Subscribe to DSC Newsletter

How to Avoid Political Blunders in Analytical Discussions

A while back I was running a data mining project for a customer and made a conversational blunder. In one of the meetings, I mentioned seeing one interesting relationship in the data. Customers who purchased one particular product tended to buy and implement a second product at a later time. I did not realize that Everyone in the room INTUITIVELY knew that there is absolutely no relationship between the two products. A big blunder. After the meeting, two friends told me that my standing in the whole organization decreased because of that one statement.

Two lessons are worth mentioning here. First, know your audience. Second, and more importantly, know how to influence the amount of freedom in the room and increase it before making a statement like I made. My mistake was not in stating the relationship in the data; it was the timing of the statement. I could have done more to prepare the audience so that the statement would have been appreciated more. This preparation consists of increasing the amount of freedom of expression in the meeting.

Freedom of expression is critical to thinking creatively during an open discussion. How do we increase freedom of expression? We increase it in two ways that corresponds to the two forms of freedom: Positive and Negative. Both are good, and both are needed. Positive Freedom is the ability to change under your own power. For example, a car with an empty gas tank has low Positive Freedom. Negative Freedom is the lack of obstacles in your way. A car stuck in the mud, deep the woods, away from the road has almost no Negative Freedom. A car with gasoline and on the freeway has both forms of freedom.

Social discourse, which is critical to problem solving and innovation, needs both kinds of freedom, just like our example of the car. Positive Freedom comes from information; current, trusted, applicable information. This information is needed for the conversation to have importance. This potential impact is what makes Data Science important. Information is fuel to a conversation as gasoline is to a car. It creates Positive Freedom.

Positive Freedom alone is useless and can be damaging. I thought I was adding to the conversation when I made my blunder about the relationship in the data. I did not realize that the low amount of Negative Freedom in the room created a hostile environment. It was like driving a car into a tree: ample Positive Freedom but no Negative Freedom. I first needed to work on the Negative Freedom in the room, then work on the Positive Freedom.

When in a meeting, watch who speaks and when. Are only a few people allowed to ask questions? Is anyone ever discounted? Do you hear words and phrases like “obviously”, “we all know that...”, “never”, and “not possible”? When this happens, your job is to encourage others to speak out, to give credit to creative statements, and to acknowledge innovative ideas. Support others and acknowledge brave statements. Spend time removing obstacles, increasing Negative Freedom, and encouraging others to participate. Then, and only then, add fuel to the conversation.

By managing the amount of both Positive and Negative Freedom in the room, your chances of successfully influencing the discussion and resulting decisions from your work as a Data Scientist improves. Your impact on the organization on the decisions, as well as the process of decision-making, becomes positive and compounded.

Views: 1905

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

Comment by Stephen Penn, DM, PMP on October 17, 2016 at 2:57pm

Hi Thomas, 

Thanks for the comment.  I do appreciate the feedback, but I can't help but think you misread my post.  I agree with everything you wrote beyond the first paragraph.  

I wasn't saying that my mistake was not trusting the data.  I was saying that my mistake was not helping others challenge their own assumptions.  I wish I had helped my customers better understand the power of data mining prior to showing them the results of my work. 

My research into what was happening was strong.  My conclusions were also well supported. However, my internal customers weren't ready for their own "common sense" to be challenged. 

A data analyst (or data scientist) has a lot of power when backed up by the data.  We need to appreciate that power and the office politics that surrounds us. You admit this when you wrote the word "tactfully."  You also admit that there are widely accepted beliefs in organizations that do not stand up to the data. I agree. That's why we have concepts like Evidence Based Management. 

You and I agree.  If you don't trust me when I say that, please re-read my post. 

Comment by Thomas W Dinsmore on January 9, 2016 at 4:36pm

Stephen, your approach is dead wrong and I question whether analytics is the right career for you.

Just because "everyone" in the room "intuitively knows" that there is "absolutely no relationship" between two products does not mean that there is, in fact, no relationship between the two.  There are many widely accepted beliefs in organizations that do not stand up to the data.  That is why we use data.

Data can be wrong, of course, and when data conflicts with closely held beliefs one should not assume that the data is correct.  The first thing you do in this circumstance is check and make sure that your data is correct.

Assuming that you checked your facts, you did not "blunder" in pointing out an association between the two products -- as long as you did not equate correlation with causation.  The relationship between the two may be spurious, or it could be real, but it needs to be considered.  

It is your job to point it out relationships in the data -- tactfully, if necessary.  If you simply cave to the conventional wisdom, you failed.

