Design for Six Sigma (DFSS), now renamed Design for Lean Sigma (DFLS), is an important part of the Lean Sigma movement. One key approach in DFSS/DFLS is Quality Function Deployment (QFD), particularly with its ability to cascade both priorities and critical design parameters down from customer requirements to design parameters to functions, parts, and manufacturing processes.
Critical parameter management is a growing concern in the defense and automotive industries (in particular, these days), as the need to reduce complex systems requirements into its many subsystems and parts is becoming more important in global manufacturing. This is not unique to complex manufactured products, however. Its interest in the service sector is growing as well.
Healthcare is one such example. The American Society for Quality's (ASQ) Quality Institute for Healthcare has offered several papers and tutorials at recent conferences. The January 2007 issue of ASQ's Quality Progress magazine included an article on Six Sigma and DFSS using 'QFD' to address hospital medication errors.*
Despite receiving coaching from a Six Sigma Black Belt (not a QFD Black Belt®), their Traditional QFD effort (clearly book-learned) described in this particular article contained many errors and weaknesses. Since I was cited in the bibliography of the article, I feel obliged to point out errors on their use of QFD as well as efficiencies they could have achieved, had they received proper QFD training from a QFD Black Belt® and used Modern QFD methods.
The goal of this project in the ASQ article was to design a standardized medication order process in order to reduce errors. They identified the nurse as the "customer" of this process. Had they known Modern QFD, the "Customer Segments Table" would have helped them identify the chain of customers correctly — such as pharmacist and the attending physician also — as other "customers" of the process and "QFD value chain."
In the article example, the voices of the nurses were gained through interviews and these were prioritized by the project team by assigning a value that they thought represented the percentage of nurses who felt the particular voice was important. An example of the voices included "quick access to medication order information," "quick pharmacy turnaround time," "process must provide a history of patient medications," and others.
In Modern QFD, we would augment interviews with gemba studies where we would observe the nurses in action and capture real-time data of their issues, not limiting them to what they can recall in an after-the-fact interview. Then, Modern QFD would use the "Customer Voice table (CVT)" to translate the raw voices into true customer needs that tell us not "what" the customer wants, but "why" they want it. For example, "process must provide a history of patient medications" describes a functional requirement of the new system, not a customer need which should be independent of solutions. The CVT would translate this into true customer needs such as "I can track changes to the medication," "I can quickly see errors in dosing," "I can see if any medications might have interactions," etc.
The prioritization of these needs would be done in Modern QFD after the nurses did an Affinity diagram, Hierarchy diagram, and used the Analytic Hierarchy Process (AHP) to derive more precise importance weights. Arbitrarily assigning percentages, as done by the team in the ASQ article, is essentially an ordinal scale process and should not be used in QFD matrices. In fact, in the article case study, their importance weights add up to 135! Completely invalid mathmatically.
The hospital QFD team in the article then converted the verbatims into "success measures" which were to be used as design requirements for the new process. Their conversion process is not detailed but includes references to faxing, printing, and defect rates. In Modern QFD, we develop technology-independent functional requirements, not success measures or failures which can only be ascertained after the new product is implemented. Functional requirements should describe performance characteristics and capabilities that the new product must achieve.
The hospital QFD team in the article then combined the verbatims and success measures into a matrix and assigned relationship weights and calculated technical importance — again, improperly using ordinal scale numbers. Other matrix errors are also present. In Modern QFD, we would have used AHP to establish accurate ratio scale values. More importantly, we would have realized that this matrix was a waste of time... A quick read of their top two success measures (Fax to MAR time — 21% priority, and Process to enter medication orders 27% priority) related to the top verbatim (Quick TAT 50% priority) yields the same results as doing the entire matrix. The Maximum Value table (MVT) would have given us that same answer with far less time, effort, and resources.
Finally, the hospital team in this article developed technology concepts and prioritized them using a Pugh concept selection matrix. While I am not an expert in the medication process, it seems to me that the proposed concepts have little to do with the top two success criteria or the top customer need. In the Pugh chart, they used the verbatims to evaluate the concepts, and for the top verbatim of Quick TAT, all the concepts were evaluated as being the same as the current process. — All that work to come up with no improvement to the customer verbatim that captures 50% of their importance! Further, all the concepts exceeded the current process for the #2 need of "95% first time accuracy." So, in fact they still do not know which concept is to be preferred.
In Modern QFD, we would indicate alternative concepts in the Maximum Value table which focuses on the top few customer needs. That way, we can assure that the concepts address the most important customer needs. We would also use AHP to select the best alternative because it would allow us to first, consider the relative importance of each customer need and functional requirement and second, to evaluate each alternative's performance on a meaningful scale, like 'time,' instead of 'the same,' 'plus,' or 'minus' that the Pugh matrix uses.
The article reports that the hospital QFD team was able to improve its medication process, and so they deserve hearty congratulations. But their use of QFD, as instructed by their Six Sigma coach, probably took far more time and resources than necessary, and because their analytic techniques and processes were flawed, the conclusion they arrived is unlikely the best solution. Further, as in many companies, even if their QFD effort may have proved useful, with the approach they took, few would be willing to do it again because of the time and effort required.
What is unfortunate here is that many of their QFD application errors could have been easily avoided and the team could have produced a more powerful DFSS/DFLS result with less time and resources, had they used Modern QFD.
Modern QFD was developed to address the problem of overworked and understaffed organizations not being able to do all the QFD they should. The goal is to sustain the QFD effort beyond the first few projects so that the Voice of the Customer can be heard on all future projects throughout the organization. Additionally, Modern QFD has rigorous front-end tools for analyzing the Voice of the Customer to identify both the spoken and unspoken customer needs, leading to more innovative solutions.
Modern QFD and the tools mentioned here are taught in the QFD Green Belt® courses. MS Excel® templates for these new tools, as well as the Traditional QFD tools, are included. The course is a modest investment for anyone who is serious about Six Sigma and Lean Sigma product/service development beyond the mechanical exercises of matrices and conventional tools.
© QFD Institute