Ex-Valve employee describes internal politics at 'self-organizing' companies
pcgamer.com
pcgamer.com
It's one of those essays that is just plain important to read for life, because you will see the forces it describes everywhere around you, for the rest of your time on earth.
Thanks for this. One of the issues I found with the Agile implementation in my current company is that there's no longer an architectural/technical leadership within the team: the manager acts as scrum master, but it doesn't guide on how things should be done. As a result, as the quoted paragraph mentions, there's progress when the team is well suited for a given set of tasks. But most of the time progress is slow, because the team doesn't want to tackle the boring ones, or challenging tasks that would require the extra mile of effort.
That extends to the leadership of the company. It seems structureless, or "democratic" organizations can only work if the will/resolve/capacity of the lowest common denominator is good enough, and there's a large pool of commonly-minded individuals.
Once you go beyond a single Scrum team most organizations do need dedicated architects who provide guidance and maintain conceptual integrity across multiple teams. https://www.scaledagileframework.com/system-and-solution-arc...
It’s a chicken or the egg problem kind of problem. For all its talk about its love “freedom,” America is a deeply authoritarian and putative society.
The institutions that define much of our default “social scripts” (e.g. school, and later, work) are run like little dictatorships and their hierarchy and despotic bent come to define and permeate broader culture. Surrendering your “fundamental rights” (like freedom of speech and assembly), the moment you cross the threshold of your workplace leads to cognitive dissonance and a diminished belief in the truth of their “inalienable” status, both at work and in society at large.
The democratic participation of working class people in labor unions used to act as a check against this orientation, but as businesses interests have succeeded at undermining and destroying unions, businesses have filled the resulting power vacuum with even more authoritarianism, both in the workplace and in the outside politics they fund and support.
To change things, you’d have to do something like what Valve is attempting, but from the bottom up and at a more baseline level of culture and social arrangement.
Unions largely have done that to themselves, as they organized into the same structures they purport to combat and act accordingly, with a capitalistic bend.
And this narrative of course absolves business interests of things like how they used the red scare to attack union leadership, buying themselves judicial and favorable NLRB rulings, not to mention how curbed they are in their power relative to unions in Europe by the Taft-Hartley act:
https://jacobinmag.com/2017/12/taft-hartley-unions-right-to-...
Agile Coach here. Yes, this is toxic to quality and technical excellence.
The typical hybrid Scrum/corporate org chart I find puts product ownership (the “business”) in control and leaves each team’s tech lead (responsible for getting the team to the sprint goal) as the last line of defense of quality and technical excellence. What ends up happening, tech leads with no technical air cover give up craftsmanship in exchange for making the product owner happy and therefore keeping their job. A number of companies are going to wake up in a few years with an agile hangover of legacy garbage rivaling the mountain of shit they used agile to replace.
Should make a lot of consultants a lot of money.
When the Agile movement started out, it was very much about empowering teams to do good work. But the form that won, Scrum, was the one least disruptive to existing corporate hierarchies. (At least until somebody invented SAFe, a comically unagile Agile process.) I have very rarely seen "we're doing Agile" shops where technical concerns weren't entirely overruled by business concerns, when the goal of many early Agile proponents was to bring them back in balance. (And here I use "business concerns" advisedly, in that they're often much more about the political concerns of managers, not the actual needs of customers or the business.)
Your point about "legacy garbage rivaling the mountain of shit they used agile to replace" is spot on. If an org does a big rewrite without fixing the problems that got them the mess, they just get another mess. Switching from manager-dominated waterfall to manager-dominated mini-waterfall with some Agile labels slapped on doesn't fix the broken feedback loops and distorted values that created the legacy garbage. It would be as if George Washington had declared himself king after the Revolutionary War, or when a left-wing dictator overthrows a right-wing dictator. Deeper change is necessary.
Specifically, the section titled "The Nature of Elitism"; and the definition of "elite", elite groups, friendship and communication channels.
It's interesting to think about as it relates to one's own experiences with group dynamics in the workplace and maybe even more so - volunteer activities where explicit structures may not be defined.
I have always looked on and decided I would create a “flat structure” in my co. To avoid “politics”.
I have since changed My view. An insight is that there will always be a hierarchy, or more appropriately recursive levels of control and moniotoring needed. The question is do you acknowledge that or not, do you design for that or do you let it spring up adhoc?
If you take a look at any biological system that must exist, sustain itself, there is always a control and monitoring system.
This model exists for example in the body i.e our regulatory system.
Stafford beer has developed a model for this, the viable system model.
This may be a sweeping statement. We do know of collections of cells, for example, the cells in the heart muscles beat together without a central monitor. Swarms like flocks of birds, schools of fish etc. do not seem to have a hierarchical order.
Steven Strogatz has a book called "Sync" related to spontaneous synchronisation. He also has given a Ted talk:
https://www.ted.com/talks/steven_strogatz_on_sync/transcript...
If I recall correctly the birds are a network, where each one would monitor the ones closest to them and react based on that?
Which is monitoring and control. Just at a very micro level.
Also note the more appropriate term used above was “recursive” rather than hierarchy.
The behaviour of a flock of birds is referred to as a complex adaptive system, with emergent properties. It can be modelled quite accurately as a vsm.
Put simply one needs control, monitoring, feedback and information flow to exist. This is needed at every level (hence recursive).
Therefore rather than contradict my comment, your post is just a further expansion.
A complex adaptive system, can itself be modelled as a vsm.
This is how Craig Reynolds' Boids model works. [0]
However, I recall from word of mouth that if you factor in a reaction time that the model breaks down; though I can't see any references to this phenomena online.
