Fluctuations on Continuously Moving Assembly Lines 3—The Value of Team Leaders

In my last two posts I described the impact of fluctuations on the duration of the work on a continuously moving assembly line. One key tool for reducing fluctuations on Toyota assembly lines is the team leader, the frontline support for operators. Team leaders reduce fluctuations in general by taking over irregular tasks like delivering material, taking away trash, and explaining situations to a manager. They also reduce especially long fluctuations by supporting the operators in their own work if there are problems, thus reducing stops of the line. This post will look in more detail at the beneficial effects of team leaders on a moving assembly line.

What Are Team Leaders at Toyota?

Team leaders are the lowest level of supervision at Toyota. A typical team leader supervises between four and six operators at the assembly line, while a supervisor coordinates four to six team leaders. Team leaders are promoted from the most experienced operators, typically after working on assembly lines for multiple years. They are normally directly at the assembly line, where their purpose is to support the operators and to support continuous improvement activities. They have a special marking on their cap (e.g., at Toyota UK two yellow stripes). Operators have no stripes, and the supervisor or group leader has one yellow stripe. Section managers have a red stripe. For more information, read my blog post On the Team Structure at Toyota.

Team Leaders Reduce Fluctuation in General

Team leaders are the direct support for the operators. On a normal assembly line, quite a bit of the work is what we in lean call non-cyclic tasks. These are tasks that don’t happen every cycle but have to be done anyway. For example, to take away packaging, to get new parts from a shelf, to get a new tool, to explain the situation to a boss, an so on. Such kinds of side tasks are a major source of fluctuation.

Team leaders at Toyota take away most of the irregular work from the operator, and hence the operator can focus on the actual work. At Toyota assembly lines, there are even signs that forbid anyone, including managers, from talking to the operator on the line—except for the team leader. All questions are directed to the team leader, who then decides whether he needs to disturb the operator. In effect, this significantly reduces the fluctuations of the operator on the front line.

Team Leaders Reduce Large Fluctuations in Particular

Team leaders help operators not only with supporting tasks. If there are problems on the work itself, a team leader is there to help. If the operator requests help using an andon, the team leader is there and will work together with the operator to solve whatever problem they have so that either a line stop is prevented, or—if the line stops—the duration of the stop is reduced. Hence, team leaders not only reduce fluctuation in general, but especially reduce the long tail of the randomness and make long events shorter.

Statistical Modeling of Team Leaders

In the last post I started a simple statistical model showing you the effect of fluctuations on the waiting time. Let’s expand on this model and show how a team leader affects the fluctuations on the assembly line.

Effect on Fluctuation in General

First, let’s have a look at the fluctuations of the operator in general. I slightly modified the random distribution of the work of the operators, reducing the standard deviation by 10%. Below are the original and the modified Weibull distributions, assuming a team leader reduces fluctuations in general. The mean did not change much, from 70.82 seconds to 69.75 seconds, and the standard deviation was reduced by 10% from 7.35s to 6.62s. As you can see, there is not much change to the distribution itself.

However, it does have a significant impact on the line. Assuming the assembly line has a speed of 90 seconds per cycle, the workers originally were able to make the speed of the line 82% of the cycles. In 12% of the cycles, at least one worker is slower than the assembly line, causing an average delay of 7 seconds if the line is stopped, or 0.93 seconds in average across all cycles.

If we add a team leader and the general fluctuation is reduced, the workers are able to make the speed of the line 91% of the time, and only 9% of the cycles the line is stopped. The average stop time is also reduced to a bit over four seconds if the line is stopped, or to 0.3 seconds in average.

Effect on Large Fluctuation in Particular

Next, we model the effect of a team leader directly supporting the operator in the case of trouble. We assume that it takes some time for the operator to realize a problem and for the team leader to come. Our assumption is that if the work would take more than 90 seconds (i.e., a line stop), the team leader is there to support at the time of 80 seconds within the cycle, and can reduce the remaining time between 80 seconds and the end of the work by 40%. This changes the probability density function as shown below, reducing a large part of the long work times.

This, too, has a large impact on the assembly line. Assuming the assembly line has a speed of 90 seconds per cycle, the workers originally were able to make the speed of the line 82% of the cycles. With the support of a team leader for the especially long cycles, the workers are now able to meet the line takt 95% of the cycles, and only 5% of the cycles are delayed. The average duration of a stop goes down from 7 seconds to 2.5 seconds, reducing the average delay per cycle from 0.93 seconds to 0.12 seconds.

Effect on All Fluctuations Combined

Finally, we combine the effect of the general fluctuation reduction and the reduction of the long work times. The initial Weibull distribution and the modified Weibull distribution having both a smaller general fluctuation and a special modifier for durations longer than 80 seconds is shown below.

Now the assembly line is able to make the cycle of 90 seconds 98% of the time, a significant improvement up from the 82% of the time of the initial situation without a team leader. If the line stops, the average delay is now only 0.2 seconds, significantly less than the 7 seconds of the initial situation. The average delay per cycle is reduced from 0.9 seconds to 0.03 seconds. The diagram below shows you the probability density Weibull function for a single worker, and the approximate distribution of a line of 10 workers with and without a team leader (estimate based on histograms… the pure math is really tough). The shaded area is the time saved from stopping the line. This shaded area is the benefit of the team leader.

The exact benefits of a team leader depend on your exact situation. The longer the line is and the more operators are on the continuously moving line (or segment without buffers), the larger the impact of the team leaders. The assumptions on the impact of a team leader could also vary, I tried here to be on the conservative side, and the impact of your team leader could be even bigger. This does not even include the chaos of constantly stopping and starting the line, preventing the operators from finding their rhythm. It also does not include the impact on sub-assemblies as well as subsequent lines on the starts and stops of the continuously moving line. You could even run the line at a faster speed!

Overall, team leaders are a significant benefit to reduce fluctuations on assembly lines, especially continuously moving assembly lines. But this requires an understanding of fluctuations… which many companies seem to lack. Many companies calculate only in averages, and, if you use only averages, there is no mathematical benefit of a team leader. Subsequently the team leaders got axed. Now, go out, understand the value of team leaders for fluctuations, maybe even use them, and organize your industry!