 

Comment by Sam Kaplan on August 5, 2013 at 3:47am

Great post--great comments.  The conflict between truth and prejudicial error is universal and eternal.  Not too many people are successful in overcoming prejudicial  views, particularly in a meeting or two--they are too firmly entrenched and regularly reinforced.  In the case you cite, they  were socially reinforced as a group norm.  So, while it is a  great posting, it is no grounds for kicking yourself.  My tendency would be to preface my presentation by asking for an open mind--some of my findings may be counter to the perceived wisdom, but if we are not willing to challenge the perceived wisdom, why are we here?  Progress in society and in any organization that is part of society requires challenging old knowledge--otherwise  we still believe that the sun  circles the earth.   So, as  we go forward, let us make a special effort to keep an open mind, and be willing to present data to support any opinion that runs counter to the  findings.  Are you willing to do that?

Comment by Stephen Penn, DM, PMP on August 5, 2013 at 3:05am

Hi Chris,

Feel free to quote from the blog.  I'm here to help. 

Yes, assumptions do get in the way.  Many times assumptions help reduce complexity, but when data is available, we should challenge our assumptions..and as you state, stop living in the illusion.

=-=-=-=-=-

Hi Wayne,

You are right in that we are all humans and all have our limitations. That proves to me why diversity within a group is so important. Anything that gets the discussion going, ideas generated, and assumptions challenged is good.  I like your phrase of "exploring the operating space."

=-=-=-=-=-

Hi Mark,

Thanks for the encouraging words. I agree with you about interviewing. In my experience, the hiring managers that do not have a quantitative background want more than just the technical skills.  They want interpretation of data as well.  But that "interpretation" requires two things: 1) a willingness on their part to include the analyst in understanding the business operations, 2) excellent communication skills of the analyst.  Your observation that an appreciation for the social dynamics is rarely seen in interviews is true in my experience as well.

=-=-=-=

Hi Fari,

Thanks for the compliment.  And thanks for the reference to Kahneman's book. Your second paragraph is very intriguing. I will have to look that up.

 

In studying for my doctorate over the last few years, I wanted to study the problem where a data miner proves something and yet conventional wisdom overrides the proof.  For illustration, imagine a scenario  where a data miner works on maximizing profit through catalog sales. The data miner shows that not every potential customer should receive a catalog. And yet, the CEO responds, "We send everyone a catalog!" and all that work of the data miner is wasted.

I completed my doctoral in management last year. My dissertation combined the typical business intelligence objects, such as data warehousing and master data management, with factors from social discourse and group dynamics. I'm hoping to continue posting on this blog with ideas that come from my dissertation.   

 

-Steve

 

 

Comment by Fari Payandeh on August 2, 2013 at 5:01pm

Stephen,

excellent post!

The following is a summary of Daniel Kahneman's (Nobel Prize winner in Economic Sciences) book, Thinking Fast And Slow. "In judging the world around us, we use two mental systems: Fast and Slow. The Fast system (System 1) is mostly unconscious and makes snap judgments based on our past experiences and emotions. When we use this system we are as likely to be wrong as right. The Slow system (System 2) is rational, conscious and slow. They work together to provide us a view of the world around us".

In short, according to what you recommend, we shouldn't shock people's system 1. We need to prepare it first so that system 2 has a chance to  get to work.

Comment by Mark Meloon on August 1, 2013 at 9:51am

Really happy to see a post like this. I'm currently looking for a new position in data science. All the applications and interviewers are focused on questions like "How many years experience do you have with R?" Fine, but there are some skills, like political savvy and a good feel for social dynamics, that are critical, difficult to learn, and very rare in applicants. Oddly enough, these don't seem to come up in the application process. So I am doubly-glad to see you highlighting an oft-overlooked attribute of a good data scientist. I'll share your post.

Comment by Wayne G Fischer on August 1, 2013 at 8:12am

"Everyone in the room INTUITIVELY knew that there is absolutely no relationship between the two products."

And everybody in the room may well have been simply...wrong!  Surely you've many times discovered from analyzing data, something that conventional "wisdom" denied could be possible?  I have.  These myths typically get promulgated by "the experts" who, because (like all humans) they can only assimilate the main effects 5 or so factors at a time, usually have *no* comprehension of the effects of interactions (two-factor, three-factor, ...), don't even contemplate quadratic effects, and have not explored the total available operating space, lack a tremendous amount of understanding about their multi-dimensional, nonlinear, dynamic (and maybe even adaptive) system.

One way to handle the “obviously,” “we all know that...,” “never,” and “not possible” nay-sayers is have them all write down their current knowledge of the system...*before* you do your data analyses.  *Then* present your findings and facilitate the ensuing discussion.   :-)

Comment by Chris on July 31, 2013 at 11:19am

Thanks Stephen, for taking time to spell out the confusion that help clarity in handling data during this transitional period where leadership is vacant. I hope you would not mind quoting your words in my future blogs. (won't do if you don't allow)

I agree with you with my own terms, ideology vs. reality (yours is positive & negative freedom).  We are so blessed with a rich past history of Constitution and Amendments, and we did share our wealth to Africa through USAID. Now living in the post modern era, ideology must back up with reality, can we afford what we want or shall we keep living in illusion?

Videos

  • Add Videos
  • View All

© 2019   Data Science Central ®   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service