06 Just give me a number

The boss doesn’t want a range.  Doesn’t want to know about risk nor probabilities.  Fair enough, I guess.  But we may demonstrate that it is in his/her interest to consider the bigger picture in order to stay out of trouble with project projections/ predictions.  Below is a light hearted attempt to deal with this one.

Suppose we have put together some solid work and run an analysis: it is 30% likely that we could complete a piece of the work in six months for $800,000.  The 50% run delivered eight months and a million dollars.  We are 95% sure to complete within eleven months and for $1,150,000.

Give a number.  What would happen if we present eleven months and $1,150,000?  Perhaps the response: “you are planning for failure”.  Suppose we present the six months and $800,000,  possible response,“can you prove to me that it can’t be done quicker” along with the sinking feeling in your own gut that comes from a 70% chance that you will be reporting bad news soon.  Even the 50/50 doesn’t resolve equitably, the problem is with the concept of “just give me a number”.

Now, lots of leading companies present a Project Management 101 course to top executives not otherwise experienced in this realm, and within that course would be a chance for such a presentation.  I know Chevron was an early adopter of this ten years ago, and Exxon has come on board, but let’s assume your company has not.  Lat’s further assume that your boss has a lot on the plate other than projects, and staff has been giving him/her “just a number” for all his/her career.

My solution: I would invite your boss over to the house for a barbeque steak dinner.  And you will want to have good data from having cooked lots of steaks all to specified doneness using an instant read digital thermometer, the ThermoPen.  Roll the tape:

Insert graphic

“Hey, boss, how do you like your steaks?”  “I don’t know, medium well will do”.

Response Option 1 (the geek): Great, so you’re not calibrated.  I’ve got some pictures, cross sections of finished steaks, color glossies at 150% life size showing how much is pink, is red, is done, want to look at these?”  “No, and what the hell do you mean by not calibrated”

Response Option 2 (the trained palette): Okay.  Let’s talk about taste and texture.  Do you like a sweet steak which is very juicy and with a center texture like butter.  Or do you like just a spritzer of juice in each bite, burgundy/ruby red but not pink, tender but without a hint of sushi-texture in the center.  Or maybe nice and firm, a hint of red center, only a trace of juice.

Response Option 3 (the empiricist): Medium is nice.  I’ll put one right on.  Just for fun, I’ll make a couple extras, one a little less done, one a little more done.  Taste ‘em and pick the one you like.  I’ll use the discards for sandwiches next week.

Next Step: you’ve smartly gone with option 3; he/she liked the one a little less done.

“Glad you liked your steak, boss.  And you know what, I can cook it just that way every time.”

“Of course, you know how long you cooked it”

“Yeah, but sometimes the fire is hotter, sometimes the steak is thicker or thinner, even moisture  content, muscle structure, fat marbling can vary. Remembering only the cooking time is not sufficient; I use a thermometer.”

“Yeah good.  What’s my number?  130?  140?

“Very good question.  The number I wrote down, and will allow most anyone with my sort of stuff to cook it exactly the same as you liked today, is 152.141.133.124.122.

“Wha?”

“Yeah, it is a series.  I use a Thermopen, which is very accurate and reads instantaneously.  The series of numbers is my reading from three points, outside in to the center, a couple of minutes before finishing, and then mid-point and center at completion.  In the past, I used a conventional thermometer, which gives sort of an average reading, which is misleading because as you can see, temperature varies across and within the steak.  I could take a reading with a conventional thermometer today, and during testing I did so, and it might give me a 130 or a 140 or something in between.  Not replicable.  Steaks within that range may or may not be the same as today’s steak”

“Sounds complicated”

“It was at first.  But my buddy who turned me on to this thermopen also gave me his cheat sheets about what temperatures he likes for various doneness of steaks, grilled salmon, chicken, ribs, the whole shot.  And I have jotted down my own notes as I cook.  Not all that tough, actually a heck of a lot of fun, and my guests generally get food pretty much the way they like it.  Especially the second or third time they’re back.  And even before I have honed in on each guests particular tastes, still the cooking is consistent.”

“Hey boss, let me take your plate.  Another glass of wine?  And, hey, can we talk a little business -  the number for the Alpha Project”

“Yes, yes and sure why not.”

“Uh, I also have a series of numbers for that, and I would like to know your taste for them”

“Wha?”

“I have a model of all the best data and its range of uncertainties and unpredictabilities for this project.  And from that, I can give you a number that will match your taste.

“Whaddya mean, taste”.

My guess is that you like early but not unachievable targets.  So do I.  That’s why when you ask for a number I usually give you the 30% probability number.  I like the challenge of being significantly better than average, and like setting these sorts of goals to the designers and builders.  We can continue to go with this.  Much tighter, though, and it gets unrealistic, you lose buy-in.

But if all projects in your domain go likewise, it means eventually you report up on cost and/or schedule increases 70% of the time. I don’t know what your taste is for that.  So I can also give you 50% numbers for you to choose action upon.

And what about the guys at the very top?  Is the 50% right for them?  These project projections go public and can affect investor evaluations.  This is the level that also deals with marketing and sales and setting timing and staffing for those guys.  Maybe they would prefer a reputation of always keeping their promises, often bettering them, rather than being delayed half of the time.  They may like 85% or 95% likelihood targets.  I know this gets into the range of “liar’s poker” – other projects or other companies who predict unachievable but faster, cheaper, better up front may get the funding and/or investor nods and then take their lumps 70% or 80% of the time later.  Again, this is a matter of taste.  Those are questions for the higher pay grades, but hopefully the decision makers are not upset to have facts at their disposal.

=======================================================================

Getting into the details?

“Well, it is sort of like this.  The number is really a prediction of the future, and while we predict the future based on real data, some of the data we rely upon has some level of innacuracy (or range) and on top of that, uncertainties crop up in every project.  To make my prediction as accurate as possible, I ignore nothing.  I have a way of using a series of numbers, called a DIST, to capture data from the past plus its range or uncertainty at the same time.  And these series, or DISTs, follow the rules of arithmetic: we can add them, subtract them multiply them, etc.  And where there are uncertainties, I model them with random numbers in a way that the random numbers conform to patterns – DIST’s, again.

Once I have the model, it’s sort of like a ladder to view the project, then I give it a good shake.  I run 10,000 or 100,000 trials letting the random numbers fall where they may (I could do more, but we tend to get convergence within these number of “shakes”).  That’s what generates the 30% and the 50% and the 85% and 90% probability numbers.

 

Notes:

Thanks to my brother, Dave, for the discussion we had on this on the way to the airport Oct 8, 2011.  In particular for your boss’ quote about planning for failure.

And thanks to Denny Westover for the thermopen.

PHB 160

Leave a Reply

Please use your real name instead of you company name or keyword spam.

*