Watson – The Frustrated CFO

Are CBS’s Holmes & Watson Too “Elementary”?

Do you know that there are people who devote their spare and even professional time to collecting mistakes and goofs made in movies, TV shows, etc.? There is a successful British website Moviemistakes.com (since 1996) whose creator has built himself an entertainment career and a money-making vehicle doing just that.

Specialists officially distinguish eight classes of mistakes, including ridiculous audio problems and crew visibility. However, themost frequent ones are continuity errors (the yellow Porsche's side was smashed in the previous scene and then in the next one it's absolutely fine) and "revealing" mistakes that remind you it's not for real (Edward is seating under the bright sun during his honeymoon but his skin doesn't sparkle like "the skin of a killer" should).

I have to say, this type of bullshit is completely lost on me. I mean, how many times (and how intently) do you need to watch Commando to notice that thing about the car? And realism of Twilight? Pahlease! Unless I am watching Bergman, Fellini, Kubrick, Tarkovsky, Lee, or Noe, whose every shot is the result of conscious artistic effort, I swear, it's unlikely I will notice visual errors even if I absolutely love (or hate) the movie.

Plus, these mistakes are the consequences of poor production quality and low work standards, and my readers know very well that I expect that and discuss it all the time. Who the hell thinks that this prevailing trend of our lives doesn't apply to the entertainment industry? People are people everywhere. I don't even get surprised by plot holes anymore, even though it's impossible to ignore those. I'm like, "Oh, it doesn't tie well? Surprise, surprise!" You know how it is: the screenplays get nipped and tucked by everyone to such an extent that the story originators cannot even recognize their own creations anymore.

However, I feel differently about factual errors (an official class as well). I notice them all the time. You see, those don't come from producers, editors, the crew, and it's unlikely that actors ad lib them. No, they are products of sloppy writing. I guess I have different standards for writers than I do for everyone else: I get upset with idiotic mistakes made by screenwriters, journalists, novelists, as well as their research helpers and fact checkers. A writer's job is very hard – to construct a flawless plot is incredibly difficult. But to verify the correctness of some piece of information? That's a matter of care and respect for your audience. I take my time to watch or read your thing and you disrespect me? Fuck you!

Of course, different blunders create different levels of annoyance. I don't curse out loud (or at least, not anymore) about the eternal confusion of Chapter 11 and Chapter 7 bankruptcies. I got used to it. Someone always says how "the company has filed Chapter 11 and will be gone, like, tomorrow." Well, no: Chapter 11 means that the company plans to stay in business, already found funding, and is reorganizing itself. This is what Bloomingdale's (or rather the company that owns them – ah, never mind… ) successfully did back in 1991 and it is still operating, thank you very much. But if it were a Chapter 7 filing, then the company would probably be gone already. Of course, no David E. Kelley's show would allow an error like that, since he holds a J.D. degree from BU, but it is very prominent in many a police procedural.

Also pretty low on my scale of discontent are silly foreign-culture mistakes: Like giving a last name Petrovna, which is actually a patronymic and cannot possibly be a family name, to a Russian cyber-genius girl; or attributing a French chanson to the wrong chanteur who never sang it; or redrawing world borders by claiming a German town for Austria. Okay, I am an irritable person, so these things get me annoyed, especially because there are way too many of them. On the other hand, what else can I possibly expect from people who listen to Taylor Swift and Lorde? So, I grumble under my breath, as if I can telepathically correct these factual glitches, but that's about it. I wouldn't turn such bullshit into a blog post.

However, some facts are such a commonplace, they are so prominently a part of the collective consciousness that it is hard to imagine anybody having informational lapses about them.

Case in point: The episode 3.13 of CBS's contemporized take on the adventures of Sherlock Holmes, Elementary, revolves around a murder of a "debt collector." I probably can spend at least 10,000 words listing all factual errors that assaulted me during 40+ minutes of the episode, but this post is already running too long, so I will restrain myself and point out just a couple of things.

You know, how Conan Doyle famously attributed Holmes's success as a detective to his perfecting of the deductive reasoning (aka logical deduction)? Simply speaking, Sherlock starts with a broad premise and elaborately eliminates all possibilities until only one right conclusion remains.

