Working Up

by Dwayne Phillips

How specific should a system specification be? How many resources should the spec consume? I propose a measure to help answer these questions – the Return on Specification Investment.

A few years ago, I was reviewing a specification for a system that someone in the office had written. I came to one part where the engineer was specifying a controller. All he wrote was:

a rugged laptop

Hmmm, not very specific for a specification. A contractor could go a lot of different ways with that vague statement, claim compliance, and not satisfy the user. We had to do something, or did we? This brings the question,

how specific should you be when writing a specification (spec for short)?

Consider the “rugged laptop.” Panasonic makes the flagship Toughbooks as listed here. There are a few other companies that make laptops that are “rugged.” That usually means that you can drop them and they still work. Some of these rugged machines can take a cup of coffee spilled on the keyboard; some can take a few hand fulls of dirt on them. Others can take more heat and cold than I can.

So what is important? Well, that depends on the system with which you want a rugged laptop, and that brings us back to the situation where we started. The engineer specifying the system didn’t know what to write. So, in my opinion, he was lazy and wrote “a rugged laptop.”

What could he have done? One thing is go to the Panasonic web site, as one example, print the marketing information for their Toughbook, and use those numbers as the specification. That is a little less lazy, but not quite creative. The short cut there is to specify

any computer meeting or exceeding the specifications of a Panasonic ToughBook model ABC

I guess that would work. A similar solution would be:

Use a Panasonic ToughBook model ABC as the system’s field controller

Okay, enough with the rugged laptop. The general concern is

How much resources should be spent on specifications?

The answer is

Enough to justify a return on investment, but no more.

Sort of vague, huh? Let’s discuss this for a few lines and see what we have.

A specification costs money. It costs time and effort to write by the specifier. The big cost is for the builder or the organization that delivers the system. They have to track the specifications all through the system build and have to test to the specifications. If the specifier lists all the numbers from Panasonic’s marketing, the builder will have to run all those tests to ensure that the laptop chosen meets them. If, however, the specifier writes, “Use Panasonic model ABC,” the builder only has to say, “yes, we provided that model.” See how much cheaper that method is? Of course, that is cheaper if the specifier was correct in stating which Panasonic model to buy. There are many possible problems that come with that type of specification.

Consider the case where the builder buys a Panasonic model for a few thousand dollars. If they buy the wrong model, they can toss it and buy another one. Only a few thousand dollars are wasted. How much money will the specifier and the builder spend if the specifier gives tiny details in the spec? If that is tens of thousands of dollars, that is a poor return on investment.

I guess that is the point, the return on investment or what I will call the Return on Specification Investment (ROSI for short – pronounced like rosey or ro-zee).

If the resources needed to specify an item and test the item to that specification are greater than the cost of the item, you have a poor ROSI.

If the resources needed to specify an item and test the item to that specification are less than the cost of the item, you have a good ROSI.

That seems to make sense. Using these guidelines requires calculating a few things before starting. The hardest part of the calculation would be to estimate the resources the builder will spend testing a specification. That cost is usually much higher than most people who haven’t built a system to a spec would think.

Anyways, give this a little thought. The concept is still swirling in my head. Maybe one did it will come to something.