Saturday, September 26, 2009

try forecasting this

In the 1976 edition of the Sunday league, the last playing day had three matches scheduled. The winner of the league would be decided only after all three matches were over. The problem was that there was only one trophy and it had to be given to the winner immediately after the match. All the three grounds (Cardiff, Edgbaston and Maidstone) were approximately 200 kms from each other.

BBC, the sponsor, hired a helicopter, and kept the trophy at their Birmingham studio and very close to Edgbaston. They decided to fly the trophy to Cardiff but changed plans and sent it to Maidstone. In a thrilling game, that went to the last ball, Somerset lost at Cardiff and the winner was Kent at Maidstone.

Read here for the complete article:

http://www.cricinfo.com/magazine/content/story/426659.html

Luckily, BBC's decision to send the trophy to Kent turned out to be right. Read the previous sentence again - the reason for the right decision was 'luck.' The last ball could easily have nicked to the third man for four runs and Somerset would have been the winner. They would then have to wait for almost 3 hours to get the trophy back from Maidstone. It could very well been a disastrous decision. Cricket experts would have surely written reports of how BBC should have sent the trophy to Cardiff and not to Maidstone.

In business, we face so many situations like this everyday. Every decision has some assumptions. If those assumptions come true the decision is hailed. If the decision does not come true, academicians like me will write a case study on how 'elementary business principles' had been violated.

I have always taught forecasting from a premise that it could always and easily go wrong. Investing in good forecasting systems is good, but effort also should be invested in creating agile systems that can survive inaccurate forecasts. We must never forget that most forecasting methods, whether time series of regression based have an inherent assumption that the past will repeat itself. Given the dynamic nature of markets, I am not sure if this is a valid suggestion.

And, at the end of the day its okay for forecasts to wrong. We could create a tracking signal and so long as the variation is within limits we should be okay. So, should the forecasting method be reviewed if the tracking signal is violated? As a good academician I will leave this question with a simple answer - "It depends".

One more thing, what would you do if you had to take the trophy decision in 1976.
1. Would you have housed the trophy in Edgbaston?
2. Would you have directed the helicopter to Maidstone on the basis of gut feel?
3. Do you have a third out of the box solution?