The writers of the series tried to employ pretty much the same method: As soon as it is established that the victim was an attorney who lost his job at a big law firm and turned to debt collections as a source of his daily bread, the plot's deductive challenge is revealed: there are millions of suspects. According to the writers, every debtor in the victim's collection portfolio is a potential murderer and it's up to Holmes to find the proverbial needle.

You may think that this is a farce (hey, our life nowadays is a farce!), but no, this is a "serious" show! Then what about the materiality threshold? I mean, I know there are people who kill for less, but, seriously, what is the likelihood of someone in Nebraska coming to NYC to kill a collector for $10K? The size of the debts alone should've narrowed down the pool of suspects to a size of a soup bowl in a shorter time than it took Holmes to pin up all those endless lists of names on the wall.

Further on materiality: It is implicit that the debtor must've been severely threatened to go for a kill, right? But what's the big scare here? While larger debt recovery organizations employ or contract multiple lawyers pretty much in every state and therefore can litigate even small balances, it is impossible to imagine that one attorney can stretch himself over more than 20-30 cases at the same time.

All that is just common sense. Unfortunately, it is an unattainable commodity now. So, how about the technical side of the business that could've been easily researched? The reality is that no one can survive in the highly regulated asset-recovery industry without a decent collection software, expensive access to personal records database (such as Lexis Nexis), and on-demand automated call distribution solution (like LiveVox). You see, violations of collection rules and standards, including the timing and the language of calls, can be sufficient grounds for law suits. In fact, there are lawyers who specialize in suing collection agencies on behalf of debtors. Therefore, there is a necessity to digitally record all collection efforts for evidential reasons. Checking those records would've instantly limited the list of suspects even further - to only those potential murderers who's been actually contacted by the victim's firm.

My spirits were lifted a bit when Holmes correctly re-labeled the victim as a debt merchant, i.e. someone who buys, at a 95% discount, portfolios of consumer debts written off by financial institutions and makes money successfully collecting a small portion of them. My elation lasted exactly 2 seconds. You see, respectful writers who gave their viewing audience a modicum of intellectual credit would've left it at that: one has to be from Mars to be completely unaware of the concept.

But noooo! These hoodlums sent Holmes on a ranting explanation of the debt-trading plague's basics to… Watson. Most ridiculously, her scripted response to Holmes's briefing is that of bewilderment: "Really?"

Wait a minute! Wait a minute! Are you kidding me? Are you telling me that a sophisticated New Yorker, a former successful surgeon, and, at this point, a full-scope PI who lives and breathes research has never heard of sub-prime mortgages, multibillion-dollar bank write-offs, and consumer-debt securitization that led to 2009 global financial crisis and bailouts? Was she actually on an interstellar journey? Or were you, the writers, out to lunch?

Now, to connect the victim to the final suspect the writers had to create an unfathomable possibility of somebody being able to pluck out a specific bad debt portfolio containing a specific stale mortgage. Well, that would be like looking for a needle in a haystack for real! The delinquent debt industry is vast and sophisticated, with leaders such as SquareTwo Financial (owned by Collect America Holding) buying billion-dollar packages of charged-off receivables directly from financial institutions and distributing them for collections through their national franchises. After a time, uncollected accounts are further repackaged and resold with deeper discounts. To trace a single debt in this dark labyrinth would be absolutely impossible for the perp in question.

The part of this ignorant bullshit that turned out to be far more disturbing than the factual errors was the resolution of the case. At the end, Holmes deduces that the debt collector/debt merchant was killed because he realized what a "disgusting business" the collection of financial delinquencies was and tried to erase debtors' liabilities.

Note what the liberally confused writers found appalling: not the packaging of unrecoverable receivables as marketable instruments by opportunistic financial brokers; not the investment of people's savings and pensions into this imaginary "securities" by brainless money managers, but the straightforward effort of making consumers pay for goods, services, homes, etc. they bought and used. Hmm… Taking something out of a store and not paying for it – isn't it, like, shoplifting? Ordering something online and not paying for it – isn't it, like, mail fraud? Residing somewhere without paying for the space – isn't it, like, squatting?

