Manufacturers around the world have worked on standardizing practices and making assembly tasks less complex in an effort to keep their employees productive on the job. Wharton professor of operations, information and decisions Ken Moon co-authored a paper that examines the wisdom of that approach. He and his fellow researchers looked at a manufacturer in China that produces millions of units of electronic consumer goods per week but had an employee turnover rate as high as 300% annually. In their paper, “Manufacturing Productivity with Worker Turnover,” the researchers discuss not only why such turnover affects manufacturing output, but they also found a way to stem that turnover, potentially saving companies millions of dollars. Moon joined Knowledge at Wharton to talk about his team’s research.
An edited transcript of the conversation follows.
Knowledge at Wharton: In your paper, you and your co-authors say that worker turnover was once endemic in the early days of manufacturing and is now making a comeback. Why is that so?
Ken Moon: The historical aspect is quite interesting. The problem is new today, but it wasn’t always. In the early days of manufacturing, the workforce was very transient in some sense. People had a number of explanations for this, but it was very typical for companies to be turning over their workforce at least once [a year] in the early part of the 1900s, when assembly production and other ways of producing were just coming into play.
A great example that offers parallels to our setting is the Ford Motor Company. In 1913, Ford had a turnover of 400%, which meant that they were filling every seat in the factory four times over with a new person over the course of a given year. In 1914, that turnover rate suddenly dropped to 56%. That’s when Henry Ford instituted his famous $5 wage [which was substantially higher than before], so compensation made a difference. The question — and if you look at the past literature, it’s still a bit of a mystery — was why? Why did he offer a pay wage that was higher than necessary?
Knowledge at Wharton: It was $5 for eight hours of work, correct?
Moon: Yes, and that was for very unskilled labor where each task was very simple. It didn’t require much training. It was very hard to make mistakes. [The jobs didn’t] require skills, so many people were coming in.
There’s a recent article in the MIT Sloan Management Review on present-day re-shoring, which is bringing back manufacturing to the U.S. The No. 1 concern cited in that article is turnover. I’m paraphrasing, but [the production managers] say, “We’re hiring people on Monday, we train them Tuesday and Wednesday, and we find their badge often in the trash on Thursday or Friday. We never see them again.”
“Is it really the case that because these production processes are so thoroughly de-skilled, the worker is replaceable?”
…Companies have come to understand that we need specialized labor, so they created HR departments, benefits, all these things that are part of our life today but didn’t exist before. They decided to retain these workers. [Yet fast-forwarding to today,] you see this reversal, and not just in manufacturing. Distributors, retail — turnover is a frontline issue [everywhere].
Knowledge at Wharton: What else are companies doing to try and stem turnover, and have those measures been effective?
Moon: I think the shortest answer is that they haven’t been doing an effective job. You do see a lot of movement here in the U.S. towards what are called flexible workforces, or a workforce where people aren’t necessarily tied to the company all the time. That is [happening] not only with contract workers or freelancers, but it’s also the case within companies.
Another trend is that there is so much more data about [a company’s] workforce, and that’s being used to raise productivity. For example, I went on a trip with Morris Cohen, a professor of operations, information and decisions at Wharton, to one of the largest pharmaceutical distributors here in the U.S. They raised worker productivity dramatically over the past 10 years, but their turnover had gone from single digits, to low double digits, to over 40%. [Companies] are focusing on … the incredible data they have about their workers and trying to make them more productive. But it seems that there’s maybe a missing piece [in] not addressing the issue of turnover. Part of it is also that it’s hard to measure and to know the effect turnover [has on their firm].
Knowledge at Wharton: Your paper discusses that at length. The assumption might be that because manufacturers have worked so hard to make assembly tasks less complex — you call it “de-skilling” the process — it would be easy just to slot one person in for another, and the effect would be minimal. What did you find is the effect on these companies when people leave?
Moon: That’s a great question, and one that we wanted to answer. Is it really the case that because these production processes are so thoroughly de-skilled, the worker is replaceable? Here, again, we’re looking at contract manufacturers located in China. They’re producing consumer electronic goods like smartphones and tablets, and producing millions per week. In this setting, we actually find an effect.
Take a [production] line that suffers a typical amount of turnover — typically, one line per week is suffering almost two standard deviations below its normal performance in what’s called yield. Yield represents the percentage of units that are actually viable, that can be put out for sale for a customer, so the product has been put together in the right way.
But we’re able to go a bit further. We’re able to find the source of this effect: It’s not that the new worker necessarily performs worse than the one who left; it’s that the workflows and the social relationships on the line matter.
The way that we look at it in the paper is to think of the following: Imagine that there are three production lines. On one production line, you have no turnover at all, so the line’s staffing in terms of workers’ experience is left intact. On the second line — the same week and the same type of line — you have 10% turnover. You put new folks in. Once production starts, you expect that the new folks maybe aren’t quite as acclimated to that particular line. If one of them is making small mistakes, that affects the entire workflow. It affects her co-workers downstream and possibly upstream. So, you expect to see an overall productivity effect.
