Hacker News Reader

I'm not sure what the big mystery is. Management exists to provide an abstraction for information cascade (both up and down) and centralized command and control. The amount of management needed decreases for rote work, and increases for more creative or novel work in competitive industries. In a healthy, well-functioning org a manager will have a reasonably deep understanding of everything going underneath them. This ladders up to where a CEO can get in a room with 5-6 execs and discuss the biggest problems for the company with a group that is A) small enough to have an actual dialogue and B) have as much breadth and depth of knowledge as is reasonably possible for such a small group. Like a balanced search tree, this structure means any necessary detail can theoretically be fished out by going through log(n) layers, which enables scaling to tens of thousands or hundreds of thousands of employees.

Of course you could also do this without managers by just having certain individual contributors be the subject matter experts of record, but such systems end up being more complicated because you lack the clarity of every person regardless of role or position having one and exactly one boss. Now this doesn't mean this is how work gets done, it just provides a substrate for decision making. In terms of actual work, more complex (read: non-DAG) networks are always needed (see Conway's Law), but when conflicts or inaction arise the manager hierarchy is what allows problems to be resolved decisively.

Now obviously in the real world people are people, and management is often the first place dysfunction shows up. This is partly because of the Peter Principle, partly because it's just easier to hide a lack of competence and play politics when you're in the fog of the middle layers, but that does not invalidate the function of management. The implication of this is that the primary job of senior leadership is bullshit detection.

1 is not a particularly good explanation in my mind for two reasons:

(a) people are great at working not in their own interest, together, on something greater than themselves. In fact, I've observed a negative correlation between this and management, in that the spirit of being "in it together for the greater good" seems to be weakened in organisations with more management.

(b) if people inherently only work out of self interest, then so would managers and the whole thing would break down anyway.

> people are great at working not in their own interest, together, on something greater than themselves.

Only really true of small, tight-knit groups of people engaged in very similar kinds of work, where everyone knows that their contribution will be evaluated by everyone else. Hence small partnerships, co-ops or very early startups probably don't need management. These roles start to appear as scale increases.

> (a) people are great at working not in their own interest, together, on something greater than themselves. In fact, I've observed a negative correlation between this and management, in that the spirit of being "in it together for the greater good" seems to be weakened in organisations with more management.

This is entirely true, and almost nobody with either corporate or government power agrees with it. They're putting in massive effort to keep people in line because they're afraid of what happens when they aren't kept in line, despite evidence that people accomplish things just fine without constant threat of external discipline.

> So far the only way we know how to prevent this is to assign humans to guard the interest of the business.

Managers very much work in their own self interest, along with the rest of us.

In an interview, an engineer might be asked something like, "how many nodes did you manager/how many lines of code/how many microservices/etc.," as a way of gauging their level of skill and expertise. There is an implicit understanding -- on both sides -- that the complexity of the systems the engineer dealt is directly correlated with their expertise and talent, even if that kind of proxy is irrational when you really think about it.

Managers deal the same thing, except their questions are more along the lines of, "how many people reported to you/how big was you budget/etc."

In fact, I'd argue that self-interest is even worse on the management side, because the career stakes are so much higher. I have seen, on several occasions, top management commit to a strategy that was objectively failing with no practical path to success, because admitting that the strategy had failed would have gotten them fired by the board. It can be extremely catastrophic to the company when it enters that kind of failure mode.

There are many reasons we have managers.

Workers are not an abstract unit. They are human beings with human concerns and human availability. This needs to be accounted for separately from their work output. Managers allow you to create departments which are abstract units as a convenient interface over this problem.

Information flows are not always efficient or desirable to have broadcast to all individual workers. Managers serve as a node in this company information graph to help manage the reality of these concerns, and to provide a mechanism where information can actually flow in the opposite direction without creating chaos.

They're am imperfect solution to a worse problem.

>Why do we have management?

Management's purpose is to ensure that an organization, remains and functions as an organization. So basically 1, but not exactly...

Basically, when you have a herd of sheep that you need to move, and you need to ensure all the sheep move someplace and none wander off. What do you do? You get a sheep dog, to ensure all the sheep move in unison and towards the same place. That's what a manager is, a sheep dog.

As another example, if person A is rowing the boat backwards, and person B is rowing the boat forwards, and person C decides to stop rowing because he doesn't know what direction to row in, you're not really working as an organization and are not going to get anywhere. Management is there to ensure everyone is rowing the boat together in unison and towards the goals of the organization. If something is preventing someone from rowing, he's suppose to investigate and remedy the situation, as well as find a way to prevent the obstacle from occurring again.

The irony is that they themselves become obstacles in a way.

I read the book "Reinventing Organizations" by Laloux a few years ago and I can't recommend it highly enough for a non-cynical take on how to create healthy organizations.

The first few chapters talk through the history of human organization, from dominance-based tribes ("I'm the biggest man so do what I say"), through to modern corporations, with nominally-meritocratic hierarchy of management. And organizations where "we're all equal". He doesn't hold his punches when talking about the downsides of these organizational approaches.

Then the rest of the book does a deep dive on a lot of organizational approaches that are being tried out in different organizations around the world, to talk aspirationally about what he thinks comes next in how humans coordinate.

I think his answers to your two questions would be something like:

> 1. Principal-agent problem. Who do the employees work for? Self interest.

No, this sounds circular but the employees work for each other (and your customers). Healthy teams care about each other as people, and if they can form emotional connections with your users they'll care far more about making them happy than they will about ever care about the annual budget. Its a cultural thing - but it can be supported by process. Set up a lot of conversations between the developers and the clients. A manufacturing company started sitting sales people with the line operators. If the sales person didn't sell enough parts for the team, there physically wasn't work for the operators that they're working with.

> 2. Failure mode. ... This is why I doubt that management can be automated.

Yeah, hard agree on that. So don't try to automate management! The job of management in healthy companies in Laloux's book is to support the actual work being done on the ground, mediate conflict, provide advice when asked and set up conversations between people. The people doing the actual work will always have more concrete information about whats happening in a day-to-day way than management will. So let engineers make and own decisions.

In the book they talk about the Advice Process, where basically anyone in the company can initiate any change - so long as they consult with people the change will affect and take feedback before doing it. (The bigger the change, the more feedback). Management processes need to be things people who aren't management (but are affected by the processes) actually want.

I think we (obviously) need to experiment more with this stuff more in the startup / tech space. Traditional management structures are nowhere near the ideal form of how we make companies.

>I'm the biggest man so do what I say

This is still shockingly common method of how modern corporations are run. I've been in board rooms on numerous occasions where I was the shortest person by several inches and I'm statistically tall. Its not statistically possible for 20+ people to be in the 99.999th percentile taller than me unless its not random. In other words not much has changed in thousands of years.

Because control systems are only sometimes a good replacement for human relationships (even formal / limited ones, such as you find at work).

There are natural scaling limits around this.

If you think about what a "manager-less" work experience looks like, it looks like being an uber driver. I think we have existence proofs that manager-less work experiences are possible at scale. But job responsibilities need to be well defined.

Naturally, uber head office does not work like this.

Manager-less might also be a partnership, for example lawyers, hair stylists and sometimes software engineers. But say with software engineers as the “company” grows it is legally much simpler and also less risky to hire new people than to make them partners.

Why are you melding agency theory with pop science systems theory to replace sound economics (transaction cost theory of the firm)? I think you'd like to read Coase (https://en.wikipedia.org/wiki/The_Nature_of_the_Firm)

Benkler has also been posted here:

https://news.ycombinator.com/item?id=5723395