I don't expect laymen to be fluent in the mayhem of American economics as it has been for the last 20 years. All I'm asking is a little awareness. Are all these people really this ignorant and stupid? It's too bad they don't read my blog. Four years ago (oh, my God!) in my post The Infinite Wisdom of Trey Parker and Matt Stone I was already referring the confused masses to South Park's episode Margaritaville (2009!!!) - the most genius breakdown of the financial crisis in popular culture. They are your colleagues, people! And they made it ELEMENTARY for you!

I'm just grateful that my knowledge of cryogenics, genetic re-breeding of extinct animals, cloning of rare plants, drone operations, and some other topics that feed (or are fed to) entertainment writers are only rudimentary. I am suspicious, but at least I can pretend that they may be represented correctly. But financial stuff? Culture stuff? I can't help myself there. I hear that doctors usually have conniptions when they are exposed to shows like ER, House M.D., etc. I totally understand: the idiotic errors – they are unbearable.

IBM Predictive Analytics, or Are We There Yet, Watson?

I have no clue what a Venn diagram for conjuction of this blog's readers and tennis fans would look like – probably unnoticeable to human eye, but whether you care about professional tennis or not, bear with me and I'll explain its relevance to the subject at hand.

Those who watched Australian Open during the last two weeks of January on ESPN2 (or any grand slam in the past few years) incidentally exposed themselves to various elements of IBM's "Let's Build a Smarter Planet" promotional campaign. IBM has been sponsoring ATP and WTA tours for years. They don't limit themselves to the ads that run during commercial breaks, but it's a good starting point for our topic.

The current versions of these TV spots are a lot of fun – very slick, very futuristic, very high-tech, very white-surface, and they feature Dominic Cooper (whom I first saw on stage [I cannot help myself, can I?] back in 2006 as Stuart Dakin in Alan Bennett's The History Boys)! Handsome, in the contemporary sense of the word, and serious, but tongue-in-cheek, British actor promotes IBM's cloud, smart day in the office, and… PREDICTIVE ANALYTICS.

Aha! Very interesting: Did IBM branch out into Astrology Charts and Crystal Balls? The advertisement's copywriters are appropriately non-specific and vague so not to bore an averagely challenged viewer, but let's look a bit closer at the core issue here.

Predictive Analytics belong to the realm of other popular concepts, such as business intelligence and performance assessment – none of them new ideas, by the way. Data warehousing may sound like a novelty, but collecting and organizing records in a particular order for easier access existed for centuries. The concept of information as a key to business success is millennia old. In ancient times tribal chiefs were estimating how many spearheads their craftsmen needed to make in order to win a battle.

Yet, in the past 10-15 years, the obsession with gauges, graphs, and all other forms of Key Performance Indicators (KPIs) have become pervasive to the point of silliness, especially in business, financial, and data-management spheres. Everyone and their mothers are absolutely convinced that they MUST have KPIs or they will be running a danger of failing, while analysts and software developers just know that they MUST provide KPI capabilities or their services will be deemed obsolete.

Truth be told, the escalation of urgency is totally understandable. The pace and
vulnerability of the business environment and entire human existence have increased exponentially. Today, more than ever, we need to have timely and valid data that has a power of springing us into immediate corrective actions.

Of course, as it always happens, the form obscures substance and common sense. In most cases, the procurers of reports with colorful dials, bars, and pyramids end up just staring at pretty pictures, nothing more.

Now, the Predictive Analytics are a very special brand of information manipulation because they claim that on the other side of chewing up and digesting tons of historical data they can poop out specific recommendations for WHAT NEEDS TO BE DONE IN ORDER TO SUCCEED, i.e. an action plan for a prosperous future.

And I have to say, when it comes to chomping massive inputs of data, while allowing flexible and customized outputs, various IBM Business Intelligence Suites, including Cognos, are probably the best performance analytics software out there in the land of information management - terribly expensive, of course, but awfully powerful in terms of the facts-and-figures consumption (imagine Coneheads at breakfast).

And as noted, IBM wants to take you a step further on the road to the future of business intelligence. In 2009 the computing giant bought SPSS, which stands for Statistical Package for the Social Science, a software product that was created in 1970 to deal specifically with the analysis of what I personally call “data with a psychological twist.” In other word, it digests information that doesn’t necessarily have dollar signs or volume units attached to it. It was widely used in sociology, marketing, health care, education, and government.

