Tuesday, September 26, 2006

Beware the Dreaded “Model”


The model you have is wrong because sales were off by a factor of 1000. The associate who input numbers on the revenue line was thinking millions when the model was otherwise scaled to thousands…

We used to do this without computers.

At least, without personal computers, that is. All this stuff—company presentations, research reports, even earnings models—used to be done without the aid of Microsoft PowerPoint or Microsoft Word or even Excel.

Slides were created by graphic artists with sensitive natures who took their creations very seriously.

Word processing was done by secretaries using giant word processors in—consult your Wikipedia on this one, kids—what was, for its time, the wonderfully hi-tech “Wang Room.”

And while most earnings “models” consisted of a fairly basic income-statement projection with numbers obtained by hand on an HP calculator, any analyst with a big, complex “model”—for a North Sea oil and gas producer, for example, with lots of tax and royalty considerations; or a pharmaceuticals company with many new drugs in the pipeline—could take his work to the fellows in the data center.

There, the numbers would be input and later spit out on giant runs of folding paper with perforations along the sides. The analyst would stare at the numbers and make changes if needed and send them back until he got his “model,” which would then be, literally, cut-and-pasted by the graphics people for inclusion in the back of the report.

But the IBM PC and a program called Lotus 1-2-3 changed all that.

With Lotus on your desktop, you could—like a photographer with a digital camera and a one-gig flash card—try as many variations on the model as you wanted. Any oil price, any tax rate, any royalty calculation, any growth rate at all could go in…and out would come a beautiful five-year forecast down to the nearest penny-per-share.

Not that it was any more accurate than the hand-made version. Garbage in, garbage out, as they say.

After all, Enron had a model and eToys had a model and so did WorldCom. In fact, just about every fraud or flame-out ever recommended by one of Wall Street’s Finest has had, at the back of the report, “The Model.”

Which is why I was not entirely surprised to receive the response (quoted at the top) to a question regarding “The Model” I’d been puzzling over. It was for a drug development company we’ve been working on, and I couldn’t understand why the analyst’s numbers showed a massively increasing rate of loss in future years, even assuming the drug under development received FDA approval.

Turns out “The Model” was wrong, “by a factor of 1000.”

There you have it: garbage in, garbage out.



Jeff Matthews
I Am Not Making This Up


© 2006 Jeff Matthews

The content contained in this blog represents the opinions of Mr. Matthews. Mr. Matthews also acts as an advisor and clients advised by Mr. Matthews may hold either long or short positions in securities of various companies discussed in the blog based upon Mr. Matthews' recommendations. This commentary in no way constitutes a solicitation of business or investment advice. It is intended solely for the entertainment of the reader, and the author.

2 comments:

Rob Dawg said...

Lotus 1-2-3?

Oh, you mean that new fangled program that wants to replace my trusty VisiCalc?

Aaron Koral said...

Jeff:

Equity models, from my understanding, are based on a set of assumptions. Those assumptions can be either be macroeconomic (interest rates, GDP growth) and/or microeconomic (revenue growth, earnings estimates).

As those estimates increase in both number and complexity, the greater the likelihood that an error in one of the assumptions will result in the "garbage out", as alluded to by the response above.

When developing models, one should keep a razor handy, not by Gilette, but by Occam, to cut the complexity, and remember the late, great John Keynes quote that "it's better to be vaguely right than exactly wrong."