Viva The Revolution in Predictive Analytics! A Shameless Parody of Bogart’s “Treasure of Sierra Madre” and Hemingway’s “Old Man & The Sea.”

Okay, I poke fun at the rising elite of consulting, i.e. those involved with predictive analytics or “the revolution in analytics.” Yet it raises serious issues around the capabilities and limitations of certain kinds of analytical modeling.  I am, by the way, a “fan” of quantitative methods …so keep smiling kick back and have a Corona.

The Old Man & The Blogger

The Blogger sat in The Cantina on a hot, lazy afternoon sipping a Corona.  Managers and marketing types gathered around him, hanging on his every word.   He spoke of “integrated business planning” and “executives getting fast answers to important questions.”   Then he crossed the line. “I think we’re about to witness a revolution in how companies use analytics in business processes. I don’t use that overworked term lightly. I expect this to be as revolutionary as the impact that client/server computing had on transaction processing and related systems such as ERP and CRM.” Oh, yeah?

Revolution!  The Old Man, a bearded Operations Research analyst in the back of the bar stirred from his siesta.  He moved slowly, almost painfully as ancient bones strained to come alive. But his eyes were coal-black, glinting with some sort of passion or maybe just crazy from the heat.  He tipped his sombrero and squinted..the eyes said it all: “I’ve been in the mountains with Fidel!”  A relic of the past hung from the Old Man’s belt, a HP RPN[1] calculator.

They said he belonged to the old Operations Research Society of America (ORSA).   They said when he has too much tequila he does matrix math with a pencil and paper.  He scares the undergraduates.  Maybe he had been in the Sierra del Escambray’s with Fidel doing some sort of weird linear programming or optimization?  Rumors?  No one really knows.

It’s About Insight    “Revolution in Analytics?  Revolution in Analytics?!  We don’t need no stinking Revolution in Analytics.” he mocked.   “Hey Amigo, we’ve been doing quant math and running yield management optimization programs since before you were born.”

The Old Man was getting more and more agitated…  “It ain’t about crunching numbers!  It’s about insight!  Ain’t nobody s-plained to you about complex adaptive systems or CAS? Eh?   No? Well you can kiss your assumption about equilibrium auto-correlation, and stationarity goodbye!   No more deterministic solutions!”  Getting in The Blogger’s face now: “Do you know anything about indeterminacy and heuristics?  Emergent properties? Yeah! I didn’t think so.”   The Old Man was on a roll now and there was no stopping him.  He was animated, a cross between Zorba the Greek and Hunter S. Thompson.  It was frightening but fascinating at the same time…we were spellbound.

 “You go down to Santa Fe (he meant the Santa Fe Institute which does research on CAS) and ask them what they think of predictive analytics!  And they will tell you that CAS solution landscapes DANCE!  Because the system adapts, it morphs, and it laughs at us!”

But now the Old Man was spent.  He was tired and needed rest. So he went down to the sea to watch the waves dance and was lost in the hubris of the past.

Hopeful but Cautious – Don’t “Pull A McNamara!”

The pros know the capabilities and limitation of analytical modeling.  They understand stationarity, emergent properties, and adaptive behavior in CAS, etc.  But others…well not so much.  And that can cause problems.

There’s a lot of hype going around about analytics.  And that’s to be expected in fields that re-emerge every twenty or so years into managerial mainstream.  We had Operations Research during World War Two, the RAND years of quantitative systems analysis during the Kennedy and Johnson Administrations, Chaos Theory and the hedge funds during the 1990’s, and so on.  When good data and good theory converge the results can be astounding, as in the airlines yield management and optimization innovations.   We also can look at predictive analytics in maintenance and equipment failures, and I recently wrote a piece reviewing a JAMA study on how performance analytics were reducing the risk of heart attacks mortality during hospital stays. Even though this is a “healthcare” orientation, the methodology and epistemology is brilliant and can serve as a template for virtually any industry.

Although I poke fun at combining quantitative analysis with the high levels of uncertainty accompanying CAS, I am optimistic we can accommodate both views simultaneously (see my CERTS posting).  However, when bad theory dominates, as in the RAND/ McNamara employment of Planning, Programming & Budgeting during the Vietnam War….well, in that case a lot of good people died. We can never let that sort of analytical hubris happen again…. see  Marke “Approaches to Risk Under Conditions of Uncertainty and Complexity” presented at the Society for Risk Analysis 2007 for a critique of McNamara’s approach.

So here’s the deal: if you position yourselves as a bunch of quants who do regression analysis or multivariate data analysis or whatever (being defined by the tool kit rather than the problem) you are doomed to work on relatively deterministic problems conforming to stationarity and living in equilibrium. Alternately, if you proceed to highly stochastic or indeterminate problems, the chances of hubristic failure are very high.  The two small power point pages at the bottom of this posting will graphically explain the typology. Remember the folks at Rand and don’t “Pull A McNamara.”

I think there is a middle ground.  It focuses on a multi-disciplinary approach that is not dominated by the tool kit.  Yeah, yeah, I know….you already spent a lot of money positioning this “new” service as “analytical.”  Re-positioning is gonna drive the marketing-types nuts.

Or maybe it’s just time to go down to the sea and watch the waves…

Here are slides from the SRA presentation (referenced and linked above)  that may prove helpful (click to enlarge). They were adapted from John Sutherland’s A General Systems Philosophy for the Social and Behavioral Sciences; 1973.  I am happy to discuss at length should there be interest, and I would also be happy to work with “quants” professionals who would like to incorporate complexity and CAS theory into their work.


[1] RPN standing for “Reverse Polish Notation”

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s