November 08, 2012

Simulations

This image tells us a lot, doesn't it?

The blue line is the prediction of a United States forecasting model, five days prior to Sandy making landfall.

The red line is the prediction of a European forecasting model, five days prior to Sandy making landfall.

The white line is what actually happened.

Both forecasts apply complex laws of physics.  Based on assumptions and initial conditions, the forecasts diverge, quickly.  Neither forecast was right, but the European forecast yielded the same outcome, and that's what matters.

You can hardly blame forecasters for trusting the "US GFS" model forecast above ... I mean, when does a hurricane abruptly turn left and head directly into the prevailing winds?

It almost makes you wonder how blind we are to "left turns" in marketing?

Speaking of marketing ... we read stuff like this on Twitter, nearly every day.
• By 2017, Woodside Research predicts that 80% of e-commerce transactions will take place on mobile devices, with Facebook influencing 50% of all purchases.

Wow!

You've observed how many marketing projections are generated, right?  I have.  We see a trend that looks something like this:
• 2009 = 2%.
• 2010 = 4%.
• 2011 = 8%.
• 2012 = 12%.
Based on this relationship, we moisten our right hand thumb, put it in the air, and derive an estimate of the future.
• 2013 = 18%.
• 2014 = 27%.
• 2015 = 40%.
• 2016 = 60%.
• 2017 = 80%.
And then folks publish the forecast, making people pay \$495 for access to the information.

Have you ever witnessed how e-commerce sales predictions are often made?  We take traffic, conversion, and average order as our key metrics.  They we start guessing.
• 2012:  Traffic = 10,000,000 ... Conversion Rate = 5% ... AOV = \$100 ... Demand = \$50,000,000.
• 2013:  We'll guess that traffic increases by 5% ... and our online marketing folks have all sorts of tricks up their sleeves to increase conversion, so we'll bump that up to 5.5%, and average order value will increase to \$105 because of our cross-selling strategies ... so 10,500,000 * 0.055 * \$105 = \$60,637,500.
Then we buy merchandise for the year, and wonder why we are seriously overstocked three months into the new fiscal year, having to discount everything to "move inventory"?

In the real world, simulations play a huge role in shaping industries.  In weather, slightly different initial conditions are simulated, so that we can see how forecasts vary.  Pilots are challenged to unusual conditions in flight simulators, giving them practice that they can apply in actual situations.  Population Biologists simulate the impact of human factors and weather on animal populations (i.e. what happens to the insect population if bat populations are reduced by 50%).  Traffic analysts simulate what might happen if a major freeway is closed during rush hour.  Governments frequently execute "war games", simulating combustible situations without putting lives at risk.

One wonders why we accept such flimsy forecasting tools in marketing, when so many others in so many professions happily apply simulation tools to their daily jobs?

It might be time for us to focus on what simulations can do to help us see the future, don't you think?