Sunday, February 21, 2016

Pervasive Management Loves Numbers, Not People

Data and analytics are buzz words in the C Suite.  Many firms are hiring Chief Data Officers (CDO).  That CDO is different than collatoralized debt obligations, but both have their root in complexity and mathematics.

Modeling is not new in the financial world.  Wall Street used statistical algorithms to destabilize trading exchanges and convince investors inherently risky products were safe.  Consider the case of David Li whose statistical models somehow made junk securities Triple A rated.

He just invented the model. Instead, we should blame the bankers who misinterpreted it. And even then, the real danger was created not because any given trader adopted it but because every trader did. In financial markets, everybody doing the same thing is the classic recipe for a bubble and inevitable bust.

Correlations are statistical associations which can change at any time.  They are not cause and effect.  Consider the Super Bowl winner's impact on the stock market.

You may not know it, but the winner of this year’s Super Bowl will be a pretty reliable predictor of how the stock market performs in 2016.  

This phenomenon has an amazing success rate: 82%. In fact, the winning NFL team has tracked how the markets fared the last seven years in a row, as well as for 40 of all 49 Super Bowl years.

Robert H. Stovall, an analyst who’s popularized the predictor, admits it has no grounding in fact. “There is no intellectual backing for this sort of thing,” he said in an interview with the Journal, “except that it works.”
Corporate executives do many things with data, a number of which also have no intellectual backing.  They optimize executive compensation to the detriment of the company and its employees.

HBR noted the latest management obsession and suggested asking questions about algorithms/models.

1. What was the source of your data?
2. How well do the sample data represent the population?
3. Does your data distribution include outliers? How did they affect the results?
4. What assumptions are behind your analysis? Might certain conditions render your assumptions and your model invalid?
5. Why did you decide on that particular analytical approach? What alternatives did you consider?
6. How likely is it that the independent variables are actually causing the changes in the dependent variable? Might other analyses establish causality more clearly?
Knowledge of variation, analytical vs. enumerative statistics and subject matter expertise are also critical.   It's important to know how numbers are generated/calculated.  It's important to look at data over time with an eye on improving quality.  And it's critical to remove threats, fear and intimidation so numbers can be honest and accurate.  Most important leadership should be in relationship with the people they support.

That hasn't been the case for quite some time in America's hallowed halls of leadership.  Management likes numbers that produce selfish rewards.  They prioritize additional debt and interest expenses over employee raises and maintaining benefits. They spend on faster computer systems and complex algorithms so they can deal with numbers, not real people. 

Senior executives hire consultants who know better than to return bad news.  Thus, they craft questions and surveys that limit employees and customers ability to give feedback because the C Suite really doesn't want to hear what people think.  Executives shoot the messenger when the rare brave soul penetrates the C Suite's Castle walls with salient observations.

Analytics enable executives to avoid interacting with pesky customers or employees.  Many senior managers will delight in replacing employees with algorithms and machine learning.  UBS predicts the Fourth Industrial Revolution will benefit the richest.  The more things change the more they stay the same.