In 2008, I wrote a mini series about Kaizen and Software Development. Kaizen or continuous improvement asserts the centrality of quality as the only way to achieve the trifecta of quality, speed and cost. Simply put incremental improvements around quality will result in lower cost and higher speed. Much of agile development has really been the re-invention of techniques long known in manufacturing. In my series I cover lot size one, no inventory, visualization, root cause analysis and customized tools.

I am excited to be facilitating a workshop this morning for the team from Shapeways about all of these topics. For Shapeways continuous improvement is doubly relevant because it applies not only on the software development side but also for their own 3D printing, their network of third party printers and their logistics operations tying it all together. In addition to the topics above we will also be covering statistical process control, which I somehow forgot about in that series of posts.

So what is statistical process control? The basic idea is simple: try to decompose variation in process measurement between inherent variability and excess variability that indicates a problem. A key tool used for this is the control chart which allows for the visual detection of excess variability and also of patterns that suggest systematic sources of variation.

An obvious example application in software development would be to measure and graph a web site response time. Response time will show natural variation because of the variability of the many underlying systems. But a big spike outside the 99% confidence interval should be investigated. Similarly a pattern where the site slows down in a predictable fashion – say every 10 minutes – should be investigated (even if that slow down stays well within the 99%).

I am often surprised by how little of the performance data that companies collect these days winds up actually being used that way. That may be another example of where we have to relearn the lessons from manufacturing. At many companies we have had the luxury of simply letting Moore’s law take care of things for us. Throwing ever more hardware at a problem is in many ways the equivalent of having inventories.

If you are an expert on continuous improvement or know one, please send them my way.  Also, Shapeways is still looking for a Global Director of Production and Distribution.

In 2008, I wrote a mini series about Kaizen and Software Development. Kaizen or continuous improvement asserts the centrality of quality as the only way to achieve the trifecta of quality, speed and cost. Simply put incremental improvements around quality will result in lower cost and higher speed. Much of agile development has really been the re-invention of techniques long known in manufacturing. In my series I cover lot size one, no inventory, visualization, root cause analysis and customized tools.

I am excited to be facilitating a workshop this morning for the team from Shapeways about all of these topics. For Shapeways continuous improvement is doubly relevant because it applies not only on the software development side but also for their own 3D printing, their network of third party printers and their logistics operations tying it all together. In addition to the topics above we will also be covering statistical process control, which I somehow forgot about in that series of posts.

So what is statistical process control? The basic idea is simple: try to decompose variation in process measurement between inherent variability and excess variability that indicates a problem. A key tool used for this is the control chart which allows for the visual detection of excess variability and also of patterns that suggest systematic sources of variation.

An obvious example application in software development would be to measure and graph a web site response time. Response time will show natural variation because of the variability of the many underlying systems. But a big spike outside the 99% confidence interval should be investigated. Similarly a pattern where the site slows down in a predictable fashion – say every 10 minutes – should be investigated (even if that slow down stays well within the 99%).

I am often surprised by how little of the performance data that companies collect these days winds up actually being used that way. That may be another example of where we have to relearn the lessons from manufacturing. At many companies we have had the luxury of simply letting Moore’s law take care of things for us. Throwing ever more hardware at a problem is in many ways the equivalent of having inventories.

If you are an expert on continuous improvement or know one, please send them my way.  Also, Shapeways is still looking for a Global Director of Production and Distribution.