Sensitivity Analysis: A Lost Art

I like numbers and statistics.  But as a tool, not the end all and be all.   Which is a good reason why my Masters was in Decision Support Systems.   So when looking at a new system or a new product I would do a sensitivity analysis for the financial outcomes which gave probability results.  Imagine my surprise when no one understood it at all!   I had greatly over-estimated the intelligence of management ( Scott Adams makes a living by going in the opposite direction). 

I've seen two types of behavior in this arena.   The first I call "Does it Stick?".  These managers hear something cool, throw it against the wall, if it doesn't stick, they keep doing it over and over again just knowing that one day they will get it right.   The second are the financially challenged.   They think by knowing some Excel spreadsheet functions:  sum, add, subtract, multiple, then Viola they have the insight into the future.   Everything is linear.   When asked about their assumptions there answer: "It's an assumption and I know it's wrong" and that's all the farther they get.  So they save the spreadsheet and use a fraction of a penny's worth of disk space and that's about what that model is worth.

What to do?

Compromise of course.   When you build out your model make sure you have three values for each assumption: Best Case, Worse Case, Most Likely Case.   Both on the revenue side and on the cost side.  Now you can have some fun.    First off run a table with everything turned Best Case, Worse Case, Most Likely Case.  You can see the spread of possibilities (put it into a graph).   Now just start flipping assumptions one at a time.    By looking at the resulting graph you'll get a really good idea what are your critical areas of influence.

Next iteration.   Turn everything to Most Likely Case and just run the three cases on your critical areas.    Chances are you will see your future.   Make sure you can survive on the downside, but never, ever manage to that downside because it becomes a self fulfilling prophecy.

And this will prevent you from making a typical rookie mistake.   You think that 20 good qualified customers will generate 3 orders.  Chances are the number is 200.   And this approach will keep you from going down that rat hole, because you will have a strategy to shift priorities if this becomes reality.

Comments

[Mary Schmidt]

Ah - and then you take it in the clueless empty suit CEO and he (or she) throws it all out and says" here's what I want." One reason I no longer work in big, bad Corporate America (and am increasingly picky about who I work with as a consultant.)

And - certainly - don't manage to worst case. But do keep something in the magic cookie jar - just in case. Particularly if you're in early start-up mode and gambling with your own money. Nobody does their best work if they worried about losing the family home.

[Felipe Bazon]

I had never thought of having three values for each one of my assumptions, I will keep that in mind for my next models.

[brucefryer]

Felipe,

I hope you do try this. It's also a great way to manage a client's expectations on a new program you may be suggesting.

[Felipe]

Bruce,

For sure I will be using it, by the way great blog first time I read and I will be around more often.