The Middle-Manager Rotation: A Modest Proposal, Or Possibly A DevOps Parable

(This is almost-verbatim a memo I recently sent to company leadership.)

If we think the on-call rotation is working well and is an efficient use of personnel, I’d like to propose we fire all the middle-managers and replace the current management structure with an analogous “middle-manager rotation.”

My team’s developers spend—on average—about one day each month on-call for our products, and have about a month between their on-call shifts. We don’t currently have any full-time SREs assigned to respond to incidents for the product we’re building, work on observability, follow up on post-mortems, or other operational tasks. Instead, it’s expected that everyone chip in on this work when not responding to pages while on-call.

We could just as logically have each developer spend about one day/month as a director or VP, and likewise eliminate those positions.1 This would help level-up management skills amongst the team, give everyone an appreciation for middle-management’s role, and also save the company a bunch of money on payroll. The line managers could form a backup rotation, in case someone on the primary rotation is unavailable for their scheduled shift or is unable to respond to an email from the CEO within 15 minutes.

Of course, there are some reasons why this may not be the most efficient or logical way to structure an organization. I’ll address a few possible objections here:

First is the argument that since the industrial revolution, workers have historically specialized in their chosen vocation. A manager—or an SRE—has a specialized skillset that overlaps with, but is different from, the skills required to be a good software developer, just as a medical malpractice lawyer and a surgeon have different but overlapping skills. Certainly some software developers go on to become managers or SREs, but we generally acknowledge that these are different roles. But just as specialization may be an argument against expecting people to do a job for only one day per month, we should recognize that developers are smart and I’m sure they can figure it out. Being VP of Engineering for a day can’t be that hard, right? And how do we expect developers to learn, except by doing? I’m sure they will take the lessons learned from their management rotation into their day-to-day jobs and become better engineers as a result.

Similarly, it can be argued one needs to do a job regularly in order to hone your skills and keep your tools sharp. Despite rising inflation, most carpenters don’t moonlight one day a month as auto mechanics. A toolbox full of woodworking tools is not particularly useful for auto repair. (Although if every problem looks like a nail… or an ICS incident…) Time spent setting up and getting familiar with specialized tools—such as PagerDuty and Rootly and Teleport and Opal, or Excel and Powerpoint and Greenhouse and Workday—gets amortized over the time spent using those tools. It’s possible that it’s not the most efficient use of human capital to expect an entire department to spend half a day each month verifying access, installing software updates, reviewing runbooks, etc. ahead of their once-monthly middle-manager (or on-call) shift. But surely the payroll savings are worth it.

Finally, to be effective at a job, I think we can agree that one needs sufficient context. That’s why the team has thrice-weekly hour-long “operational review” meetings. A diligent member of the on-call rotation thus spends ~12 hours/month in meetings (or watching meeting recordings) to have sufficient context for their one day per month on-call shift. If we were to replace the middle managers with a rotation, we’d probably also need to expect people to spend a similar amount of time in management meetings from which individual contributors are currently excluded. Is this a good use of ICs' time? Should we be spending more time meeting, or more time doing? Who’s to say?

I realize there may be some downsides to this proposal. But as we get ready for our upcoming product launch, now is the time to figure out how we can be more effective and efficient as a team! Let’s get rid of the middle managers and see how it goes. Who wants to sign up for the first shift?


  1. The first draft of this essay included line managers, but to be analogous to the on-call rotation from a headcount perspective, they probably should not be included. We would only need a handful of SREs to replace the current on-call, so the middle-manager rotation should aim to eliminate a comparable number of positions. (Also, this is satire, and it’s funnier the more clearly it’s “punching up.”) ↩︎