What Changed

A fundamental question in data analysis.

Sales are down this quarter over last year. What changed?

Student test scores are down this year over those from five years ago. What changed?

I feel terrible today, but I didn't last week. What changed?

I'm making more money this year than I've ever made before. What changed?

The time targets on this project are slipping more than our previous projects. What changed?

And so on.

Now, the composition of the change may be its own study, but the key to understanding why you are where you are begins with the "what changed" question.

Sales are down. Hmmm. We changed advertising agencies just before the down-trend.

I feel terrible. Hmmm. My mother-in-law arrived three days before that.

Student test scores are down. Hmmm. We started teaching "self esteem" instead of fundamentals just before that.

I'm making more money. Hmmm. I stopped watching television just before that.

Project targets are slipping. Hmmm. We adopted a new project management system just before that.

And so on.

And here's where the metal meets the meat: CorrectionRequiresHonesty?. If you just make up something to fill in the "what changed" blank, things are not going to get better. So, if you want actual correction, you gotta get honest with what's changed.

There are other factors, like getting the time frame right, like understanding how the change would create the currently observed effect, like knowing that "painted new stripes in the parking lot" is very unlikely to be responsible for the plunge in print quality of your brochures. Domain expertise is assumed. Indeed, required.

And why, since this is so Bloody Obvious (tm), am I bringing it up at all?

Well, evidently BloodyObvious != CommonSense. "What Changed" morphs into "What Can We Tell Management (tm)" so that they will not take us out back and shoot us? Occasionally, the morph is to "Who Can We Blame (tm)" -- and I guess I'll leave it at that.

-- GarryHamilton