The Law of Averages

Here is a simple business problem. A 2-lane hand car wash business needs to determine the right number of people it should have to provide adequate service.  The business can employ between 2 and 8 people at a time depending on number of cars waiting. The problem is to determine how many service people to have at any given time because there would be ups and downs in the number of cars arriving, service times, service person availability etc.

Well, one easy way is to keep record of the number of cars, their arrival times, and service person’s availability, hours worked etc. then determine on average, how many service people should be required. Easy, right?  Err. Not really. It is actually a fairly complex stochastic operations research problem. You need all of the records to determine reliable car arrival rates, service times, absent records, disruptions, bottlenecks, weather, and a number of other pieces of information to have a reliable estimate to dynamically plan and allocate resources.  Yes, you can use averages but expect to get average results only. That is the law of averages. Sometimes too much and sometimes too little but on an average it will look right. There may be too many cars and not enough service people or too many people but not enough cars!

I mention this because recently I was made aware that the mining industry has very simple analytical problems for equipment operations and maintenance that are dealt with by using ‘simple’ calculations with average values. Really? Because when I think about the problem and the factors involved (including machine run times, breakdowns, maintenance history, operators and their productivity data, and a large number of environmental factors) that may impact this complex dynamic environment, it sounds pretty complex to me.

Sure, one can always use average numbers to solve these problems but as the law of averages dictates, the solution will be short of any meaningful outcome because most of the times it will be either above or below the mark. The proper way will be to have all of the detailed level data that may potentially impact the equipment operations or maintenance and use advanced analytics to get estimates with much higher confidence than mere averages.

That is the way Teradata customers typically will approach and solve their business problems. Get all of the relevant data at the most granular level in one place and use analytics to find the most suitable solution. How are you handling your business problems? Do average values give you as good a solution as you could have with integrated and detailed data? I would be interested to hear about any similar business challenges where average values are a good enough solution that you do not require detailed data.

Najmul Qureshi

Caption: Cartoon by Jeff Danziger DanzigerCartoons.com