The Operator's Blog

Usually at this point the operator is the first one to be sacrificed at the altar of misguided expectations because it must have been execution, and therefore the operator’s fault.  After all, the team that built the model, with a combined IQ in the stratosphere, in-depth business case experience from great schools, and modeling skills unmatched in the industry, could not have possibly made a mistake, and even if they did, they are nowhere to be found since they never run anything, just modeled it.

I can’t really share how many times and at how many clients I have seen this play out and how often we have been called to fix the “problems with execution” when the root cause was the GGF trap in a model closing around an operator’s neck.  I will however sanitize the names to protect the innocent, and alter some of the facts to share how it happened in a classic instance not too long ago.

First let me explain the GGF trap, which is simple to avoid if you are an operator but easy to fall into if you are not.  For simplicity let’s imagine you know nothing about the operating issues of lawns and grass in general.  Your lawnmower experience is limited to looking at their pictures in the Sunday circulars from your local hardware store and wondering if you need special driver training to run one of them.  You have a house with a nice lawn, and your spouse wants you to spend some money for privacy bushes in front of the windows in the front yard.  Being a good operator, you decide to find the least expensive solution to the problem.  You can always buy some bushes and have a landscaper plant them, but why not use existing yard resources better.  After all, since doing more with less is a standard mantra at your company, why couldn’t it work at home?  So, how about using your grass more efficiently?   It’s there; you already have someone taking care of it; and if it can grow high enough to hide the bottoms of the windows there will be no need for additional plants. But your spouse will not be easily sold on the idea and you decide to build a business case for it. The first thing you need is a good model.

So you ask your lawn company to let the grass in a small part of your yard grow for a month without cutting it.  Every few days you measure a good statistical sample of the blades and discover that it grows at a rate of ¾ of an inch per week.  You then enter the data in an excel spreadsheet, create a forecasting model for the next twelve months, and low and behold, your model indicates that by next year your grass will be over five and a half feet tall.  If you are a really determined modeler, you actually include temperature, rainfall, and a dozen other parameters and your complex multi-dimensional model shows that the grass will be 5.63 feet tall with a plus or minus 8% degree of certainty.  Using the output of this very sophisticated model, you attempt to convince your better half that you do not need to plant privacy bushes in front of the windows of your home.  Just let a patch of grass grow in front of each window, and within a year all you have to do is trim the top of the grass blades so they don’t grow taller than they need to be.  And, of course, you can use the forecasted savings to buy a new remote control helicopter.

Sounds logical, doesn’t it?  After all, the model is extremely sophisticated and all the calculations have been verified more than once.  Needless to say, logical or not, most people can tell you your plan will not work for many reasons.  Any lawn care company operator regardless of their ability to understand mathematical models can easily spot the errors in the output of your model.

Regardless of the mathematical soundness of your extrapolation formulas, the outcome is only possible in the model.  Regardless of how well the trend look in the model, in real life, grass roots are not deep enough to support a six foot long blade; And even if they did, the blades are not strong enough to continue pointing straight up at that height; And even if they did, the base of the each blade of grass would have to be a couple of feet wide to reach that height and there is no grass in the world with that characteristic, etc., etc., etc.  The concept and silly GGF analogy is so simple, no one with sound mind, let alone an analyst with an MBA from a prestigious university, would fall in it when constructing a model, right?  Well, that would be a wrong assumption.  Sad but a true story, here is a classic example and a case Dilbert would love.

A couple of years ago, a division of a major corporation was experiencing low single digit margins when everyone in the corporate office thought they should be in the double digits.  So the manager from the corporate finance organization responsible for financial oversight of the division -  a staff type and not an operator by any stretch of the imagination, who was actually slightly intimidated by operators and generally avoided them - decided to build a model to analyze the business unit’s performance for the upcoming budget cycle.  So he engages a bright chap, a recent MBA graduate, to work with him to develop a model and a budget plan on how the financial performance of the division can be improved.  Without an operator in sight, the young MBA, from a very prestigious school and well versed in modeling and simulations, builds a complex financial model that shows that the division can drastically change its profitability in less than twenty four months without any major changes or investments through a slight reduction of cost and a small increase in revenues.  Without a review by the business unit, the plan is presented to the corporate budget committee where, based on the output of the model and a great slide presentation, it becomes the division’s performance objectives for the coming year and is included in the corporate plan presented to Wall Street. 

