The Value of Information
05/28/2020 Category: Network Design
One of my favorite stories (stop me if you’ve heard this one before) concerns an aged power station. One day the station failed, leaving the small town it served completely without electricity. The town managers called in several engineers, none of whom could resolve the problem – the station’s equipment was older than anything they had worked with before.
Eventually someone remembered the name of the long-retired engineer who had maintained the plant almost since its construction; a quick phone call was made asking for his help.
When the old man arrived, he spent 15 minutes wandering around the station, inspecting, listening, and touching. Then he took a hammer from a toolbox, approached one piece of equipment, and tapped it gently three times.
The entire station immediately came back to life. The engineer tossed the hammer in the toolbox and left.
A week later the town managers received a bill from the old man for $10,000.03. The bill was itemized as 3 cents for three hammer taps, and 10,000 dollars for knowing where to tap.
Information, as that retired engineer knew, has value.
I’d like to conclude my short series on measurement as a means of reducing uncertainty by discussing, briefly, the valuation of information. The context in which I’ve been writing these articles is determining how, and what, to measure to make a better business case for networking projects. Networking projects are expensive: Reducing uncertainty (not, as I’ve been emphasizing, reaching 100% certainty but simply becoming less unsure than before the measurement) has distinct value in helping to make better decisions about how the project can be executed with less cost and less risk.
The value of the information you acquire from the measurements depends on how much you already know. If your level of uncertainty is high, almost any information that reduces the uncertainty is valuable.
But there also is a point at which more information becomes less valuable. In a previous post I gave an example of a carpenter who can measure a cut to a 32nd of an inch using simple tools. He could make his cut much more accurate with highly calibrated scientific instrumentation, but the increased accuracy is not worth the cost of the equipment. A 32nd of an inch is close enough for most carpentry work.
In a network study, perhaps you are making incremental samples of traffic flows to understand where you need to add bandwidth. The first 50 samples are highly valuable in helping you get closer to the answer. The next 50 samples are less valuable; the 50 after that even less so. At some point further information is not worth the time and effort of taking the samples.
Determining the value of subsequent measurements usually depends on the results of the initial measurements. Unexpected results might sustain the value of further measurement, or they might cause the value of further measurements to drop almost to 0.
The results of initial measurements might also shift what we measure and how we measure it. Things we thought were important might turn out to be unimportant, and things that we thought unimportant – or immeasurable – might turn out to have a heavy economic impact on our project.
It’s also important to understand the cost of measurement and the value of the resulting information within the context of the costs and risks of the project. If your estimated project cost is $10 million, a $100,000 study to clarify the “intangibles” of the project is only 1% of the expected project cost. If the study results in a 1% reduction in cost, it has paid for itself. If it results in a 5% savings, it has become a very good investment.
The bottom line is this: Most things we think of as intangible in networking can in fact be measured; if it truly cannot be measured, it has no real value.