Posts Tagged ‘making measurements’

Finally, validated data sets to “crunch”!

June 9, 2009

I am excited by the announcement of WolframAlpha. Wolfram claims that this new dataset database will be validated, and therefore more valuable to the user. My hope is that the validation helps me “make a measurement, and not just a “reading”. For WolframAlpha to be truly validated, I think that four elements are necessary:

  1. the source of the data must be known and published;
  2. the actual data must be verified correct and accurate;
  3. other similar datasets must be identified as well, and it must be clear why they are similar, and also how they are different; it will be essential to know the complete and explicit schema; and
  4. some data may be missing from the dataset; for example, certain US Census datasets may omit responses from rural areas and towns smaller than a certain threshold population – in other words, there may be “context” to consider, and this is the real challenge for Wolfram – to explain the context of the dataset completely enough.

If Wolfram is “religious” about this kind of validation, their WolframAlpha data services will be tremendously valuable to researchers and analysts – these users can be more confident that they are making “measurements”!

What the computer says “must be true”

May 31, 2009

To quote a recent Microsoft customer service rep, “Yes sir, that’s what my computer says, it must be true!” That’s dribble. Two errors were made by this customer rep: 1) my registration records were checked, but initially not the correct records; and 2) their records were incomplete – even when the correct records were found. The service rep had 100% confidence in inaccurate and incomplete records. This problem has been a common one for a very long time for many companies, and it persists to my great irritation – everywhere…

This post is another commentary on the value of context. My mentor many years ago drilled this meme into my head: “Just because you took a reading doesn’t mean you made a measurement.” In other words, don’t blindly believe what you see on the test equipment display – or on your computer display, for that matter. That engineer at Collins Radio Company was Bill Thompson who died of cancer not too long ago. I will miss him until my own death – he was a truly great engineer.

Bill’s message to me was to be sure I was measuring the correct signal with a properly configured test setup measuring the relevant characteristics that are necessary to answer the pertinent questions so that I draw a valid conclusion. Until I was certain of all of those considerations surrounding my test equipment reading, I had not made a measurement. I had to be careful to “think sufficiently” about my objectives and know enough about the context of what I was trying to do.

My conversation with Microsoft during the week failed to accomplish what I set out to do. Ultimately, their dataset was incomplete. Microsoft had acquired a software company, and that company’s old registration records were not completely incorporated into Microsoft’s own CRM systems. Microsoft provided no context to the customer support staff that registration data for my product might be incomplete. “Our records simply do not show that you are a registered user (you’ll have to convince us…).” They “took a reading, but they didn’t make a measurement”, and it is my problem to convince Microsoft to the contrary.

Well, this isn’t over, yet. I found my e-mail copy of the original purchase and the company’s confirmation of receipt of my registration information on my e-mail archive DVD! But is it worth two more hours of my time on the phone with those buggers at Microsoft?