Estimating software: a rule of thumb

I have an even simpler rule of thumb. Try as hard as you can to think "how long should this take". Then double it and a half.

It's never let me down.