Category Archives: statistics

PowerPoint is finished when there is nothing left to remove


Inspiration for your 2008 board presentations: The Minard Map shows the attack of Napoleon’s army, from 1812-1813, as the defending Russians scorched the earth and anything else that might feed or shelter the French. Edward Tufte lionized this graphic for its use of six different dimensions: the width of the line shows the number of French remaining as they march into and out of Russia; color of line shows direction in and out; the line itself shows the course of the march; the underlying map shows geography; and the scale at bottom shows both the frigid winter temperatures and time of the return trip.

Here’s a New Year’s challenge: Make at least one 30-minute PowerPoint presentation with only 1 slide, make the slide this beautiful, and see how it improves your story.

What will Google do with a time machine for world health?


The concept is brilliantly simple: The best way to understand data is to animate it so you can watch trends over time. Go to Gapminder and you can plot variables against each other, and watch bouncing balls shift and wobble as the years roll by.

Gapminder is really Trendalyzer, software developed by Swede Hans Rosling to plot world health trends. Google purchased the software back in March, and we’re waiting to hear if Google will do anything with it other than juice its web analytics tools. Rosling was just honored by Discover Magazine as a Notable Scientist of the Year.

The way data jumps off the page is quite startling. The above screen shot, for example, is one part of an animation showing how faster economic growth in China vs. India correlates with increases in carbon emissions. The software also helps assess if there is no connection between two variables; liberals in the United States will be pleased to find that increases in military spending appear to have no impact on U.S. fertility rates.

Until Google releases an update, healthcare policymakers can be tempted with limited data sets here.

Predict all you want, but remember: There’s no free parking


OK, some bozo ran billions of computer simulations to discover how to win at Monopoly. The foolproof way to beat the odds — stop reading now if you ever want to recapture the childlike joy of snaring Boardwalk again — is to buy railroads, not utilities, collect a same-color three-set of properties on sides 1 and 2, not sides 3 or 4, and quickly build out to three houses, no more, no less. This tips the odds in your favor until the cash cascade crushes your opponents.

Dang. We love taking such mathematical forecasts into media planning, but it occurs to us that trying to predict everything squeezes some lifeblood out of the Marketing game. We once knew a CMO who would never make any move without spreadsheets proving to him the ROI. He didn’t move much … and eventually he got moved out. Predictions are necessary, but what about gut and inspiration?

Malcolm Gladwell writes in Blink about rapid cognition, or the theory that human intuition is actually completely rational, a fast product of our deeper mind processing data and drawing conclusions in a way that our immediate consciousness cannot fathom. It’s why you jerk the steering wheel away from an accident in a rain storm before you understand the danger, or why art experts can detect fraudulent carvings in an instant when they can’t really explain why. Marketers caught in their calculators should pause occasionally to consider what they really think — yes, results are down, and the media plan may have issues, but … does the creative just suck? Have you thought that the ad message itself is simply not breaking through? Is a competitor making a huge move that is sinking your own little ad campaign?

So bring on the Monopoly math. Sometimes you have to predict odds carefully. And other times, you just need a 7-year-old who joyfully knows how to glance at the big board, and then roll the dice.

Correlation does not equal causation. Smokers rejoice.


If you dig data, you could still choke on the 25-pound Historical Statistics of the United States, available for only $940.50 at Amazon.com. P. J. O’Rourke dissects it in The Atlantic pointing out the problem with data. Americans are eating less red meat and more vegetables, but still getting fatter. Divorces are down, but suicides are up. And — get this — from 1973 to 1994, smoking rates fell from 4,148 cigarettes per capita to 2,493, yet lung and bronchial cancer diagnoses are up 34%.

The message: Stopping smoking is dangerous, because without cigarettes you may get cancer.

The trouble with data like this is it is too easy to jump to conclusions, and to assume a change in A must drive a resulting change in B. Marketers run into this all the time measuring campaign results. Three common mistakes in advertising measurement are (a) setting metrics up too broadly so you cannot accurately track individual responses to individual ads, (b) neglecting to consider the impact of competitors on your results, or (c) failing to evaluate how shifts in ad channels affect each other. We’ve seen many clients with declines in print response AND increases in internet response who cannot connect these dots. In other cases, we’ve seen advertising results fall for no apparent reason — until we begin tracking the major competitor moves whose gravity is causing our client’s prospects to swing out of orbit.

Measurement is hazardous, because the wrong assumptions can lead to the wrong decision. Cause and effect, or effects with no cause? Until we figure it out, it’s probably best to stop smoking.