Quality Function Deployment: where to go after House 2?

I've used QFD within a Japanese company, specifically during the development of the Sumo Image setter. The reason I mention this is because their approach tends to be different to what one might see in American textbooks.

The approach we took was as follows:

Phase 1 What: Define the voice of the customers (includes external users, internal designers and technical experts, manufacturing, quality department, marketing, and of course the chief engineer.
Phase 1 How: Deploy sub-systems/sub-assemblies to meet the requirements

Phase 2 How: Deploy components to meet the requirements of the sub-assemblies in the Phase 1 How.

Phase 3 How: Deploy processes to manufacture the components in the Phase 2 How.

Phase 4 How: Deploy checking systems to verify (self-assure) the manfufacturing processes in the Phase 3 How.

Note: I've omitted the Hows transforming into Whats in the next phase for brevity.

Hope you make some bread ('Life on Mars' joke :-)

Andy Urquhart

In fact the question was where to go after house 1.
It is easy to speak generally about this; any text book will tell you the classic flow down from VoC to design to manufacturing to quality control.

What you do with QFD1 in practice depends upon what QFD1 has told you regarding competitor analysis, whether you have enough functionality (Hows) to meet (map onto) your requirements (whats). Do you have any negative correlations between requirements that need to prioritised/resolved with the customers? (The 'conservatory of QFD1). Do you have any negative correlations between functions (the 'roof' of QFD1).

Are customer requirements properly prioritised (via AHP, for example)? If they are and you have completed the relationship matrix then you should have priority ratings for the functions, and so may take a subset forward to QFD2 as 'whats'.
All customer requirements and functions require constraints (How Muchs in QFD1). Do you have these? Are you making bread or marketing it? What you do dictates the content of QFD2,3,4.

Assuming that you are a manufacturer of loaves of bread typical functions will be 'bake', 'protect', mix dough, control baking, etc. These are 'solution invariant' - they are the same for loaves of bread irrespective of the 'design' of the loaf. QFD2 involves determining design parameters for the loaves; there may be a separate QFD2 for each type of loaf. These would typically be things like 'ratio of yeast to flour'. Again, these require constraints (how muchs) at the foot of the QFD2. These are typically found via Design of Experiments in conjunction with some kind of optimisation. Once the target values for the parameters have been determined, along with tolerances which give an acceptable variation in performance of the loaf, QFD3 is concerned with taking the most important design parameters as the 'whats'(highest ranked, most difficult to achieve in practice, ones you've had trouble with before, ones involving new technology etc), and finding manufacturing solutions to ensure that the design targets are met with good process capability to the design tolerances. These manufacturing solutions form the 'hows' of QFD3. Initially QFD3 may be high level process steps, but ultimately they are detailed process steps consisting of process parameters with specifications. This may involve process simulation studies, pilot studies, DoE etc. QFD4 is then a risk assessment of the most important process parameters to define the level of quality control needed to maintain capability to the process variables. The higher the risk factor the more QC is required, via Statistical Process Control, sampling inspection, operator instructions, product audit, 100% inspection etc. etc.

Phil Rowe, Consultant, Six Sigma Group