It has been said that the first step in making improvement is to measure. However, it is more productive to consider the first step as taking responsibility. When implementing a program of improvement it is necessary to correctly assign responsibility for improvement to the various people best located to effect improvement. These are the people day in and day out whose responsibility it is to operate the manufacturing system i.e. the production operators, front line technicians and supervisors. It is therefore important to engage, empower and energize the right people and make them feel responsible if onewants to improve productivity. 
The production management and technical support staff are responsible to support all efforts in correctly operating the processes as designed, and to redesign the processes if given pertinent information suggesting such a change is necessary. One of the leading questions when performing a review of a production area is – how do the production people know that they are doing a good job? Perhaps a good starting point could be that all the tasks and procedures an operator should perform in their day to day work should be explained and documented. But just as important some key metrics should be established with graduated performance goals explained along with suitable escalation actions to be taken if those goals are not achieved. Along with this a method should be available to allow the production team to easily measure their performance against thesekey metric. --------------------------------------------------- Keys to KPI success The challenge in manufacturing has always been getting access to all the correct data, collating and processing this information to provide meaningful measurements on a processes performance, and to do this in a timely manner. So a successful improvement program must: • Collect the right information
• Process the information in a timely and meaningful manner
• Communicate this information in an easy to understand manner
• Engage the production people as selfactualizing high performance teams --------------------------------------------------- Productivity
Does one consider a production line to be operating effectively based on the fact that the machines are always running, or is there more to it than that? Take for example figure 1 showing a line with a high level of productive time of more than 94.9 percent in an SMT line. 
In this case this was an effective line with very few assists and interruptions as shown with the blue and red areas of the graphic. However, if half of the nozzles on each head were disabled or the accelerations for many of the components slowed down, the line would still show itself to be operational and have similarly high levels of productive time. However the actual output of the line would be significantly reduced. So clearly this KPI percentage "productive machine state” is not an accurate reflection of productivity and a higher level of productiveness is required. The main function of an SMT production line is to place parts on a PCB so a simplified top-level metric for this can be the total number of components placed over a certain time period. All other metrics are the result of factors that influenced this key metric, such as the ratio of productive time to scheduled productive time. The detracting factors for this KPI can be found in both the productive time of the line (how well a program is optimized for a line) and the non-productive time (stoppages due to lack of operator attention, or changeovers, etc.) This is a relatively easy setof data to collect by simply collecting the number of components placed per hour and calculating an average for thetime period under review. 
Figure 2 is an auto generated report that provides many of the KPI and input to KPI for a placement line. From this graphic one can see the components placed over a defined time for average components placed per hour calculation, and also for an hourly view of this performance. By monitoring this it would be easily apparent if the line had changes made to it that slowed down the placement rate such as nozzle skips or acceleration restrictions. But also downtime due to maintenance, breakdowns, and changeovers can be observed allowing these issues to be controlled. Perhaps the only exception in the use of this metric as a KPI could be in the case of an NPI line engaged in extremely small batch NPI activities. In this case perhaps times taken for activities such as; setup and teardown, line changeover, first of verification, etc could be used as the KPI. However, components placed per hour may still be a valid sub KPI as it reflects the optimization off the line and therefore potential performance potentials when the product moves to volume production. Process quality performance
In a high mix environment one product may vary from another in a significant way, such as component complexity or component quantity. It is therefore not so easy to use product yield as a direct measure of process performance. For example, a product with 100 components will not have the same yield as a product with ten times as many opportunities for defects such as a product with 1,000 components. Similarly two products with the same component count will not have the same yield as another if the first only places 0603 devices and the other places components by a factor of four smaller such as 01005 devices. And from one time period to another, these product to product differences may have an effect on the overall yield even if the underlying core processes have not changed. Hence it is necessary to switch to a more proportional process metric such as one of the following: defects in proportion to opportunities; defects per thousand components; defects per million components; or defects per million opportunities. Without a sophisticated set of automated data collection and analysis tools it would become very tedious to collect and collate all these information for production. Therefore it is suggested that a few high volume products are chosen and every week an activity sample of process performance is measured in order to establish a baseline and drive improvement. Of course when there is a chance to have direct feedback such as "inline” inspection then that is better and the use of statistical control tools also can aid in process improvement. In cases where in line automatic inspection machines are used then direct DPMO measurements can be made by correlating the components placed per time period versus the defects found per time period. Material control
Here one can use some basic metrics such as: scrap per process stage in quantity or value; material shrinkage rates; and work order tact time through processes. Directly at the SMT line, one can also measure the identity and vacuum errors per station as this is a leading indicator for process quality and productivity. There are additional benefits from actively tracking and controlling this metric. These include less chance that a particular component shortage is encountered due to one feeder in the department wasting excessive quantities from some avoidable process setting or corrective action. Even in the case where component pick up errors are slightly higher than normal this can directly affect the bottleneck cycle time of the placement process leading to underperformance in productivity. Cascading the KPI
In order to fully engage the production team one must also know how performance on a particular issue effects and influences the performance of all other systems and the success of everyone at all levels. Therefore, it is necessary for one to understand across all levels the KPI and how each section’s performance can effect and influence every other area. One method to illustrate this is by linking the KPI at all levels and areas to each other through a system of cascaded KPI as shown in figure 3. By cascading the KPI up and down the levels of the organization one can establish the importance of the metric and if the KPI is carefully selected its importance is relevant to all. In the case shown to drive productivity in SMT the average components placed per hour is used. In order to make this KPI relevant to the production team, only the average over time actually available for production is used, time due to material shortages or equipment breakdowns are excluded. By measuring at the department level an overall baseline can be established and long term improvement goals can be defined. Certain projects and initiatives can be implemented at the engineering and management level to have an overall effect on this KPI. Any benefits from these projects can be visually demonstrated on the KPI charts, and the improvement success can be celebrated by all enhancing the overall team morale. By measuring at the shift level, the performance of each production shift can be compared to another. If significant differences occur then analysis can be undertaken to find ways to raise all shifts to the highest performing level. This can serve to act as a communication aid between managers and shift leaders; production and engineering support, and even between different shift teams to positively encourage competitive improvement. By measuring at the line or team level the metric and performance can become personal. The individual as well as the production team can see directly how their performance cascades up and affects the overall KPI metric. Here again comparisons can be made between teams to search for ways to positively encourage competitive behavior and improvement. This system of review and dialogue at all levels and between levels should serve to energize and drive improvement in a positive manner. The use of simple and clear visual control systems can serve to enhance this process. Going back to the point about responsibility one of the most powerful tools available to ensure people feel responsible for their own performance is to simply ask them personally to hand write on a large white board how they are performing. The fact that they personally are recording for all to see how they are performing against clear targets reinforces the feeling of responsibility instilling a sense of pride when achieving good results and a desire for improvement should the performance is below par. Conclusion
By the careful selection of KPI, implementing a cascaded system of performance measurement and improvement goals, combined with visual management programs it will be possible to develop an energized and driven organization well prepared for the improvement programs that are essential elements of a build-to-order system. EM About the author
Robert Gray is the Product Portfolio Manager at Siemens Electronics AssemblySystems. For further information, email siplace.sg@siemens.com | 
