Out of the 18 crews in the department, our crew was the most productive. Out of nearly 20 other facilities spread across the country and world, our plant was the most productive. 

That remained true regardless of who occupied the CEO seat. CEOs came and went, but our results stayed the same.

Yet that wasn't true when certain managers moved or left.

Take Steve, our supervisor. He didn't try to control things he didn't need to control. We were already (maybe too) competitive; he didn't need to drive our performance. We were constantly looking for ways to improve efficiency; he didn't need to encourage innovation or creativity. We were really good at training junior members of our crew; he didn't need to think about employee development. 

Instead, he focused on controlling things we couldn't control. Component availability. Job scheduling. Not to go all metaphorical, but smoothing paths we couldn't smooth, and eliminating roadblocks we couldn't clear.

Then Rudy became our supervisor, and our performance slowly fell apart. 

The same was true on a broader level. To be honest, while I'm not sure what Dick, our plant manager, actually did, whatever he did (or didn't do) clearly worked. 

When Dick retired, I was all too aware of what the new plant manager did. Transformational leadership initiatives. New training programs. New interdepartmental process flows. He put his stamp on everything.

And our performance, as a plant, slowly fell apart.

I know what you're thinking: Individual leaders, especially at lower levels in plants with more than 1,000 employees and in companies with over 30,000 employees, don't make that much of a difference. 

Especially when CEOs like Elon Musk send emails to employees that include lines referring to "a lot of middle-managers adding costs but not doing anything obviously useful."

Except they do.

A 14-year study of automobile manufacturing plants found that replacing a bottom-quartile (think "poor") plant manager with a top quartile (think "great") plant manager decreases the time required to build a car by approximately 30 percent. (Interestingly, plant-specific tenure has a dramatic effect, further proof that promoting from within is almost always the way to go.)

But wait, there's more: A paper published in Strategic Management Journal by professor Ethan Mollick, my go-to source for leadership research, found that managers in the video game industry can account for more than 20 percent of the variation in game revenue, a greater degree of influence than any other role.

And more: Startups that add middle managers become one-third more likely to develop product innovations

And the kicker: Replacing a manager at the bottom 10 percent of working with people with one in the top 10 percent lowers turnover and reduces overall cost by 5 percent. People often quit -- especially the people you can least afford to lose because the best employees always have choices -- because of their managers.

No matter how wonderful you, the founder or CEO, might be.

All of which makes sense. Mary Barra may be an extremely effective CEO, but she has little impact on the average GM line crew's productivity. The same, to a lesser extent, is true for that crew's plant manager. The farther removed you are from a person doing a job, the less impact your leadership -- no matter how outstanding -- has on that person, especially in terms of their productivity.

As Netscape co-founder Marc Andreessen says, "If you are super scrupulous about your hiring process, you'll still have maybe a 70 percent success rate of a new person really working out -- if you're lucky."

So, by all means, be thorough, thoughtful, and deliberate when you make hiring decisions. But be even more thorough, thoughtful, and deliberate about mentoring, coaching, and developing the people you put into leadership positions -- and dealing with ineffective or, worse, toxic bosses.

That will make a bigger difference than anything else you do on how well your employees do their jobs. 

And on how they feel about coming to work every day, because happy usually leads to productive.

And vice versa.