If You Think You Can or Think You Can't, You Will Be Right

Long story short. She came home from school one day and excitedly reported that she had passed all her time tests! Mom asked her how she finally did it and she said, “Well I remember Dad telling me ‘If you think you can or think you can’t, you’ll be right,’ and I decided I can, and I was right!” More than twenty years later, she can still do complex computations in her head better than most people with a calculator.

One of my most enjoyable jobs was as the engineering manager for an OEM production line running a mix of printed circuit board assemblies (PCBAs) for use in producing our medical devices. The company had a mix of legacy and relatively new products on market, and therefore the circuit board line built about 100 different models of PCBA at various production rates ranging from 100 units/year for some legacy boards to 3,000 units/month for high volume new products. The engineering team consisted of 5 engineers, mostly EE with one ME, plus 3 technicians who handled machine maintenance and repair, setups, fixture maintenance, and the like. When I arrived at the company, the team was already in place, and most of the team members had been in their job for 4 years or more.

The team members were essentially home-grown – they had come to the PCBA team from other parts of the company, and did not have outside experience in board assembly. Procedures were stable, the guys knew what to do day by day, and things ran pretty well. My prior experience, however, led me to view things a little differently. I was accustomed to the Electronic Manufacturing Services (EMS) mode of doing things. In the EMS business you cannot tolerate defects in production, because that means you have built something but you are not going to get paid for it; the customers only buy the good ones. I had been steeped in the methods of Lean and Six Sigma for driving thruput, reducing waste, and improving product quality. The first couple of site quality review meetings I attended made clear that the PCBA line had a wealth of opportunities for improvement.

I started the job in November, and in January all of the business units had to set their goals for the coming year, focusing on quality as measured by first-pass yields. The final assembly business units each had one or two product models to plan; however, the PCBA business unit produced all the boards for each product, meaning 3 to 7 boards per product. Net, we had 16 different PCBAs for which yield goals needed to be set. I reviewed the previous year’s plan and actuals, and saw that the yields fell far short of the established year-end goals. We held several team meetings to figure out where to set the current year goals.

Four long meetings and lots of hallway conversations later, we established what in that realm were considered to be pretty aggressive goals. With a little gentle arm twisting from the newly-hired leader, the PCBA team set the goal of reducing the failure percentage of each board by half. That meant that for a board with 90% yield and therefore a 10% failure rate we set a year-end target of 95%. And a board with 70% yield (we had a few) and 30% failure rate got a new target yield of 85%. The Operations Director poked and probed, pushing for a little more on some boards and in a few cases asking whether the targets we proposed were achievable. My gut feel at the time was that if we did business as usual, we would fall short on every one of them. But I saw an opportunity to move the PCBA team’s capability to a higher level and felt it was better to set a high goal and possibly fall short than set a low goal and make it.  (Or even worse, set a low goal and fall short.)

The single biggest change we made in the team’s production line support activity was to pay daily attention to the yield of each board running in line, and use 6-sigma methods to set control limits and create failure Paretos, rather than only looking in aggregate (passes and fails) at the end of each month as had been the custom. The high-mix line meant that we rarely built a given PCBA for more than two shifts, so initially it was difficult; by the time we had a handle on what was causing failures on a given PCBA model, the run was done, and the line had been changed over to run another model. The opportunity to implement a corrective action and check its effect was gone until several days later when the board model would run again.