Where it gets interesting is the third line, where let’s suppose there is 10% turnover just like that second line. But now we look for replacement staffing from around the factory by sourcing workers who already obtained experience elsewhere. Essentially, we replace all of the line’s workers who have left with new people who are just as experienced. How does this line compare to the first line where nothing happened and no one left? How does this line compare to the second line where a lot of people left and you’ve plugged in these novices?
We find that the third line performs much closer to the second. It seems that [the right focus is] not just, “I need someone to ably carry out the task at hand.” It’s on the disruption to the workflow, and the disruption to these relationships through which people are able to manage small problems that arise on the line.
“One thing to recognize is that a number of factors affect when workers decide to leave.”
Knowledge at Wharton: It’s an impact on the team, correct?
Moon: Exactly. This sort of disruption is what’s costing the company in terms of dollars.
Knowledge at Wharton: In the company you looked at, how much was at stake? What were they losing from this turnover?
Moon: It was a larger effect than we anticipated. It was around 5% of variable costs. For this particular company, because they’re producing at such high volumes, around 5% is in the hundreds of millions of dollars.
Knowledge at Wharton: You said about $135 million per product. That’s pretty substantial. What kinds of tweaks could these firms make to help mitigate this?
Moon: There are [several] things a firm can do, depending on the situation. One thing to recognize is that a number of factors affect when workers decide to leave. [For example, workers] are affected when other people leave their lines, and then they tend to leave as well. It’s a little bit worrisome. When someone leaves, you have to think, “How do we stem the tide?” One way is through compensation, which ties back to the story of Henry Ford.
Again, we run into this question of why install this new $5 wage? People suggested a number of potential explanations. They said, “Well, there are costs of replacing people. If someone leaves, I have to recruit, I have to train that person.” But compared to the cost of the $5 wage program, that cost was very small. Also, it was hypothesized that in this new assembly manufacturing workplace, which remains very similar to the setting we see today, people needed to work very quickly and consistently and exert a great deal of effort. Maybe this higher wage was a way of incentivizing them to do that. The fact that turnover in Ford’s factories dropped from 400% a year to 56% would then be a nice side effect [of the new wage], but not Ford’s main objective.
But from our study, what we’re finding or suggesting for that scenario is that maybe [lowering turnover] was the manufacturer’s goal, because turnover disrupts assembly manufacturing — the reason being these workflows, these relationships that were becoming more and more critical in manufacturing. Now when you think about compensation, it no longer necessarily centers on one person’s performance…. It also becomes, “How do I think about the performance of the process? How is the company creating value from workflows, and how should I think about protecting that?” It’s a nuanced and involved analysis that’s needed.
Knowledge at Wharton: If you’re a large company and you have to make even an incremental increase to try and retain people, does that mean you need to shrink your workforce? Is that OK?
“[If] you keep workers around longer by better compensating them, you will hire less. You go through fewer workers, and your workforce shrinks just from that.”
Moon: It’s interesting because these facilities are huge. You have hundreds of thousands of people going to work in a single facility. In some of these buildings, essentially an entire wall would be a door, and still companies stagger start times because people would get stuck there in a bottleneck. You have a lot of people moving, a lot of product moving [into and out of these facilities]. When a workforce of that size [continually turns over], there’s even the question of how long is that sustainable?
What we find is, given [this kind of] facility and production process, if you keep workers around longer by better compensating them, you will hire less. You go through fewer workers, and your workforce shrinks just from that.
But there’s a second effect. A lot of turnover introduces uncertainty…. You don’t know who is going to leave when, and how that might trigger cascade effects on a production line when workers start to leave in part because their friends have left the firm. To deal with this uncertainty, you over-hire. That’s part of the analysis that we go through in the paper, and that has a very real cost.
In other words, when you start to control turnover, there are two main [results]. One is, “I raise compensation, and I keep people around longer [reducing disruption].” Your firm is no longer as much of a rotating door or turnstile. The second is, “I have less uncertainty, so I hedge less. I don’t need an excess workforce sitting around or to prepare for [production shortfalls caused by] turnover. I’m no longer putting out fires.” It becomes a much more stable and productive environment for the firm and for the workers. It looks to us like it could be a win/win.
Knowledge at Wharton: What are you thinking of next in terms of this research?
Moon: There is no shortage of questions with a data set like this. This is the first time that this company, and I think many companies, are putting together this type of production data, where it’s down to the person level – with compensation data, planning data like production plans. It opens up a lot of questions. First, what will companies do with this data? Particularly, what good can come of it? There is a lot of opportunity to identify the key ingredients — including the workforce — of a company’s processes for delivering value. How can data help? Potentially, that can be good for all stakeholders.
