Seven years after one of the worst financial panics and recessions in living memory, the U.S. economy appears to be back on track. The headline unemployment rate dropped to 5.3% in June, with the Federal Reserve giving hints that it soon might raise interest rates -- a sign that they believe that economic growth is sustainable. But as a wise man once said, “It's tough to make predictions, especially about the future,” and for the people who manage and operate production facilities, caution is the watch word.
The problem for American manufacturers is that while it looks like somewhat of a return to normalcy, on the international front there are storm clouds. Europe, for all intents and purposes, has been an economic basket case, showing habitual weakness, even leaving aside what is currently going on in Greece. This uncertainty has caused the U.S. dollar to appreciate, coming to within 5 cents of parity with the euro in April of this year. Great news for honeymooners looking to visit Paris. Bad news for U.S. producers looking to export their products to the Eurozone.
We Decide to Investigate Inherent Instability
To stay competitive, manufactures must lower costs and improve operating efficiencies any way they can. We did an investigation into how managers could possibly take advantage of the instability inherent in many production lines, with some surprising findings. The lines studied were unpaced manual production lines, i.e., a series of connected stations in which workers were free to operate at their own pace (giving rise to imbalance issues) and in which the stations were subject to machine/tool failure.
A series of simulation experiments were done in which multiple factors and variables were looked at. The Windows-based simulation software program “ProModel” was used to create a hypothetical (but representative) production facility and observe its behavior over millions of simulated hours of operation. The goal was to compare the behavior of balanced lines to that of unbalanced lines to see if there was any way to take advantage of the relative instability that is present in almost all production processes.
Almost any operations management textbook (along with almost any operations management consultant) will tell you that a balanced line configuration is best. That is to say, all workers and machines should be functioning at the same rate, since a slow station will cause a “bottleneck,” with a resulting decrease in production rate.
But perfectly balanced lines are almost never found in real life. Furthermore, there is research going as far back as 1966 showing that by having a fast station in the middle of a production line and a slow station at both the front and back (something known as the “bowl phenomenon”), output could actually be increased.
We looked at lines in which the operators could be designated as slow, medium or fast, according to their average processing rate or meantime (MT). It is a well-known fact that people, due to having different levels of knowledge, skills, competencies, abilities, motivation, etc., perform tasks at different speeds. This phenomenon was incorporated into our study by way of four patterns of worker MT imbalance:
- A decreasing order (\): The slowest worker is at the beginning of the line, with workers getting progressively faster as you move toward the end.
- An increasing order (/): The fastest worker is at the beginning of the line, with workers getting progressively slower as you move toward the end.
- A bowl arrangement (\/): The fastest station is in the middle of the line with the two slowest stations located at both the front and back.
- An inverted bowl (/\): The slowest station is located in the middle of the line with the two fastest stations located at both the front and back.
It is important to note that no matter what pattern of worker imbalance used, the overall total average for all of the individual processing time for all of the workers was always 10 minutes. This was done for comparative purposes.
To make the study more realistic, the stations were also subject to random failure. Since it is logical to assume that tools and machines can break down with varying frequencies, the chance of a breakdown was set at three different levels, with a station being considered as having a high, medium or low degree of reliability. The breakdown and repair values used in this study are listed below in Table 1.
For machine reliability imbalance, once again, four different patterns were used:
- A decreasing order (\): Tools and machines get less reliable as you move down the line.
- An increasing order (/): Tools and machines get more reliable as you move down the line.
- Inverted bowl arrangement (/\): The most reliable tool or machine is in the middle of the line.
- A bowl arrangement (\/): The least reliable tool or machine is in the middle of the line.
No matter what pattern of station reliability employed, the total amount of time that a line was down was always the same (in this case, 4,000 minutes for the 20,000-minute production runs used in the investigation). Once again, this was done so that comparisons could be made between the relative performance of balanced lines and unbalanced lines.
All possible permutations of combined worker imbalance and station unreliability were simulated. Both 5 and 8 station lines were looked at, with inter-station buffers set at 1, 2 and 6 units. This resulted in almost 400 different simulations being performed.
We found that when one looks at line throughput and average buffer levels, the performance of an imbalanced unstable line isn’t that far off from a perfectly balanced line. In fact, there were certain patterns of combined worker and unreliability imbalance that actually improved performance.
With regard to the production rate or throughput rate (TR) of a line, while some improvements in performance were possible through imbalance, they were not enough to be statistically significant. This would lead us to say that in terms of output, a balanced arrangement is most likely best. But at the same time, where performance was found to be inferior (compared to a balanced line), this too was not statistically significant.
In fact, if one wanted to make a ranking of the “best” imbalance pattern, it would be a bowl-shaped (\/) MT arrangement in conjunction with an inverted bowl (/\) for reliability (i.e. the fastest and most reliable station is in the middle). In many instances these patterns perform just as well as a balanced line.
Looking at the average amount of material being processed in a line, otherwise known as the average buffer level (ABL), the results were less ambiguous. Here, many occasions for significant improvement were found thanks to imbalance, with the best arrangement being a descending (\) MT pattern combined with a descending (\) allocation of reliability (i.e. workers get faster and machines get less reliable as you move down the line).
A second best configuration was a descending (\) MT pattern plus a bowl arrangement (\/) for reliability (i.e. stations get faster and the least reliable station is in the middle. For the best and second best combined imbalanced patterns, it was found that reductions in ABL of up to 74 % and 71% (respectively) were possible.
Overall, our findings bring into question the conventional wisdom that balanced lines are always the best or that resources should be expended to bring a line into balance. For output, certain unbalanced configurations performed almost as well as a balanced line. For ABL, we saw vastly superior performance through deliberately unbalancing a line. The problem is that it appears that managers are forced to make a choice between maintaining their output or vastly improving their ABL. This decision will most likely be informed by the type of industry one is working in (high labor/product demand vs. high inventory costs/lean operations).
Sometimes it might pay to balance a line; sometimes it might not pay to balance a line. In terms of ABL, we found that unbalancing (using the unbalanced patterns that we described) provided serious improvement. In terms of TR, less so. But even here, sub-optimal results from the unbalanced patterns we suggest were not statistically different from those of a balanced line.
The main purpose of this article is to suggest that you might want to question the kneejerk reaction to always try to bring a line into balance, something that almost all operations management textbooks tell you to do. But every production line is different, hence we also offer a word of caution about making any changes -- your results may vary.
We're not saying that we have "The Golden Rule" that will solve all of your problems, but rather that there are other possibilities that you might want to investigate.
Tom McNamara is an Assistant Professor at the ESC Rennes School of Business, France, and a former Visiting Lecturer at the French National Military Academy at Saint-Cyr Coëtquidan, France. Before joining academia he worked at a Fortune 500 energy company.
Sarah Hudson is an Associate Professor at the ESC Rennes School of Business, France. Her research interests are in the field of industrial and organizational psychology and production science. She is the co-author of several articles in the area of production line efficiency.
Sabry Shaaban is a full professor in the Department of Economics, Strategy and Organisation at ESC La Rochelle (La Rochelle, France), where he teaches courses in operations management, production and inventory management and business statistics. His research interests include studying the performance of a wide variety of production lines.
Hillier, F. S., & Boling, R. W. (1966). Effect of Some Design Factors on Efficiency of Production Lines with Variable Operation times. Journal of Industrial Engineering, 17(12), 651.