On This Day in 1983, Data Analytics Might Have Been a Fail

On 26 September 1983, Stanislav Petrov took a stand against what his systems were telling him and he may have changed the course of history.  Petrov was working as a duty officer at the command center for the Oko nuclear early warning system.  This is the place where the Soviets monitored incoming attacks, much like the US command center you remember from War Games.  Earlier that month, the Soviet Union shot down a Korean commercial jetliner over the Sea of Japan, claiming that it was on a spy mission.  269 people died in that incident, including a US Congressman.  Some at the Soviet Union were fearful of a retaliation strike by the US.  Cold War tensions were high.

At the command center, Petrov was getting data that a launch of five missiles had been made in the US towards the Soviet Union.  But instead of just reading that dashboard and acting he actually used his own inner analytics system to process the data and decide not to report or react.

Had Petrov reported incoming American missiles, his superiors might have launched an assault against the United States, precipitating a corresponding nuclear response from the United States. Petrov declared the system’s indications a false alarm. Later, it was apparent that he was right: no missiles were approaching and the computer detection system was malfunctioning. It was subsequently determined that the false alarms had been created by a rare alignment of sunlight on high-altitude clouds and the satellites’ Molniya orbits, an error later corrected by cross-referencing a geostationary satellite.^[5]

Petrov later indicated the influences in this decision included: that he was informed a U.S. strike would be all-out, so five missiles seemed an illogical start;^[1] that the launch detection system was new and, in his view, not yet wholly trustworthy; and that ground radars failed to pick up corroborative evidence, even after minutes of delay.^[6]

- Wikipedia contributors. "Stanislav Petrov." Wikipedia, The Free Encyclopedia. Wikipedia, The Free Encyclopedia, 26 Sep. 2012. Web. 26 Sep. 2012.

I’ve always wondered if the system he was using had a bunch of fancy dashboard features, like shiny 3D pie charts, moving average lines and drill down capable reports if he would have been able to not trust the data.  I’ve seen this sort of over-trust of data with data model diagrams.  It seems the prettier or more advanced the presentation of the data is, the more people want to believe it is right.  In fact, I’ve learned to present draft documents to people on my teams with hand-written notes/comments on them to sort of "break the ice" to show people that they are drafts.  A modern solution might have included some sort of decision making guidance that say "Confidence Factor of Attack: 99%" or something like that.  And it would have been highlighted by some sort of red bar, showing just how confident the system was based on the data – bad data, it turns out.

More details about Petrov and his actions in the video above from History.com

Stupidest Bar Chart of 2011 – Congrats, Klout!

I’m no expert in graphic design or data visualizations.  I do know that pie charts are evil, as are most gauges and uses of 3D in dashboards.   However, when I saw this bar chart from the fine folks at Klout, I knew it was a winner.  Even the name of this category is wrong: Top Influential companies?  No, from their description, this is a list of companies that were most talked about the most often.  That doesn’t make a person or an organization influential.  Otherwise, I’d be dressing like Kim Kardashian and collecting Justin Bieber dolls with weird "try me" portals. 

At best, these companies influenced people to say things about them by doing something well…or in the case of Netflix, doing a lot of things poorly.  But they didn’t do the influencing.  People on the social networks did.

This data visualization is a list of companies who were most talked about…so they’ve used a bar chart.  Bar charts are supposed to be used to show quantities, usually over time or some other measure.  But this ranking  is not a quantity.  In fact, the bar chart is emphasizing the wrong thing, too.  Notice how 11th place Facebook has more bar?  But it’s in the worst place in the list.  Or is it?  We don’t know because the bar chart is showing us conflicting information.  It could be that the top 11 companies are being ranked 11-1 in descending order based on their influence mentions.  From the list, I’d think think that the influence was descending, but I don’t know.  If someone doesn’t know and the visualization adds no more insight than just a list of the companies, don’t add a chart.

Jen Stirrup, BI expert and Microsoft MVP, has some more to say about bad bar charts. 

Don’t use bar charts to show rankings.  Put down your data viz tool and take a walk.  Notice the real world around you, then come back and think about using visualizations to help a reader understand the data better. That’s how to love your data.