It sounds easy and well understood as best practice, but allowing the data to drive decisions is often times difficult. Personal agendas, beliefs or past practices can infiltrate the conversation and provide alternative thinking. Good sound data is the best decision making tool. Whether related to a personnel decision, business process or outsourcing decision, accurate, valid data is the key.What is good data? Many managers believe it is that which supports their gut feel and opinion, however, good data is objectively collected, effectively analyzed and acted upon. Objectively collected data is that which is not contaminated by factors other than pure randomness. If there are external, assignable causes present within the data you could be directed to false conclusions. For example, let’s assume we have a transactional process where we are monitoring errors in the number of transactions completed. We started tracking the data in January 2006 and the data is as follows:
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Errors 20 18 21 22 25 25 31 28 35 38 38 37
In reviewing the above data one can easily see a moderate trend upward. I realize this is basic and I am not questioning your statistical intelligence. The intent of using data is not the utilization of a Monte Carlo study, design of experiment, multiple regression or response surface methodology. I am referring to simple, basic statistical tools that can go a long way in driving correct, congruent decisions. Referring back to the data above, anybody can tell you have an unfavorable upward trend that requires some immediate attention, right? Not so fast my friend. Before you claim yourself winner of the “Jump to Conclusions” game, consider this, in May, torn documents were added to the list of errors, prior to that they were not counted. Torn documents average 5 per month. In September, a customer complaint added another error code to the transaction process. Timeliness of the transaction was included as an error code in September. Prior to that it was not counted as an error. Timeliness or the transaction errors average about 10 per month.
With the new error codes included the data looks like this:
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Errors 35 33 36 37 35 35 41 38 35 38 38 37
That looks a little different. You may conclude that the increase in transactional errors was caused by the change in the error coding. There was no change in the transactional errors, there was a change in the measurement method. There is a difference. This sounds basic, but often times measurement methods and data integrity are not understood and false indications are recognized.
Once we have ensured the data is objectively collected, now you must analyze it. Take a look at the data below, which represents rpm’s of a motor:
4600 3400
3200 4400
3400 3600
4600 4400
4400 4200
3400 3400
Looking at the data, not much initially jumps out. however lets put it into a simple histogram.
RPM Occurences
3200 1
3400 4
3600 2
3800 0
4000 0
4200 2
4400 4
4600 1
Hmm, that looks a little different. Somewhat appears as if there could be two distributions occurring. A bimodal distribution could be the result of many causes, either way it requires further investigation. My point is not to insult your statistical intellegence, my objective is to reinforce the power that lies in simple statistical tools. Basic data analysis can go a long way in discerning fact from fiction when it comes to what the data is telling you.
There is a place for more advanced statistical tools and they are invaluable, however these basic tools have a distinct advantage when they can be applied. For one, they are simple to use. With some initial training, the basic statistical tools can be utilized with good effectiveness. Number two and most important they are easy to understand. Top management can understand and grasp average, run charts, histograms, pareato charts and the other basic statistical tools. Try getting your point across when you’re blabbing on about a two way analysis of variance or an F-test.
When the job you need to complete, entails putting a nail in a 2×4, you don’t need a hydraulic, computerized numerical controlled force application device, a hammer will work fine.
