3.3 COLLECTING THE DATA

< Day Day Up >

"Yes!" shouts the MD or SEO, thus giving you the management commitment you may believe you need, "we want information." What do you do next? Unfortunately many groups and organizations who have started an initiative to provide management information have gone out "data hunting." The average dedicated data hunter is nothing if not thorough. He or she has two guiding principles: if it does not move then measure it; and if it does move, pin it down then measure it!

The average data hunter is also something of a fanatic who wants to involve everyone else in their abiding interest. The data hunter will lay traps with forms and procedures seeking to include everyone in their passion. Only when everyone is spending great chunks of time and effort supplying information to the data hunter will they be truly happy.

And what is the result of all this happy hunting? Well, the first thing that people realize is that it tends to be costly. You see, one thing about the average data hunter, they are impatient. Not only do they want lots of information, they want it now. This means that they often get teams of people to sift through old data or even possible sources of old data. Joy to a data hunter is a 50,000-line COBOL listing (OK, I am showing my age but there is also a confession coming). There are so many ways the data can be cut! Sorting through any form of old documentation, except code which is at least available electronically so that the hunter can get lots of data quickly, takes time and effort. Lots of it. This costs money.

Now having got the data, what do you do with it? Obviously you analyze it! Do you know how much raw data you can get from a 50,000-line COBOL listing? Let me put it this way, do you have a spare room? And all of this data needs to be analyzed when, despite the advances in electronic data analysis support, working with more than about ten variables (ten being the number of subjects the preconscious brain can handle), is pushing your luck. So you need huge amounts of effort to carry out the analysis of all this data which increases the cost again. In fact, some organizations have rooms full of data that they have neither the time nor, to be honest, the inclination to analyze. Such organizations fell into the hands of data hunters.

Even if you do manage the data analysis the results may cause you even more heartache. You see, if you have lots of data and lots of variables, something is almost bound to relate to something else just by chance. And by the way, do not fall into the trap of believing that the laws of probability apply in anything but their fullest sense. This means, to paraphrase Murphy's law, "if event E has a probability P of occurring on a specific day this means that event E will occur when it can do the most damage or on the first day after counting begins, whichever has the greater disaster potential." In other words, the probability of two data items relating to one another through chance may be mathematically small but this means you will probably go into your senior managers meeting and state that programmers who keep hamsters make good testers, because that's what the data shows! Chance got you again.

The final problem of the data hunting approach I can vouch for from painful experience. As someone who did once dabble with data hunting and someone who also believes that data, somehow, must be of some use if only I had the time to... (I told you there was a confession coming) As I was saying, the final problem is that you will always miss the most important variable. Murphy's law strikes again!

More seriously, organizations who do operate under a data hunting regimen tend to have very disgruntled employees because all they see is bureaucracy and yet more bureaucracy.

But if data hunting is not the answer, then what is? One of the key points that I hope this book can get across is that measurements should only be made to satisfy specific requirements for information which in turn should be linked to business objectives. Management information is no exception to this rule: you collect data because your modeling process indicates that the collection of that data will satisfy an information requirement.

< Day Day Up >