“The stuff you engineers design is too good,” the guy from Marketing said, “I need shoddier! Can you make them less efficient and cheaper?”
I was working for a small company that made electric power conditioning equipment. Our devices protected against spikes, sags and brownouts in the AC power for mission-critical equipment like air traffic control radar and the computer rooms of Las Vegas casinos. Our engineering team was justifiably proud that we provided the best power protection on the market. “OK-Power, our toughest competitor, is eating our lunch,” Marketing Guy continued, “I can argue specs all day long, but when I’m out of breath, the customers say OK-Power’s box works good enough and costs 20% less than ours.”
“But OK-Power uses 30-year old technology and is only 72% efficient. Our new technology boosts efficiency to 93%. Sure, our equipment costs more, but the lower electric bills pay for that in less than a year!” the engineers answer, flabbergasted.
“I know that,” Marketing Guy said, “and I explain it and the customers even understand it. Then every single one of them tells me the equipment cost comes out of his budget, but somebody else pays the electric bill. They get their bonuses for cutting their own budget, not somebody else’s. So they buy the cheaper, less efficient one from OK-Power.”
My old employer is pushing up corporate daisies now, perhaps because of short-sighted customers, perhaps because Marketing pandered to that short-sightedness, or perhaps because stiff-necked engineers couldn’t bring themselves to design inferior circuits. But the problem lives on everywhere that purchasing decisions are made by people other than those who use the product.
Take, for example, (and here I finally start to get to the thermal part) the discussion I had with one of my previous bosses about buying a CFD package to help us with thermal analysis of electronics. CFD stands for Computational Fluid Dynamics, the theoretical basis for computer programs that numerically figure out fluid flow and heat transfer, given some geometry and material properties and boundary conditions. You probably know it better as the trademarked commercial codes FLOTHERM, Icepak, Coolit, Maya, and so on. After enduring many demos and then trying out the CFD packages on real-life electronics cooling problems for a few months, I was ready to make a case to my boss.
His eyes glazed over at my table comparing CFD temperature predictions against experimental results and at my chart about avoiding costly re-designs if we could predict thermal performance more accurately. But they lit up again when I showed him a CFD color temperature map of a circuit board. He said, “We’ve got to have this! But which one? Sounds like a perfect job for the Decision Matrix!”
His method of deciding things like this, including the best place to go to lunch each day, was to make a spreadsheet. In this case, the first column listed the CFD packages under consideration, and the top row listed their various features. Each feature would be weighted with a relative importance from 1 to 10, and then each package rated on a scale of I to 5 on how good its version of that feature was. The spreadsheet added up all these subjective ratings to give an objective score, and the CFD package with the highest score would win.
I listed the features that I, the primary user of the package, thought were important, like Ease of Use, Accuracy of Results, Solver Convergence, Solver Speed, and Quality of Tech Support. Then my boss insisted on adding his own features: Cost (I had no argument with that), Post Processing (the ability to make impressive color temperature maps), and the most heavily-weighted one of all: CAD (Computer Aided Design) Interface.
“But I don’t even use CAD!” I protested, “Why is a CAD Interface so important?” “Concurrent Design, son,” he said, even though he was only 6 months older than me, “That’s the buzzword that drives our industry this year. You want to be able to seamlessly exchange design data with our electrical and mechanical CAD systems.” “That sounds good,” I said, “But the geometry used by the CFD package is usually a very simple subset of the geometry in those CAD databases. So I’d end up deleting 90% of all the data I’d import. If I ‘exchange’ my CFD model back to CAD, then that engineer would have to spend all day putting all that deleted information back in.” The boss waved away that worry with, “I’ll bet they’ve got translators for that.”
“CAD designs don’t have the thermal data I need, like power dissipation or conductivity,” I said. “We’ll have them add that to the component libraries,” he answered. “Maybe,” I argued, “but the main reason I’ll hardly ever use a CAD interface is timing. Thermal analysis should be done very early in the design to be of the most use. By the time the other engineers have finished creating their CAD designs, they’re ready to start building hardware. That’s too late do run CFD, because it’s too hard for them to change the design if I do find a thermal problem. A lot of my input data now comes to me on napkins or in e-mail, not from CAD. James usually brings me his power numbers written on the palm of his hand.”
My boss shrugged. “Doesn’t matter. I’ve been tasked by the Process Processing Committee with organizing all our engineering tools for Concurrent Design. I can’t report I’m doing that if your CFD-thingy doesn’t have the best possible CAD Interface.” He had the power to sign the Purchase Order, so he got to assign the weights to the features in the spreadsheet. He was a busy manager, though, and left the actual construction of the decision-making spreadsheet to me. Later I gave him a printed copy to review. Even though the CFD package I preferred scored low on CAD Interface, its other features somehow added up to the highest overall score. The boss was satisfied that the spreadsheet had sorted out the correct choice, so he approved the Purchase Order.
Then he had me run the lunch spreadsheet. “How come it never picks Burrito Castle anymore?” he wondered as we headed for the door. I shrugged and said, “It also says it’s your turn to pay again.”
