A striking lesson in calibrating needs with funds.
It began as it often does: A product emerges from the chrysalis of laboratory and prototype conditions, and is starting to gain market traction. Time to ramp up to production thinking.
The customer had a benchtop functional test setup. Perfectly serviceable when you need to test five or 10 boards. Very small board, roughly 1.00 x 0.200". Excite a gas discharge tube. Apply 0 to 300V in a one second pulse. Monitor the output pulse when the tube fires, as well as input capacitor discharges from a spark gap trigger. Make sure everything is connected and works as advertised. Four parts: a capacitor, two resistors and the discharge tube. Red light/green light output: either it works or it doesn’t. Keep ’em if they do; trash ’em if they don’t. No diagnostics. About as simple as they come.
The product is selling. Now comes volume. Time for grownup manufacturing. The little board is accommodatingly placed in an array of 100 boards. The customer wants to build 200 arrays at a time. Some expanded thinking is required to scale up testing to adapt to the new efficiencies mass production dictates. This is going to require serious test engineering and a well-defined statement of work.
We get the call and the SOW. Twelve pages. It includes an entire opening page of ANSI, IPC, Mil-Spec and customer-specific internal governing documents referenced. Specific equipment required, down to the make and model of oscilloscope and power supply. Very detailed environmental test conditions, pass/fail criteria, required data and datalogging. Even a corrective action procedure, in the event a wayward action needs correcting. Very little thinking required. Just follow, and do it, no questions asked, and you will be on the right path.
The good thing about such a painstakingly detailed statement of work is that it leaves little room for introspection or misinterpretation. You just quote according to the rules and deliver a price.
Which of course, upon delivery, sends the customer into cardiac arrest. Seems they had caviar aspirations on a Top Ramen budget.
We dutifully followed directions and, after several weeks’ work, proudly delivered a quote for a full-service functional test setup, LabView-programmed, for $30,000. The customer, upon further review, decided he wanted to spend $5000. Tops.
Confronted with cost, minds concentrate.
Maybe, the customer concluded, those tolerances in the SOW were too tight and could use some loosening. And a number of the power-on requirements weren’t really necessary. Certainly the 300V programmable power supply they specified wasn’t really needed.
Upon even further review, he concluded maybe his old, flimsy, one-up 300V benchtop kluge of a setup wasn’t so bad after all. The one he was ready to abandon one month prior, when we were first invited in. Not so fast. It’s looking pretty good now. Maybe the production ramp won’t be as steep as was initially anticipated, and can be managed using existing (i.e., paid for) resources.
It is always instructive when an engineer confronts the fact that new things cost money, and usually far more than was budgeted.
Wouldn’t it be interesting if GDP figures could be parsed by economists into subgroups like Quoting Exercises Wasted? I would imagine that would be a rather large figure when spread over all industries. In our own limited case above, several weeks’ effort were devoted to designing and quoting to the customer’s specifications. Now those specifications are no longer binding because nobody gave any consideration to the cost.
Another time, another place. Same symptoms, similar outcome. Yet another consequence of failure to think through the implications of requirements. Once again a very elaborate, 10-page SOW. This customer wants to combine basic analog/digital in-circuit testing with load current testing. In the tester. Together. He insists. It must be a complete solution, in one location. ICT combined with a functional load test. Load power per the SOW is specified as 100A @ 12.5V. That’s nominal value. On an Agilent 3070. Before specification and system tolerances. Factor the latter and you are rapidly approaching 1500W (Ohm’s Law). Not a happy situation for a 3070. External power supplies are needed to deliver the required current, and additional probes are necessary for operational safety. Not what the 3070 was intended to do. So we customize.
Assume for a moment we do this. It gets unhappier in this scenario if the board fails. Actually, catastrophic is the correct adjective. I mean flames. Under load testing as described above, a failed board declares itself in what could be charitably described as a destructive situation. Not only for the board either; fixture and test system run the risk of severe damage. We’re talking Chernobyl conditions. In one fiery instant the OEM would be the proud owner of a somewhat charred, seriously depreciated 3070. Probably not in the capital equipment budget, but there you have it.
Side note: I personally love detonations. Gomez Addams was a childhood hero. They really, definitively, unequivocally (in a Fourth of July kind of way) tell you something is wrong with the product. Yet combustible testing of this kind simply does not lend itself cost-effectively to mass production.
Pity. But I digress.
We painstakingly explain the realities and implications of the situation to the customer. Confronted with these bleak, and literally explosive, facts of life, and inconveniently accurate mathematics and physics, the customer relents and allows the functional fixture to be made to stand alone from ICT. Except now he isn’t sure what he wants, so we need to help him with his options, from simplest (benchtop manual plug-in and power-up) to the most sophisticated (fully automated, with push-button operation and datalogging). Four different options, from the most manual to the most automated, with a cost range assigned for each.
Meeting time. An opportunity to flesh this baby out and determine whether this is real or imagined. Explain in considerable detail to the customer in person the full range of options, so he can explain them to the Boss, who will then give direction. We describe the range of pricing, from about $5,000 for the benchtop solution to $70,000 for sky’s-the-limit automation. Boss is absent when this presentation takes place. However, his reputation for wanting everything and not wanting to pay anything precedes him. Our job is to throw cold water on his fantasies, which we do well (we think). We speak slowly and deliberately. Intuitively we believe this project is going nowhere, but to a certain degree you need to go through the motions and bring the customer to the same preordained conclusion.
There is, of course, always the risk that the customer would take the results of our engineering, our carefully explained methodology, and share it with a competitor to obtain the same result at a much cheaper price. Imagine that. They wouldn’t do that, would they?
Apparently not this time. One week later comes the email confirming our intuition: Functional test project has been shelved for the time being. Run quantities presently do not justify the expense.
Now they tell us.
The next customer who calls with a functional test requirement gets the privilege of being the beneficiary of the two recent negative experiences above. We look at their RFQ and SOW, and tell them almost immediately that to meet the requirements of this SOW, as written, they must be prepared to spend $XX (a large five-figure sum). Customer thanks me, says she’ll get back to me. She doesn’t of course. I have to contact her about a week later. “Oh yeah, thanks for emailing. Our customer decided to live with their existing test setup. They don’t have a budget anywhere near the one you described in your quote.” Shocked, shocked. They almost never do.
Thank you very much.
Scaring them works. Less engineering work for us that way too.
At least it frees us to move on to real work. You know, the kind that produces revenue.
Too bad we can’t bill them for the time wasted. At least in the last instance it was only one email.
Never forget that GDP is understated.
firstname.lastname@example.org. His column runs bimonthly.is president of Datest Corp., (datest.com);