The Agile MPG illusion

There's a very quick online quiz going around called the "MPG Illusion" which explains why I think EverydayAgile is so, so important.

Seth Godin's has written a variation which, I think, is a little easier to read, so I'll quote it. It starts:

Imagine that half the cars in the US get 10 miles per gallon. And half get 40 miles per gallon. Further stipulate that all cars are driven the same number of miles per year. Now, you get one wish.

You can give every low-mileage car a new set of spark plugs that will increase fuel efficiency by 5 mpg, up to 15. Or you can replace every 40 mpg car with a car that gets 75 mpg, an increase of 35 miles for every gallon driven.

Which is better?

Think about this for a minute.  You'll find the answer here.  Were you surprised?  I was.

Seth* continues:

As marketers, we're tempted to tweak the already tweaked, to turn the 100 to 101, to optimize for the peak performances.

There's a huge, huge, lesson here for the Agile community.

What if Seth's blog entry started:

Imagine if 90% of software development teams in the WORLD get 10 miles per gallon.  The remaining 10% get 40 miles per gallon. ...

Would you - if you wanted to make the world a better, richer place - focus on further optimising the the already efficient 10% of the population (a.k.a. the high hanging fruit) or would you focus on improving the remaining low hanging fruit?

Any thoughts?

* Forgive my use of Seth's first name and this BLATANT self aggrandizement, but Seth once referred to me (yes me!) as a "Purple Cow".  You may call him Mr Godin. :)

Comments

[Jim Bullock]

It's a bit more intuitive if you talk about gallons per mile or total gallons consumed. Once you say this:

"Further stipulate that all cars are driven the same number of miles per year."

It is about gallons spent. Most people leap right to miles traveled for an implicit fixed amount of fuel.The trick here is thinking about gallons spent in stead.

This is really a design problem in disguise. All optimization questions are really design problems. So, set up the trade-offs several different ways before leaping to a conclusion.