Among the strange quirks of human psychology is this: While most people know that argument by analogy is invalid, we’re universally vulnerable to how the analogies we choose influence our thinking. For example, those describing corporations and their behavior have, among other metaphors, likened them to jungles, predators, economies, giant human beings, prosthetics, robots, factories, plants, farms, computers, and amoebas.

If you envision your employer as a prosthetic — as a mechanical extension of the will of its owner — you’ll have very different expectations of how it will and should behave than you will if you think of it as a self-directed machine. Among the differences in perspective is how you’ll think about optimization — the subject we’ve been pursuing (beating to death?) the past several weeks. If you consider the enterprise to be a machine, you’ll think about optimization and wonder how many decimal places you can take it to. If you figure it’s an economic system you will, Adam Smith fashion, establish a network of charge-backs and expect the invisible hand to optimize everything without anyone having to worry about it very much.

What’s the best way for you to think about the part of the enterprise you lead? Personal opinion: Push all metaphors out of your head. Relying too much on metaphors is like wearing blinders.

Oops.

As pointed out in last week’s column, few situations allow you to optimize for just a single variable. Imagine you’re running a small company that manufactures a single product. Imagine, for the sake of argument, that you know, with a certainty, exactly how variations in each of your product’s attributes — price, durability, performance and so on — will affect its attractiveness to the universe of potential customers. And finally, to prove it’s a fictional scenario, imagine you’ve created a system of simultaneous equations that perfectly accounts for all aspects of your business, including the cost of manufacturing and distributing your product and how the cost changes with each variation in the product’s attributes.

By manipulating your system of simultaneous equations you can posit any number of product variations, predict exactly how much each will cost to manufacture, and what selling price will maximize your profits for that product variation. With this excellent command of your enterprise you’ll have no trouble formulating the exact situation that maximizes your profits (I’ll leave the math as an exercise for the reader).

Except ,,, even if you do end up with a single solution that maximizes profits, that solution might not do much to position your company for future success. And since the future is intrinsically unpredictable, you’re still making guesses, even in this idealized scenario.

So much for fiction. No company has this level of understanding. No company has anything remotely resembling this level of understanding, and there’s no point in pretending any company does.

How does this relate to your situation? You don’t run a whole company — you run a department or division in the company. The company doesn’t have a system of simultaneous equations that shows exactly how changes to different departments affect overall profitability. All it has is a budget, and nobody even understands that.

So to take the first step toward optimizing your part of the company, forget about the whole notion of optimization. It’s a chimera.

But don’t give up entirely. Just because you don’t have a system of simultaneous equations to optimize doesn’t mean you have to sit on your hands doing nothing.

Here’s a sane starting point, which I learned from my research advisor back in my electric-fish days: Understand your subject at three levels of description. Learn your personal level of responsibility in depth. Develop a working knowledge of the levels above and below. I studied behavior, which meant I had to understand behavioral theory in depth while developing a working knowledge of neurobiology and sociobiology. Three levels of description.

In a business situation (I’m guessing you aren’t all that interested in the behavior of electric fish) it means you should thoroughly understand what you’re personally responsible for while developing a working knowledge of the responsibilities of those reporting to you (level below) and how your organization contributes to the ones it directly connects to (level above). This more limited understanding of the universe and your place in it won’t let you achieve global optimization. Understanding the level above will, however, help you develop plans to improve your own organization in ways that are likely to contribute positively to the company as a whole.

Understanding the level below means you won’t make those who have to make it happen burst into hysterical laughter.

I’ll wrap up this subject next week … I hope!

Harriet Wasserman writes:

We just came back from Stockholm where we visited the Vasa ship exhibit. In 1600 something, the king modified plans for his giant ship, to make it bigger and with more guns. His shipwrights (comparable to our engineers) told him he couldn’t add enough ballast and would have problems. He insisted, and they constructed the giant ship, covered with gorgeous carvings and carrying about a zillion guns. It got out into Stockholm harbor, tipped because of too little ballast, all the guns rolled to one side, and it sank. It was retrieved in 1963. The design story is all too familiar.

Core Principle #1 of the KJR Manifesto states that in order to optimize the whole you must sub-optimize the parts. It isn’t original (although I’ve been unable to track down its author). And it raised an objection in the correspondence I received in response to last week’s column that introduced the KJR Manifesto‘s core principles.

The objection was entirely reasonable. Paraphrasing, it’s that optimization never makes sense except in the concept of the larger whole, but that a naive manager, reading this catch-phrase, could use it as an excuse for poor operating discipline.

It’s a fair criticism. It can be applied to nearly any succinct statement of principle. That’s why each core principle is accompanied by explanatory text. It also raises a larger subject that’s well worth exploring: That in organizational design, when it comes to parts and wholes, “whole” is purely a matter of choice.

Consider General Electric. It’s pretty much run as a holding company. The point of optimization for GE isn’t the whole corporation — it’s the individual business units. They aren’t evaluated according to their functional contribution to the larger entity. Each must be one, two or out all by itself. That stands in contrast to Costco or Nucor, whose parts optimize Costco and Nucor, respectively.

Or take the pre-SBC-acquisition AT&T. Its point of optimization was the shareholder, and optimization was defined in terms of the price of a share of stock. It certainly wasn’t the company itself or any of its business units. The word that comes to mind when describing those is “pathetic.”

CEOs are responsible for clearly defining what gets optimized — the enterprise, individual business units, product lines, or what-have-you. Call that the Optimization Unit. Whoever is responsible for each Optimization Unit defines optimization priorities (perhaps cost, then speed, then quality), and aligns systems and structures with them.

Those who manage the divisions that make up an Optimization Unit deal with more complexity than the Optimization Unit manager, because any change … any improvement they might want to make in their own division … could have an impact on one or more of the other divisions in the Optimization Unit, as was the case with the Vasa and its improved armaments. And unlike physical engineering, organizational engineering is an imprecise discipline under the best of circumstances.

This analysis leads to the potential for conflicted IT. Imagine an enterprise that adopts a holding company structure, like GE. Now imagine a central IT organization that tries to support the enterprise. One IT organization that has to support multiple, autonomous Optimization Units won’t be able to sub-optimize its own operations in a consistent rational way.

Which is why, I think, CIOs in this position usually choose to either establish IT as an internal outsourcer, or to structure IT as an “organizational mirror” with separate departments, each aligned to a business unit.

Most managers, up to and including the Board of Directors (and, it appears, some Swedish kings) don’t understand this principle at all, figuring global optimization is nothing more than the sum of local optimizations. So they set up systems that assess the performance of sales, procurement, manufacturing, and distribution in isolation. I know this from personal experience, having been responsible for the manufacturing budget once upon a time. We were to develop our budget in isolation from sales forecasts. Predictably, the CEO pushed Sales to increase their forecast, and simultaneously pushed Manufacturing to reduce its budget. You can guess what happened next.

With all of the above in play, I don’t mind “sub-optimize the parts to optimize the whole” at all. It is catchy. More important, the phrasing requires executives and managers to figure out what constitutes the “whole” and their relationship to it, and to recognize that they can’t make choices in isolation.

A longer essay doesn’t do the job, because they’ll never read it … unless the catchy phrase first grabs their attention.