequencing, where a person’s chemical building blocks can be catalogued—according to one website—for scientific purposes such as exploration of human biology and other complex phenomena. And whereas DNA sequencing was formerly a costly endeavor, the price has dropped from $100 million to just under $1,000 per genome.

These advances are built on the back of Moore’s Law where computation power doubles every 12-18 months paired with plummeting data storage costs and very sophisticated software for data analysis.  And from a predictive analytics perspective, there is quite a bit of power in discovering which medications might work best for a certain patient’s condition based on their genetic profile.

However, as Stan Lee’s Spiderman reminds us, with great power comes great responsibility.

The Financial Times article mentions; “Some fear scientific enthusiasm for mass coding of personal genomes could lead to an ethical minefield, raising problems such as access to DNA data by insurers.”  After all, if indeed there is predictive value via analyzing a patient’s genome, it might be possible to either offer or deny that patient health insurance—or employment—based  on potential risks of developing a debilitating disease.

In fact, it may become possible in the near future to assign a certain patient or group of patients something akin to a credit score based on their propensity to develop a particular disease.

And something like a predictive “score” for diseases isn’t too outlandish a thought, especially when futurists such as Aaron Saenz forecast; “One day soon we should have an understanding of our genomes such that getting everyone sequenced will make medical sense.”

Perhaps in the near future, getting everyone sequenced may make medical sense (for both patient and societal benefit) but there will likely need to be newer and more stringent laws—and associated penalties for misuse) to ensure such information is protected and not used for unethical purposes.

Question:

• With costs for DNA sequencing now around $1000 per patient, it’s conceivable universities, research firms and other companies will pursue genetic information and analysis. Are we opening Pandora’s Box in terms of harvesting this data?

Ring in the New Year with New Data Products

For web-based businesses, and of course, those with a web presence (which is just about everyone) there’s a goldmine of behavioral data accessible with the right tools. The trick is getting past static web analytic reporting (bounce rates, page views, session times etc.) and going further into unlocking the rich treasure trove of machine data, text and weblogs that create “big data” insight.

In your business, gigabytes if not terabytes of multi-structured data are likely just waiting to be coupled with your imaginative thinking and analysis to create new data products that ultimately help drive customer interactions and revenues

For example, take a look at what LinkedIn is doing in creating new “products” with data they collect and analyze with a MapReduce approach and other techniques.

According to a recent whitepaper “Building Data Science Teams”, LinkedIn’s former Chief Data Scientist shows how smart thinking can be paired with compute power and huge quantities of multi-structured data to create innovative new products such as:

• Products that provide personalized content (which makes customers feel products/services are handpicked for them based on their wants/needs)
• Products that drive the company’s value proposition (For LinkedIn, it’s their “People You May Know” or “Jobs You May Be Interested In” algorithms which drive further customer engagement)
• Products that facilitate an introduction to other products (to funnel customers into other relevant areas of your website and thus lower your bounce rates)
• Products that prevent dead ends (ex: smart algorithms that suggest other potential purchases, i.e. “People like you also bought…”)

And of course, many of the above “products” are more than simply focused on nebulous metrics such “customer engagement”—they can directly tie to revenue improvements.

There are even opportunities to drive news cycles with unlocked insights. Companies such as LinkedIn can use information gleaned from their web server farms to build press releases such as:  “Top Ten Phrases Recruiters Want to See” or “Top Ten Job Growth Areas in the United States”.  These kinds of press releases are interesting to local newspapers, bloggers and media outlets, especially if there’s a unique angle relevant to readers/viewers.

There are plenty of companies turning to outside firms, crowds, and even their own customers for innovation. And there’s certainly nothing wrong with any of these approaches. However, those approaches may be trying too hard – especially when there’s a goldmine waiting to be unleashed just a few web servers away.

Will Cloud Computing Change Your Business Model?

Cloud computing is changing the manner in which consumers and businesses buy, manage and use technology. However, the impact of cloud on technology providers is causing an even more pressing adjustment—as business models shift from simply selling and servicing technology to instead helping companies consume it.

The business model for plenty of technology companies hasn’t changed much over the last fifty to sixty years. Sell equipment or software, install it, provide a bit of training, and reap contracts from subscriptions and/or maintenance.  And if it isn’t broken don’t fix it, right?

The rise of cloud computing is changing this paradigm.

In an October 2011 report, Bo Di Muccio and Thomas Lah of TSIA Research suggest a drastic change is coming to technology service providers.  Instead of simply installing and managing technology (via shared or managed services), cloud computing will force companies help users “consume” or use technology to achieve business benefit.

Prior to cloud computing, companies buying technology had no choice but to accept “complexity”, say Di Muccio and Lah. To reap benefits of technology, business line managers enlisted system integrators or consulting firms to install, integrate and manage technology on their behalf.  In addition, companies had to write an upfront check (capex) for hardware, software, training and implementation.

Cloud changes this model.  Instead, as more business managers get comfortable with cloud and its inherent benefits, Di Muccio and Lah argue technology service providers will be forced to adopt a “consumption economics” model where they no longer receive payment for shipping, installing and servicing a box, but instead receive revenues based on usage (pay per use).

Di Muccio and Lah also mention that cloud based computing shifts “risk” from buyers to technology service companies. For example, in previous years a business line manager might pay half a million dollars to a vendor, whether he or she uses the technology or not. With cloud computing’s pay-per-use model, the risk shifts to the vendor which only gets paid when technology is consumed.

To be sure, the shift in mix of complexity vs. consumption is not occurring overnight. However, with adoption of cloud computing increasing exponentially, technology and service providers must make plans today to meet tomorrow’s business expectations.  This means development of new pricing models, products, services, skills and training to make companies more than “buyers” but also successful “consumers” of technology.

Questions:

• What new types of skills and services will be needed to help companies “consume” technology?
• Di Muccio and Lah say the move towards simpler cloud based products will reduce the need for core professional services. Agree or disagree?