The operating executive at the division initially protests that he did not sign off on the plan, but having limited ability to argue with the very sophisticated model that created his budget, he is told it’s just a matter of working a little harder and a little smarter so he accepts the model’s output as the baseline budget.  After a couple of months, things are off to a good start according to the model.  Or so it seems…  Six months into the year, things are not moving according to plan, and twelve months later, the division is still producing lower single digit margins and experiencing a flat year in all aspects.  During the last two quarterly analyst calls, the CEO is asked to explain why the company is not meeting its own quarterly earnings guidance, and the division president, a good operator in many respects, starts fearing for his job.  After fourteen months, the CEO fires him and engages us to help him “understand where execution of the plan went wrong” and “develop an operating strategy to bring the division back on plan.”

No problem.  That’s what we do for a living. 

So we start digging in and after a couple of weeks of data gathering inside the company, we finally have all the material we need to start deconstructing the performance objectives and profitability scenarios for the division.  Even though we find a couple of areas we can improve operating and technology performance, and some opportunities for outsourcing, without a major overhaul and significant capital investment, the contribution to the bottom line will be marginal at best.  It seems the operator in charge run a tight ship without much inefficiency with what he had to begin with.  We are all scratching our heads trying to understand why the expectations are so far removed from the reality of the situation when one of our partners with a forensic accounting background suggests we ask for the raw data behind the budget.  We do, and within a couple of hours we get a copy of the truly outstanding presentation deck that solidified the budget, the output of the model used to support the presentation, and the model itself.  After digging into the model for a couple of days, we all agree the model’s calculations and formulas are correct and the assumptions are generally reasonable from an outsider’s perspective.  A few things stand out however after we look at the inner workings of the model.  The major component driving the Costs Of Goods Sold line item down is a “small” reduction in raw material cost from 14% to 11% consistent with a trend from the last three years. This trend is so consistent, the model assumption is not even highlighted in the footnotes.  Similarly, one component that drives the revenue line up is a “slight increase” in market share from 5% to 7% over two years, not an unreasonable extrapolation for market share growth based on the actual data for the last three years, and also barely noted in the footnotes.  Other than that, nothing really looks out of line.  So how could execution of this plan go so wrong?  Why isn’t reality conforming to the model?

After some discussion, in a case of “a brilliant insight into the obvious”, we decide to check with some operators.  So we consolidate all the assumptions and trend calculations from the model and build a matrix that lists all of them.  We then obscure the client information and set time to discuss this assumption set with a couple of operators we know in the client’s industry. 

The meetings are an eye opener.  In two separate meetings each of the two operating executives in the industry, takes less than fifteen minutes of looking over our matrix to spot the problem with our “theoretical company turnaround.”  When the operators looked at the model they both zoomed into the same two things in a list of over fifty major and minor assumptions, raw material cost, and market share increase.  Even though mathematically the raw material cost forecast was correct and the trend based on historical data realistic, it turns out the reduction of raw material cost trend had run its course since an offshore player entered the market five years ago and it stabilized in the last year before the model was built.  The modeler, having no contact with the reality of the business unit, or any operating experience for that matter, did not know that.  They also pointed out that if the trend continued to the price points used in the model, the raw material suppliers could simply switch to selling to another industry that used a cheaper substitute but would buy their product if the price was competitive.  In a classic GGF case, the raw material forecast was not connected to reality but only to a mathematical formula in the model.  There was no base for extending the trend beyond the last data point and that “small” reduction was a non-starter at the get-go.  To quote one of the operators smartass comments: “why don’t you just assume they will give the raw material away for free and make your company turnaround really great.” 

The next point both operators zoomed into was the increase in market share.  They pointed out the entire industry segment only grew at 5% a year for the last few years and based on their experience it was going to be flat for the next couple of years.  New competitive products entering the market addressing the needs of their clients would require most players to retool or consolidate.  For our  "theoretical" company to grow its market share by almost 50% it would have to win all the new business, and take away significant market share from the dominant player in the industry, which would be a “pricing blood bath” as one of them put it.  Even though possible, without a significant investment in sales and marketing, and even a potential acquisition of a couple of smaller players, extending the trend would be almost impossible to accomplish over the next couple of years.  Another mathematically correct model created by an analyst not connected to operating realities bites the dust and takes a good operator with it.

During our report to the CEO, we pointed out that in the end, those two trend extension assumptions made the model an exercise in futility and wishful thinking and the plan was doomed to fail from the start.  Nothing anyone did in operations, short of re-tooling for the new product lines, could create the output the model predicted. It was not an easy meeting, but in the end, we worked with the client to prepare and sell the division to their major competitor and helped the fired division president find a job with that same competitor.  We also added another GGF story to our "modeling gone bad" category.