In its current developmental stage SPSS has become an integral part of IBM's predictive analytics solutions, which presumably can be applied in any field. According to IBM’s own description, the SPSS’s mission is “to help organizations to predict what will happen next, so that they can make smarter decisions to improve business outcomes."

Are they succeeding in this futuristic endeavor? It’s hard to tell, because there is not enough readily available feedback: everyone’s KPIs are proprietary. Even public companies will not disclose how IBM’s predictive analytics compare to the actual results. Fortunately, there is one application of IBM’s analytical intelligence that is very public. Here's where tennis comes back into the picture, offering us an opportunity to look at the system’s predictive aptitude.

Nowadays, IBM's sponsorship of the Grand Slams also includes the powering of the websites of individual tournaments, including the Wimbledon, Australian, French, and US Opens. And so, if you go on one of these sites’ home page you will see right there, in the top-right corner: Smarter Analytics by IBM.

Well, right now it says “Completed Matches” above it because we are still almost 7 months away from the US Open 2015, but if you were to track a live match during a tournament, you can observe the IBM’s predictive functionality in real time:

This here is one of the most important predictive KPIs provided by Smarter Analytics – it is designed to show most important milestones one must achieve to reach one’s goals, and supposedly it is applicable to any data set.

In tennis application they call them Keys to the Match. Right there on the screen, in just two sentences IBM delivers the gist of the chart, explaining both the terminology and the methodology: “Keys to the Match system identifies key performance indicators – what players need to do to succeed in a match. Each player’s performance is measured against their keys and updated in real time.”

This one in particular was for the 2014 Championship match between our very own champion extraordinaire Serena Williams and Caroline Wozniacki. The system isolated three performance keys for Serena: 1^st serve points, medium rallies, and the returns of the opponent’s 1^st serve.

According to the system’s algorithm these parameters are the most impactful in terms of the winning statistics. The calculated "musts" are in blue and according to the statistical analysis of the historical data, Serena won 74% of sets when she won more than 61% of points on her own serve; 87% of sets when she won more than 49% of medium rallies, and 81% of sets when she won more than 38% of points returning her opponents 1^st serve.

The red sectors show where Serena was in this match. She exceeded all requirements: reaching 62% in the first parameter (very close), 54% in the second (also pretty relevant), and 68% in the third. It is no surprise that she won the Championship match, but that wild discrepancy between the prediction and the actual in the third indicator (38% vs. 68%) makes me question its relevance.

We will come back to my doubts about the validity of the entire model in a moment. Now, let’s see how we would apply the same approach to a business. Here is a chart that identifies AZ Company’s Keys to a Successful Month: what needs to be done to achieve sufficient profitability and positive cash flow.

AZ Company's Keys (640x495) These are very familiar to all business runners and highly vital parameters. Product mix is a crucial profitability factor. The same goes for the portion of the gross profit eaten up by the overhead. And, of course, the speed with which we manage to turn our inventories into receivables and receivables into cash determines whether we can generate more cash than we disburse during a particular period, in other words, produce a positive cash flow.

The green bars represent the levels of the parameters at the moment this snapshot was taken. Orange ones are the objectives that simply must be achieved, and pinks are desirable targets that supposedly guarantee a financially successful performance.

According to the model’s algorithm the share of the highly profitable product X in the
trading mix should be at least 39%, because 69% of the months when such mark
was achieved were profitable. Obviously the overhead has far more definitive impact on the bottom line – the model confirms it by calculating a higher probability of success (79%) at its relatively lower levels. There is even closer interrelationship between receivables and inventory turnovers on one side and the positive cash flow on the other – respectively 80 and 85 percent.

Many business owners and executive managers, when they see a chart like that, get very excited by the prospective of having a system that can “see the future” – they think they’ve got the ultimate solution in their hands. The widespread assumption is that Technology is smarter than a human, calculates everything faster and with higher accuracy. And this application presumably eliminates the need to process information yourself – analyze, ponder, be anxious, listen to your gut feelings, or rely on anybody’s expertise. Just follow the model’s suggestions and everything will be fine.

