Improvement Insights Blog

Much of Six Sigma is focused on quantitative data (numbers). But what about qualitative data (text)? How can you use it to improve?

“I’m Jay Arthur, author of “Lean Six Sigma Demystified” and QI Macros [software].

“You know, most of Lean and Six Sigma is focused on quantitative data: numbers, things you can count, right? But there are qualitative data that we need to analyze. You need a little different toolkit for some of those things.

“For example, if you have a call center and reps are typing in comments from customers, that’s all words, right? How do we analyze words? Well, in the QI Macros we have a Word Count Wizard that’ll parse all of that into one- and two-word phrases, and then sort them so that you know what the most common word or phrase is. From that, you can start to determine what customers are talking about.

“When I was working with one hospital, we looked at their data like that and discovered that the most common thing people were calling and trying to straighten out was ‘duplicate date of service.’ They had two invoices going out and one was for the emergency room, one was for the admitting room, but those were on the same day. They had to combine those two things into one bill, otherwise it got rejected. Wow! That was interesting.

“Then there was a woman who was analyzing Twitter and she was trying to figure out what are the words that are most likely to go viral. She was using the Word Count Wizard to crunch all this stuff and to figure out what kind of phrases would work best on Twitter to trigger viral activity. I’d have never guessed that. She didn’t buy [QI Macros] for anything else but Word Count Wizard. Fascinating! Just fascinating.

“Now there’s also things like sales data where John in a certain zip code with a certain income at a certain age level bought a speedboat or something. Well… and then what? Those kinds of things… or Suzie with a certain income level bought a certain kind of purse… whatever. So what we want to do is figure out is there a pattern here, and can we then take advantage of that to focus our marketing efforts on people who are more likely to buy than people that are not likely to buy? For that you need something called Logistic Regression. There’s Binary [Logistic Regression] and Multi-Level Logistic Regression, and so you can crunch all that and then give yourself some things to start and play with some parameters, and see which one is most likely to deliver the best results. So we want to be able to take a look at data like that and then turn that into an improvement story, or direct our efforts in a useful way.

“That’s my Improvement Insight. Let’s start figuring out how to use that kind of data and make improvements. Let’s go out and improve something this week.”