Tuesday, April 27, 2010

Quality Process and Quality Results

How do we assure that our customers receive the highest possible quality in our products?  The first thing that comes to many people’s minds is that we need to have plenty of inspection.  Defective products will get made and we need to catch them before they get to the customer.
Unfortunately, inspection (as well as various kinds of testing) does not catch every defect. No inspection process catches every defect.  Many let 20%, 30% and more escape.  Inspections can also catch products that turn out not to be defective.
In workshops, we demonstrate the unreliability of inspection by having each participant independently count the number of a certain letter (for example, “e”) that can be found in a text. All counters do not come up with the same number. If the letter were a defect, some would not be caught in the inspection process.
Inspection does not add value.  It only catches some of the defective products and leads us to other work that does not add value either, but does add cost.
Let us suppose that we have an inspection station that catches 80% of the defective products. Twenty percent will escape and go to the customer; either an outside customer, or the next step in the manufacturing process, an internal customer.  Those defects could cause costly problems. Defects that get to the final user can increase warranty costs.
Let us consider the defects that do get caught. If they are not repairable, the material, equipment usage and labor that have gone into the product are waste. To find out if a defect is repairable an analysis may be required. This takes labor, equipment, and floor space.  These are costs. If a defect is repairable there may be additional analysis to determine how to repair it, costing more labor, more equipment, and more floor space. 
The repair itself uses resources, people and equipment to do the repair and often additional material. Once repaired, the product must be re-inspected, meaning that the inspection facility must have a capacity to inspect everything produced (100%) plus everything found to be defective and then repaired.  This typically also takes labor, equipment, and floor space. It may not be possible to re-inspect the unit that has been repaired immediately, nor is it always possible to repair a unit that has been kicked out by inspection immediately.  Thus, we must have space to store the units that are waiting. We must keep track of them, so that we know what has and has not been repaired and what has to be scrapped.
All of this is cost without adding value. What does not add value is waste. Inspection and repair do not add value. In pursuit of world class and lean operations we must figure out how to eliminate inspection and repair.
I am not suggesting that we should simply stop inspecting. We do need to figure out how to improve our process wherever we can, so that inspection becomes irrelevant.  It is easier to do in some cases than in others, but the strategy is the same in the easier and the more difficult cases.
We apply statistical and non-statistical process controls to our process.  Statistical process controls are most useful where we can measure variables of equipment and materials.  Statistical process controls permit us to reliably monitor and take action when the process goes out of control.  Non-statistical process controls have to do with creating standardized procedures for (and with) the people who do the work, monitoring the standard procedures, and continuously improving them.
In both statistical and non-statistical process controls, the goal is to minimize variation. Variation in the process is what causes variation in the product.  The variation can come from the equipment and tooling, the environment in which it operates (temperature, humidity and contamination), variation in the material (with regard to all of the specifications) and people and the methods that they use.  The last source of variables often offers tremendous opportunities for improvement.  Standard procedures that are followed consistently help make visible the variables that most affect our results.
A third part of the strategy is to use the information that we are currently getting from the repair process.  All too often we do little nothing to document our repairs.  We can accomplish a great deal  with simple check sheets to tally the types of defects we find.  Instead, we fix it and throw it back on the line be re-inspected. 
In a plant making automatic braking system components a machine inserted two rubber grommets into holes in a casting.  Some grommets would go in too far.  Others would not go in far enough.  From time to time the line would be stopped and the grommet-inserting machine would be tweaked.  The tweaking did not improve things. Most of the time an extra operator had to stand at the machine and manually repair each unit that was defective—having grommets inserted too far in or not in far enough.  Once repaired the product would then be put back on the line. This was non-value-adding repair work.
We had that operator remove every defective unit, put a tag on it indicating the time and sequence at which the defect occurred, and put it on a cart, without making any repairs to it.  We accumulated a lot of units fairly rapidly.  We conducted what we called an autopsy of all the units, recording the data on the tag, along with observations and measurements of anything we thought might be relevant to the problem.
The first thing that we found was that we had a problem of considerable magnitude.  A high proportion of the units were defective.  We were also able to discover that the cause of the problem was variations in the holes into which the grommets were inserted. The inside diameter of the holes was out of spec on both the high and the low side. Many of the holes also had flash that could hang up the grommet.  The problem needed to be corrected where the casting was milled. The grommet insertion machine was not the problem.
The fourth part of the strategy is to have and instill in everyone a mentality that variation must be continually reduced.  As long as we are satisfied with variation being within spec, there are two many opportunities for the variation to slip out of spec.  Our goal should be to keep variables centered on the nominal, with as little variation as possible, when it comes to the product in process or the variables that go into the process.  These can be kept in statistical control, using statistical process control methods. Manual processes, or processes where an operator is involved, should be always done the same (standardized) way until we find a better way.  When we find a better way we standardize on that. We will discuss process standardization again in a future post.
Notes: Jim Winings provides one example of discussions about the effectiveness of 100% inspection. Dr. Edwards Deming has influenced my thinking on process quality.  See for example, The New Economics for Industry, Government, Education and Dr. Deming, The American Who Taught the Japanese About Quality, by Rafael Arguayo.

No comments:

Post a Comment