When “Red Beads” Lead to What Looks Like Workplace Discrimination
Quite often, I used the famed “Red Bead Game” (a.k.a. the “Red Bead Experiment”) that was made famous by W. Edwards Deming.
Chapter 5 of my book Measures of Success: React Less, Lead Better, Improve More is a narrative of the game and the lessons learned in workshops where it's used.
In the game, one management fallacy that's exposed is threatening to punish or fire people whose performance is the result of “the system.” In the game, setting an arbitrary goal of “3 red beads” doesn't mean it's achievable.
Firing the bottom half of our performers (or firing the “bottom 10%” if we had ten “willing workers” wouldn't be fair and it wouldn't improve performance to bring in a replacement worker to work in the same badly-designed system.
Here is the scoresheet from the game when I facilitated this last week. After the third round, the bottom half of performers were fired. It was a “performance-based culture” and a “data-driven decision,” after all (an organization might say).
When “willing workers” dip their paddle into the container, they “produce” 50 beads and red beads are considered “defects.” It looks like this:
The expected average number of red beads on the paddle after each round of production is 10. Since each round has six workers, the totals of 60., 57, 59, and 57 show some of the inherent
In the game, we're not doing anything to really improve the underlying system, so we'd expect the variation in results to be somewhat consistent. But there's always going to be variation.
The problem is that the round-to-round variation is much higher for each individual worker. The variation is high enough that it would be tempting to blame workers for their bad performance. But, in this case performance is absolutely driven by the system, which includes variation in the distribution of the red beads in the container (it's not a random-number generator).
The “Process Behavior Chart” that shows each individual paddle shows that the results are a “predictable system” with variation:
There's nothing but “noise” in those results. There's no reason to ask, “Why did you only get four reds? anymore than it's worth asking, “Why did you get 16?”
The same system generates all of those results. Rewarding, judging, or blaming people for random performance is ineffective, if not cruel.
In this post from last December, I wrote about how Process Behavior Charts could be used to compare individuals (or teams or sites) and their performance at a snapshot in time.
We can use Process Behavior Charts to see if the performance we attribute to an individual is within the range of “routine variation” or if their performance is somehow an outlier.
In the Red Bead Game, everybody is doing the same work the same way. I observed them and there was no cheating. Just a lot of hoping and wishing for better results.
When we compare the average number of red beads from each “willing worker,” we see that their performance is all within the same range. The results are driven by the system.
The average number of red beads by each per person look like this as a table of numbers:
Would you assume that the organization would be better off if we fired the bottom three and then asked the three survivors to run double shifts? Why did Donna, Patricia, and Tina get more defects? BAD LUCK.
When we draw a Process Behavior Chart, the calculated Lower and Upper Natural Process Limits make it much more clear that none of the six are an outlier.
Again, firing the bottom half (or anybody in this system) wouldn't be helpful. It might make a manager or business owner feel like, “Well, at least I'm doing something.” I guess that's a time when Dr. Deming would say, “Don't just do something, stand there.”
Leaders need to work to improve the system instead of blaming individuals for the variable results of a system that they're not allowed to help improve.
Don't fire people based on results that are driven by the design of the system and variation in the system. It's not helpful and there are other reasons it might make you look bad.
One time I facilitated the game. the bottom three performers just happened to be women (I forget how many men we had also playing). The three women got fired. “Performance based.” “Data driven.” But not just.
Somebody commented, in an off-handed and light-hearted way, “Oh, you fired all women… that's going to be a lawsuit!”
In the game last week, you'll notice we had two men and four women playing. All three of the fired workers this time also happened to be women. And, to make the appearance of discrimination worse, I also had to fire the two African-American women who were playing.
They realized I wasn't discriminating. Nobody brought it up. But, wow was I aware of it.
Everybody knows I'm just role playing as I facilitate. But I'd hate to be accused of being sexist or racist.
There is, sadly, too much discrimination that still takes place in the working world. Discrimination is wrong, whether you want to tie that to Toyota's “respect for people” principle or not. Discrimination is bad management. Bad management can sometimes create the appearance of discrimination.
My main point here is to stop firing people based on random targets and thresholds like “the bottom 10%” or “below-average performers.” I'd be curious to look at data that might get people fired from an organization:
- Software bugs in your code
- Software bugs missed in your review or testing
- Post-op complication rates
- # of cash register scans per hour
- Sales figures per salesperson
If we created a Process Behavior Chart, would we see a signifcant difference in people's performance? Maybe. Maybe not. The old Jack Welch / GE rule of “fire the bottom 10% each year,” is completely arbitrary. You could easily have a real workplace system where the performance of the bottom 10% is still within the realm of statistical noise and routine variation.
Is that any way to run a company?