If you sense sarcasm between the lines it’s because I am very resentful to such a blind reverence of computing technology. And coming from me it’s a very serious statement, because I love progressive computerization. But I cannot tolerate the lack of common sense.

Look, the selected parameters themselves are great – no question about it, but do we need a fancy program to tell us that the business is going to be okay if we push sales of a high-margin product to nearly 40% of the volume and turn most of our beginning-of-the-month AR into cash? I don’t think so. The years of our own expertise will kick in and reassure us – we don’t really need the statistical confirmation of the possibly successful outcome.

But I have even more troubling concerns about this “keys to success” model. Let me show you that the examples we just reviewed are fraught with at least two serious problems.

The first one is the size of the statistical sample. In quantitative research, you need a sufficiently large sample in order to be able to pinpoint trends with an acceptable level of realism. Did you notice that Serena’s KPIs were related
to sets rather than matches? That's understandable. The Smarter Analytics uses just 8 years of grand slam performance. In the period from 2007 through 2014 Serena Williams played 157 matches in the four major tournaments of the year.

Women play “win-2-out-of -3-sets” matches. Sometimes it’s 2 sets, sometimes it’s 3, and sometimes your opponent retires before you get to finish the first set. Serena has a high percentage of straight-set wins, so we can safely estimate her average at 2.25 sets per match. This means that the predictions produced by IBM are based on a pool of
data collected over 353 sets. It’s not US census, of course, but it's an okay sample size.

Now, in business, to approximate the tennis-model’s reliability of conclusions we would need to look at least at 30 years of a company's monthly performance data. And that's a lot to hope for: 25 years ago the majority of small to mid-size companies were not even computerized and the paper journals and ledgers are either rotting in some storage unit or gone!

But the bigger problem is the quality of the historical data, its relevance to the very specific, very present moment in time. We operate in remarkably dynamic environments. Business conditions change every moment. One day everything is going great and another day everything falls apart. Labor costs grow exponentially in every corner of the world. Foreign exchange policies force one currency to soar and another currency to drop below everyone’s expectations. Cyclicality shifts all the time: what was true for February of 2005 may not be applicable at all for this month. And there are multitudes of industry-specific stressors.

For decades Kodak was competing with BASF, TDK, etc. for the worldwide market share of film distribution. Who knew that the biggest challenge would come from the photo and video equipment that didn’t need any film at all?

20 years ago American agricultural sector exported nearly 3 million tons of frozen meat and poultry to Eastern Europe. Then came 1999 and protectionist governments in the region declared US produce unsanitary. Prices tanked and multiple industries contracted.

15 years ago fashion industry had six major seasons. Today it operates in micro-cycles with styles changing every two weeks. This dramatically altered the landscape of apparel importation.

Well, business insiders live and breath this knowledge, but does the prediction model take into account all these factors? It simply cannot – any mathematical algorithm can only account for a limited number of time-proven constraints. How can we presume then that what might have been true historically will apply to our current conditions?

Even Serena’s keys to success, as far as I am concerned, are not too trustworthy, even though they are based on the game that have been played by the same rules year after year, on exactly the same courts, tended in exactly the same way. But was it the same winning Serena Williams in September 2014 as she was in 2007, when she fell away in quarterfinals? Or was she even the same in January of 2014, at Australian Open? Can you spot the difference that made a significant improvement to her swing?

Can IMB’s "social" software comprehend that? I don't think so!

There is no way I'm putting away the years of expertise and my business instincts to start relying 100% on some computerized predictions in my strategic and tactical decision-making. It may be a good contributing tool (again if you can afford it), but a computer is nothing more than a sidekick to the human brain; even if it's the IBM’s star artificial-intelligence child Watson. Hence, the appropriate name it's given. Yes, it can answer questions posed in natural language and wins Jeopardy! thanks to the 4 terabytes of information stored inside its metal guts. Yet it's not able to intuit the right response to a simple clue “This hat is elementary, my dear contestant!” My mind, on the other hand, serves up the answer instantaneously.

An expert with a common sense should be able to formulate her conclusions and make fast decisions herself, without any magical software, as long as she is steadily provided with timely and relevant information – whatever it is: business, sports, or arts. It's as I always say: give me sensible data and I will tell you what needs to be done.

Share this:

Share this: