[hasoleju]

The problem with too many layers of management is mainly that they stop doing what they should do. Making decisions.

The lowest management layer cannot make any decisions alone, otherwise the next layer would have no reason to exist. The next layer might even have to hand over the decision to one layer above. Up there no one really understands the decision that needs to be made because they are too far away from the actual impact. The result is that the decision is not made at all. Instead they procrastinate by asking for more details about the decision and the options.

If you have only one layer you get a decision very fast.

[tsimionescu]

There is some limit to how many people you can manage effectively. So, in a company with, say, 10,000 employees, you can't have just one layer of management between COs and ICs. You'd either need each manager to work with 500 direct reports, or you'd need the COs to work with 500 managers. If we look at 100,000 employee orgs, this gets even sillier.

If you even attempt something like this, what will actually happen will be a chaotic hierarchy arising within the larger group, with people playing politics to rise above their peers.

Edit: I should add that I absolutely believe a bottom-up organization can work, but it needs to be built with the right organizational model in place. You can't just say "no managers" and expect people to work directly for the CEO. You can put in place rules so that employees can self-organize, appoint representatives to interact with other teams, vote for collective decision making and so on. But this typically requires a huge power loss for the C-suite and owners, which is why it's very unlikely to work well outside of a co-op.

[silvestrov]

Depends on if you are managing people or you are micromanaging people.

It should be possible to manage 22 employees if you are good at setting directions and explaining why those directions have been chosen. You can have 22 people in one room and do functional group discussions, this is like a good size for a school class.

    1 CEO -> 22 level A managers
    each level A manager -> 22 level B = 22 * 22 = 484 level B
    each level B manager -> 22 level C workers = 484 * 22 = 10648 workers.

I have yet to see a company with 10_000 employees and only 2 levels of management between workers and the CEO.

If you add another management layer, it is only 10 people per manager. This is not much and quickly gets into micromanagement if the manager works fulltime at managing.

[iteria]

22 is so many direct reports. If my supervisor had 22 direct reports. I'd just check out mentally because I know I wouldn't get any real face time with them to discuss issues. Skip manager attention? Forget it. With 22 direct reports, you're not there to add your expertise and make the process better, but to be a cog in the machine.

[silvestrov]

Maybe there is a cultural difference here.

I see repeated 1:1 sessions as micromanagement, like being treated as a kid that needs their hand held.

I don't want to discuss work items on 1:1 with the manager, I want to discuss them in a group setting because it is the group as a whole that solves the tasks. I also want the input from other members in the group instead of only the managers view.

In my country we do personal "how is everything going" 1:1 meetings at most once a year.

[eitally]

No, there isn't a cultural difference.

1. If you want to discuss work items in a group setting with 22 people, it's an inefficient use of time.

2. If a manager has 22 direct reports, it becomes nearly impossible for them to do anything but manage, which make the whole team less efficient.

3. A manager ought to be expected to, and have time to, mentor their employees. This is impractical with such large teams, and the manager will always end up preferring just a couple at the expense of the rest.

4. Micromanagement is inefficient and unappreciated, but in such a large & flat team, it's very hard to ensure streamlined horizontal communication such that optimal decisions can always be made by ICs (or above them, managers).

5. There are two reasons organizational layers exist. The first is to allow domain specific experts to focus on their specialties in areas of the business that required dedicated staffing. The second is so the organization is setup to continuously mature over time. If you have a completely flat org almost entirely composed of IC SMEs, attrition will end up killing you because of the disjointedness and lack of corporate tribal knowledge creation you miss when you don't have functional teams. Note that the second piece almost always introduces some functional inefficiency in the org, but it is a reasonable risk mitigation practice for most companies.

Personally, I think "how is everything going" should be discussed at least every month, and with more casual check-ins that are more frequent than that, and task-based. I have high expectations for managers, and do expect that a significant portion of their time should be spent helping the people beneath them learn & grow.

[OnAYDIN]

Actually 20 people is around a reasonable number if you have some tech/expertise leads in the mix. 2 to 4 tech leads would be more than enough. Tech leads would be responsible for the technical decisions under their respective area and course setting for the future work but they don't do people management.

Managers do not need to mentor. Actually an independent mentor is better as there would be no conflict of interest. A person can get a mentor from another team as long as both are willing. If there are more mentees seeking mentorship than mentors, you can rotate the mentors or double/triple assign them especially to those who seek to switch to management track. In reality though, there are more mentors than mentees it seems. At least that's my anecdotal observation.

20 people management is very well possible. You don't need to have 1:1 with directs every single week. If you do it bi-weekly (every two weeks), the meeting cost to the manager would be 2 sessions a day. Managers tend to join every single meeting under the sun. That's how they fill up their time and appear(!) to be busy. In reality, they can cut down the meetings down to %25 and nothing would change.

Let's look at lowest level managers. They can meet 20 directs bi-weekly for 0.5hr. Let's add 0.5hr follow up work after each 1:1, which is way generous. So that totals up to 20 x (0.5 + 0.5) / 2 = 10hr. Let's assume their own 1:1 with their own manager to take 1hr bi-weekly or 0.5hr weekly as they need to discuss both team and personal issues. They can meet with peers and their manager weekly for 2hr. They can meet with sub-teams (let's assume max 4 teams) weekly (4 x 1hr) for status and problems. And at the end of the scrum (bi-weekly), they can meet all the directs for demos and postmortem for 2hr. If there are some other technical meetings they do not need to attend unless they are absolutely needed but this shouldn't be often. Leads are perfectly capable of handling those and report back if they're hesitant, which, again, shouldn't happen often. Meeting tally up should be around 18hr. For a 40hr week that's less than half. Let's assume they're dragged into some technical meetings for 5 more hours that's still 23hr. Remaining time should be more than enough to sifting through emails, quarterly duties (e.g promos, end of quarter status, next quarter strategy/goal setting) and occasional emergencies.

For mid-level managers (generally directors), they don't have to meet with teams but we assumed their 1:1 would cost 0.5hr/week. Their follow-up should be much less though as they should be able to make decision on most of the issues on the spot. But let's add a quite generous %50 on top which would sum up to 0.5hr x 20 x 1.5 = 15hr. I think they should meet with tech leads (skip-level) too if they're one removed from the lowest level. They can meet them for 0.5hr every 8 weeks which should take at the very maximum, 20 x 4 (leads) / 8 * 0.5hr = 5hr/wk but it'll probably take less. Tally up is 24hr (if you add peer & direct meetings). Mid-level managers may get more randomized but they need to be efficient at what to get involved and what not to. Majority of the middle-managers make the mistake here. They get involved more than they can chew as they're micromanagers and/or they respond to every single demand from their own manager, practically doing the manager's job for them. This is because their promo is generally tied to their manager's attitude toward them but this should be changed. How you can change this is a separate topic I can get into, if needed but let's skip it for now.

For high level managers (generally VPs), their time would be spent less on skip-1:1s & technical meetings but more on strategy and status meetings. I believe they would have even more time at their hand than directors if they're vigilant on the time sunk on the emails.

Then you have CEO at the top or you can have SVPs if number of your ICs is above 16K. But, yes 20 people management is possible if managers are cognizant of their time and boundaries of their responsibilities and if they actually work instead of appear to be working.

But let's face it, it ain't happening because managers are A) either people who doesn't want to work anymore but need the money so they'll just carry on until they f-up big time and then switch companies and rinse & repeat, B) or, they overstep their boundaries, C) or, they don't know how to do prioritization and time management, D) or a combination of some/all above.

[organsnyder]

> I see repeated 1:1 sessions as micromanagement, like being treated as a kid that needs their hand held.

There's space between viewing employees as children needing to be coddled on the one side and viewing them as completely rational automatons on the other.

I worked for three years for a "flat" company where we were all expected to independently define our own goals, schedules, and get support when needed. When everything was going smoothly, it worked well—I had a degree of freedom that was an incredible relief after my previous employer. But when I lost sight of how I could best benefit the company, I went into a spiral that led to months of me questioning my ability to work in this field at all (or whether I was a capable adult at all). Without a reporting chain—or official support structure of any kind—I didn't know where to turn; and knowing that it was my responsibility to correct this mess made the spiral go even deeper when I failed to do so.

At the time I was fired from that job, I was having issues with my blood pressure for the first time in my life. Within months afterward, my blood pressure returned to a normal level, and my wellbeing improved in so many other ways that can't be so readily measured.

I'm fortunate to now be in an environment that both has formal support structures as well as informal communication and feedback loops. I'm still the ultimate person responsible for my success, but my manager (and others in the company) are also invested in it. Perhaps I have less autonomy than I did (I still have quite a bit, especially compared to other places I've worked), but I find myself much more relaxed. Guardrails can be freeing.

[marcosdumay]

> Without a reporting chain—or official support structure of any kind—I didn't know where to turn

I really don't know where you are coming from. The OP stated a clear hierarchy where you'd have a single and well known manager that you can just go and talk to.

[bane]

Imagine every time a decision had to made where you were clearly the only person who needed to be met with and 20-50 people had to be plowed into a room to make that decision.

Instead of 2 person hours going into that effort, now we've just vaporized half a week to a week of person hours. A $300 decision just turned into at least a $3000 decision and the opportunity to handle 16x more point decisions just disappeared.

[marcosdumay]

Even the idea of routinely scheduled 1:1s i dreadful.

People that do that one certainly will also do routine scheduled meetings by project, and routine scheduled meetings by department. You'll quickly get more than 10 hours/week of just those useless meetings, and your manager will be completely unavailable for any real work at any time.

[mrguyorama]

22 is a reasonable class size for childhood schooling, and every single teacher ever has a clear and everpresent understanding of which kids "get it" or are gifted, and which kids are the problem. They are usually able to get face time with the kids that need it, and they do that to nearly entirely new kids every single year. Good teachers have a strong rapport with their students, and a tangible relationship with each and every one.

Why can't managers do what teachers do at a tenth the cost?

Like.... this isn't even the teaching part of teaching.

[tsimionescu]

22 is a dreadfully huge class size and only "works" to the extent that we ignore all of the children that fall through the cracks. Teachers have nowhere near enough time to have personal time with the problem children that need it.

Children either excel, tread water, or fall behind. Children who fell behind once will most likely drop out of the school system entirely, unless it was for some very specific temporary reason, or their parents can afford private tutoring.

A school system that really wanted to make every child excel would probably need class sizes of 4-8, but that is well outside the realm of possibility.

[eitally]

Two reasons:

1. Large class sizes are only effective if the instructional method is rote memorization. Otherwise, the teacher ends of having all the same kinds of issues corporate managers have.

2. Teachers literally go through years of formal education on pedagogy and effective leadership. Most corporate managers adhere to the Dilbert Principle, where they had been a competent IC whose role grew too big for one person, so they were allowed to hire a helper... before they know it, they're managing a team, until they become too incompetent at the managing part.

[matwood]

Have you managed people? Just dealing with interpersonal issues alone of 22 people would take all your time. You would have little time after to convey direction or even learn about direction from those above you. This is why someone would make 2-3 of those 22 people managers/leads...

[gadders]

You're only talking about functional management though.

You also have to discuss 22 sets of career plans, interpersonal conflicts, hiring etc and all the other aggravation that comes with dealing with people.

[master_crab]

I forgot where I read it, but a study years ago said something along the lines of 3-7 is the optimum number of direct reports someone can have.

[martin_a]

True!

In firefighting (in Germany) we usually stick to no more than 8 people under direct control. That's an amount of people you can take care of.

[alwaysbeconsing]

Mashup of Dunbar's Number with "The Magical Number 7 Plus or Minus One"

[tsimionescu]

Your model assumes the CEO and all others on the path are full time people managers. In reality, the upper layers of management have many other roles, from customer engagements to political lobbying to CSR to strategy etc. They can manage nowhere near 22 direct reports.

Additionally, geographic distribution is important. You can't have one manager in the USA managing 7 direct reports from Germany, 7 from India, and 8 from China. So, you need 1 site leader for each of these, that the 6-7 middle managers report to. These all add up very quickly to take the theoretical 2-3 minimum required layers to 6-8 layers in practice, without requiring any of those along the way to be micromanagers.

[bane]

Middle management time breaks down like this:

- 40-60% trying to figure out how to turn large strategic goals into projects digestible by your staff

- 20-30% coordinating up to get permission and sync with senior management, and then coordinate workshare with your peers

- 20-30% coordinating with your staff

- 75% of your time then gets sunk into squishy people problem stuff most people would call "HR problems"

"But this adds up to over 100%!" - yup

When an HR problem arises, at least 8 hours will immediately vaporize out of your schedule. It's not uncommon for bad situations to turn into multiple days to resolve.

On average, in my experience as a manager for the last 20 years, you will lose about 8 hours per month per staff member to "HR problems". It doesn't mean every staff member has these problems, but there will absolutely be those that have multiple repeated issues that take a long time to resolve. If your average work week has about 21 or 22 work days, you've just designed a full-time staff manager who does nothing else. Imagine performance review time in this structure where each review requires this staff manager to solicit performance input from each staff members task manager. I can tell you from personal experience it turns into a solid month of long nights and zero staff satisfied with their review.

This kind of construct does exist in many companies, but most companies have decided that it's more efficient to have the staff manager and the task manager be the same person, even if they have to swallow the fact that it makes organizations less flat as that manager simply doesn't have the time to deal with so many people as direct reports.

And voila, corporate hierarchy.

There are certainly managers who deal with having so many direct reports by simply not allocating time to staff management issues (i.e. HR problems) or not dealing with them in a timely fashion. The result is usually chaos, high-turnover, blah blah, but the actual end result is that staff become fungible units as the organization has to plan around this effective unreliability.

Staff also have fewer career growth options as the divide between worker bees and management transition becomes an insurmountable gulf. New management staff then tend to be brought in from the outside, and don't know the business or products. Enshittification occurs because the only comprehensible strategy anybody can cling to is revenue growth.

Anecdotally, the places I've worked that were flat all featured these characteristics while well balanced hierarchies were able to grow, build long-term strategies, and had high-retention rates.

For the record, I've had as high as 20 direct reports where I was both staff and task manager and it was hell on everybody. I've also had as high as 75 reports in a well structured organization (no more than 5-7 people per manager) and it was totally fine. Companies were software, R&D, or some mix.

Note: the average Googler stays at Google for 1.3 years. In my current company the average employee stays for just under 5.

[eitally]

I'm glad you called out Google in your post. I joined Google in 2015 having been a managing director of a global software engineering team at a non-tech F500, where I had teams in Brazil, India, China and Mexico, and folks scattered around the US & Scotland. I was accustomed for the prior ten years of my career, to run my org basically how you described, and I like to think I was a good manager that my team respected, and that we generally operated well. I spent a lot of my time 1) in leadership meetings trying to understand strategy and business requirements, and 2) structuring projects for my org and mentoring key individuals.

At Google, I found there to be 0 respect for management as a function. Every manager is expected to have their own personal projects and come performance review season you had better be able to point to something you did yourself, not just the outcomes you managed through your team. Perhaps this worked when it was a startup, but as an enterprise with roughly 350,000 workers, it's disrespectful to leaders to run things like this. I expect Bayer to experience similar disillusionment among their experienced management team. Overall they will probably see tactical acceleration in some areas but breakdowns in many others that will be more systemic and harder to recover from.

[Centigonal]

Golden post. I'm a manager at a company that was mostly flat for most of its existence, but is starting to add management as we grow. Middle managers are evil and bad until you find out why you need them. Much better to bite the bullet and design a system with clear roles and good incentives past a certain head count.

[adastra22]

Works for the Vatican just fine.

Your point is generally good, but not universally true. Mostly-flat organizations do work in certain contexts.

[defrost]

What works for the Vatican just fine?

It's a long way from being mostly flat, decisions are made at a crawling pace, and is convoluted enough to hide substantial theft, and shield large numbers of pedophiles .. even those rising to the Vatican C-suite level.

[adastra22]

The catholic church is famously held up as an example of a nearly-flat organization in many business schools.

[tsimionescu]

I think it's also possible that says a lot more about the shoddy metaphors of most business schools than the actual level of hierarchy in the Catholic church (beyond the other discussion we were having about how atypical the CC is as an organization).

[defrost]

Yet in reality it has as many layers as other governments and businesses of a similar size.

"Administration" in https://www.britannica.com/topic/Holy-See gives a rough overview.

There's layered mangement for multiple domains of interest.

[tsimionescu]

Well, a thousand years of refinement and a set of company values that most employees have instilled into them since birth will do that to an organization.

[adastra22]

That, and also most of the work of a church (interaction with laypeople) is self-evident, local, and easily managed. Once trained, a priest can be set loose on a community without much oversight.

Likewise, I can imagine a chain of Montessori schools being similarly organized, or a trade guild.

[cyanydeez]

If you.ignore the rampant pedophilia

[rjvs]

Does this structure explain the organisational response to the (myriad) accounts of terrible behaviour by individuals within it at many levels?

[fizx]

Hi, sometime middle-manager here. The goal isn't to have me make decisions, its to let people around me make decisions quickly and well enough that most of the time it works out, and when it doesn't, we can roll back easily.

If this isn't working out, I hire experts to improve the quality of decision-making, with me making decisions as a last resort.

[jncfhnb]

So… the goal is for you to do nothing? Except not impede people making decisions and hiring consultants when you, not the person closer to the issue, thinks it’s not working?

[kbenson]

As someone squarely in this position for the last few years (from a prior senior engineering position), I would say my job is to try to emulate the decisions and guidance my manager would give, but to more people than he has time to do it for.

This is fairly obvious because for a while he was doing that for too many people because we couldn't find a person for the position. If I can speed up answers and feedback and provide direction quickly in a way that's similar to what he would have do for people that often ended up waiting for him because his schedule was too busy, then I'm doing at least one part of my job well.

The more layers of this there are the more likely the message might be garbled by that game of telephone, but ultimately people can only handle so many relationships successfully, especially when they require specific regular action, and delegation (which is all middle management is essentially) is one approach that's been found to work around this.

When I'm doing my job well, the people I manage have clear directions and expectations, and if the company is managed well those expectations are seen as achievable by all involved, and if they aren't it's my job to communicate that either up or down the structure. I don't think it's sane to assume that will just naturally occur once a workforce gets big enough, so something is needed to help that along, and managers are one way to do so.

[toyg]

Basically you're describing a lieutenant/sergeant relationship. The sargeant will try to give each private the same commands that the lieutenant would give, with necessary adjustments for the immediate realities of the field.

[matwood]

Yep. It's frowned upon in some companies to use the military as an example, but military management has been forged often times in blood. The concept that you're hitting on is called 'commanders intent'. Basically we need to climb that hill over there, but the commander is not going to tell someone exactly how because they won't be in the situation. They trust the people on the ground will make good decisions.

I give my team all the information I have, convey my/the company's intent, and let them work. That way, I don't have to make every little decision. I trust they will make good decisions and come to me if there's a problem.

https://en.wikipedia.org/wiki/Intent_(military)#Commander's_...

Commander's intent (CSI) plays a central role in military decision making and planning. CSI acts as a basis for staffs and subordinates to develop their own plans and orders to transform thought into action, while maintaining the overall intention of their commander.

I'll also add that it builds ownership since the people doing the work are making the execution plans. IMO, it's the really the only way to manage if the goal is to get things done.

[jncfhnb]

I think this is not a good explanation. “Get on the hill” is a bad order. “We want to get on that hill to gain fire control on that road” is much better. Because that means the on the ground troops can achieve the objective through an entirely different means if the hill is not viable.

Critically, and this is why military analogies don’t work with most large orgs, the on the ground folks don’t need to confirm with top leadership that an alternative tactic is sufficient. In messy business orgs, middle managers play it way too conservatively and try to get their bosses to green light every decision by presenting them a synthesized and watered down view (“the hill is well defended, the other hill is better”). But this is slower, error prone, and at risk of compromise to minor communication failures because the condensed time and attention for the synthesized logic won’t capture the full reasoning (“I don’t get why the second hill is better. Just do the first one”).

[9659]

a time honored and tested management philosophy.

[toyg]

Indeed.

Unironically, a lot of modern war material like "Band of Brothers" is relevant to management and leadership in vertical organizations - which is what most companies are, at the end of the day.

[jncfhnb]

Ok but I think this is still just bad.

If person A manages 6 things and thinks it is too much, he should instead manage 3 things and have a new person B manage the other 3. Not elevate himself to manage person B and C who each have 3 things.

You say your job is to emulate what they would do, but also that there’s a game of telephone going on. That implies your boss is still very much in the loop of those actively communicating, and what’s probably happening is you’re bubbling up lower and lower fidelity information up the chain for them to make decisions.

You should never have the role of trying to emulate what your boss would do. If you’re going to manage managers, then those managers below you should have some sort of functional/technical expertise that warrants them being independent managers or they should have enough autonomy that your relationship is only about allocating higher level resources to them.

[kbenson]

If you have 100 people at the company you should have 19 managers with about 5 people each? And then the CEO has to manage those 19 managers, and has 19 people under them (because what is a CEO other than the person managing what's done at the top level?)? What about when the company has 1000 people?

It's all fine to you should just split it up more, but that only makes sense if you look at the part of the system (company) in isolation, which you can't do. It's intricately linked to the parts around it.

What you're recommending isn't even the waterbed theory of complexity, where you push down complexity in one area and it pops up in another, this is the ostrich theory of complexity, where if you stick your head in the sand and ignore everything then it doesn't matter.

[jncfhnb]

Org structure should not be a function of number of heads. If you follow some logic that each manager should have n reports you’ll make an idiotically tall structure. It’s basically the stupidest thing you can do. Which many organizations do. Most large orgs have a terrible design. It’s difficult.

Effective org design is about properly delegating authority. If you need to get your boss’ permission to do things, you’re friction, not lube. If your job is synthesizing others’ synthesis, so that a real decision maker can do things; you’re friction. If you’re a manager whose sole job is to try and emulate someone who is a better manager in a higher position, you’re friction.

It’s more useful to directly oversee a smaller number of important useful things that you can handle well than to hire two new people and manage them managing others.

Managers of managers are useful at the junctions where one manager doesn’t have the skill set to manage a necessary sub function at a tactical level. A CTO, CFO, CHRO, etc. at the top. Maybe a tech team lead or a design team lead or a financial team lead (varying by if you’re organized by product or function).

When a “manager” is not responsible for any tactical decision making, and instead just broad strategic direction, it’s much easier to govern a wide team of independently effective members. But if it grows too much? SPLIT the manager of manager duties with someone else. Don’t create another layer of manager of manager of managers who then by necessity also need to get split anyway.

> What you're recommending isn't even the waterbed theory of complexity, where you push down complexity in one area and it pops up in another, this is the ostrich theory of complexity, where if you stick your head in the sand and ignore everything then it doesn't matter.

Fuck off with this bullshit nothing assertion.

[pgwhalen]

A more charitable interpretation might be that their goal is to create space for good decisions to be made, and to de-risk those decisions.

Unless your mental model of a manager is "person who make decisions," than that is far from "doing nothing."

[jmopp]

It's kind of like how the north star for a programmer is to automate themselves out of the job. A good manager will bring the best out of their reports, which leads to making good decisions and executing on them.

[adastra22]

In the same sense that a machine maintenance person’s job is to not be involved in the machinery. Their job is not to operate the machine, but to make sure the machine operates correctly all on its own without intervention. Doing this correctly does take a lot of work.

[seb1204]

I like this, but I think also that a machine maintenance person has a very clear job description, defined set of tools and schedule to follow.

[xingped]

I mean you can say that, but that has never been my experience. Neither, I'm sure, has it been many other peoples' experiences either.

[sodapopcan]

Naw, it was my experience once. It’s rare, though.

[SoftTalker]

The layers above do likely understand the actual business goals better, and want to be sure that this thing you are asking for is actually going to support achieving those goals. Ideally that is the case, anyway.

[michaelt]

To an extent, sure.

If you've got the CEO, and you've got front-line workers doing the actual value-producing business, you can have two layers in between, sure. Maybe three layers.

But once you get to four or five layers? I hope your culture and productivity and ways of working were exactly where they need to be, because they're now impossible for anyone to consciously change.

[seb1204]

I work in an org with 5 layers of management. I do think CEO and top management have good strategies and big picture thinking etc. It is the bottom two layers where is see the biggest issues and challenges. We are not a software company but as an engineering/technology company most of our work happens in systems and software. Lowest level management needs to know, really know the software and concept to support workers. Else it is just smooth talking and buzz words.

[physicsguy]

It depends on the type of company I think.

There's a big difference between for e.g. Twitter where ultimately you have thousands of people working on a single product, and a conglomerate like Siemens or GE where you have very different, almost totally independent, business units doing totally different things.

[grecy]

After working at a very large Telco with 50,000 employees and management style straight from the 1950's, your comment hit the nail precisely on the head. I put faces to the positions in your comment, and it sent shivers down my spine.

Never again.

[mschuster91]

A way to sidestep this is by giving people financial authority. Say you're a team lead of half a dozen developers, so you get a blank approval for anything your team orders under 500$, 1 level of review for 500-2k$, and a third level (or HR) approval for hires and raises.

Of course, for that to work a company would have to place trust even into lowest levels of management, and that is what causes a lot of friction: people are feeling the disconnect between them being responsible on one side for the success of a contract that brings in hundreds of thousands of dollars in revenue, but apparently cannot be trusted to not waste 20$ on someone who managed to drown his computer keyboard in coffee by accident, no, that has to go through three levels of approval.