Also, to state the obvious, nonhuman social structures aren't always applicable to human social structures. Very often, the have very limited applicability to human organizations, for the simple reason that they're biologically based.
Instead you should turn to history and anthropology, not biology for advice. Sadly, there are very few (if any) examples of long lived leaderless / flat human organizations. Inevitably, a small cadre form an elite core (either through influence (Politics! Gasp!) or force / intimidation (politics by other means! Double gasp!) You can see this with various anarchist organizations, intentional communities such as communes, protest movements, even the tragedy of the commons.
[0] http://www.sciencemag.org/news/2010/04/when-pigeons-flock-wh...
This is equally flawed since this is a field swamped with risks of survival bias in anything you study.
I think this point still stands though.
I wonder if a small number of our species will come to terms with the fact that we're basically a virus cursed with short term thinking and come up with some means of actual long term survival before we're wiped out. I'm optimistic.
I still can't do anything but laugh when someone tells me they're an anarchist. No you're not, what you mean is you don't like the current structure, whatever it happens to be. You have an idea of what you think it should be, even if you can't clearly express what that is exactly or how to get there.
It wants a society where power (social, political, economic, etc) is not distributed based on a hierarchy.
Some misinterpret this as meaning that anarchists oppose the structuring of power in general (a.k.a. chaos), but in fact power can be organized through all kinds of non-hierarchical forms.
Anarchism as a broad philosophy is not prescriptive as to how power /should/ be organized, only how it should not. There are many complementary and competing ideas about how a society or group can best structure itself to minimize hierarchical power, and anarchists argue about these a lot.
They also actively experiment with them, and often the experiments fail. But sometimes they succeed, and those discoveries help all of us expand our possible modes of social organization. So basically I think even if you don't subscribe to anarchist philosophy, you should not be so dismissive of it.
Anarchist Club
Meeting Tuesday
Elections To Be HeldA control/monitoring system isn't required to be centralised, or even different from what is controlled/monitored. E.g. with groups of fish every fish just keeps his distance to other fish, resulting in the well known global as-if-centrally-coordinated behaviour of fish groups.
So I think the statement just means that there is some kind of control loop. It doesn't say how it looks.
If you don't plan for it you're going to nurture paranoia in employees.
> If you take a look at any biological system [..]
No, no, no. You can justify any systems by trying to extract analogy from `nature`.
It's worth noting that Stafford Beer dedicated a lot of his life to fixing very broken hierarchies which at their core tended to two extremes: management managing everything, crushing front-line autonomy; and management having no interest in anything outside metrics.
Beer took a cybernetic approach to designing organizational structures which is not at all unlike the design of electronic regulatory systems.
There are several of his talks online which are an accessible but easily-misunderstood introduction to the material he covered more fully in his books.
What are the other approaches?
The way I now think about it is the for organizations, the purpose of the hierarchy is not to directly control work, but to create supportive contexts with enough formal structure so that emergent organization gets the work done flexibly and humanely.
This isn't an unknown model; it's basically how western democracies relate to families. Nobody tells families what to do. We mostly trust they work it out on their own. But when things go wrong, we have a variety of interventions (from public service announcements to police to disaster response) to get things back in the good zone. To varying degrees, it's also how states relate to cities, counties, companies, non-profits, and a bunch of other entities.
For work, I think of that unit as the team. If I'm putting together a software team, I will make sure the team has all the needed skills for the work (e.g., front end, back end, ops, design, user research, product management, data science). But I don't want to design a micromanagement structure for how work flows between roles; I want to help the team self-organize in a respectful and collaborative way. If they are happy and getting shit done, my role as manager should be supportive. It's only if they get in trouble that I'll need to shift to controlling interventions.
Exactly. Trying to build a company without structure is like trying to build software without structure. If it's non-trivial software, and you insist that there should be no organization, the project will fail 100% of the time. And so, unconsciously or consciously, you or others on the project will in fact begin to structure the code.
I think the intended lesson has always been not to have a fixed, rigid structure, but one that can adapt easily to pressures to better suit the challenges it faces at any given time.
thanks for the article link.
As ismail points out, people always conflate "flatness" with "structurelessness" or essentially some version of the wild west or a slumber party. This conflation is unfortunately too common and misses the point and reasons for more horizontal organization.
> Contrary to what we would like to believe, there is no such thing as a structureless group.
> This means that to strive for a structureless group is as useful, and as deceptive, as to aim at an "objective" news story, "value-free" social science, or a "free" economy. A "laissez faire" group is about as realistic as a "laissez faire" society; the idea becomes a smokescreen for the strong or the lucky to establish unquestioned hegemony over others.
Wow, what a money line. It's like a Rosetta Stone for a whole family of naive political ideas.
From ancap and the more extreme forms of libertarian thinking on the right, thru "flat organization" in the startup world, to explicitly leaderless leftist groups like Occupy Wall Street.
What they have in common is the fantasy that large human organizations can work without defined leadership.
I would rather say that it is extremely naive to believe that the systems we have inherited are optimal in any way.
What we (all four--me, the comment I'm replying to, the Tyranny of Structureless essay, and the article about Valve) -- what we're saying is that "no system" is not an improvement.
That's not to say any are perfect or fully optimized. There is room for innovation. As problems are solved, remining problems start getting more attention. And environments change over time. But established approaches are more optimal in at least some respects than failed experiments of the past.
New schemes are certainly a good idea if they actually improve on the current models and are based on reality and not fantasy or malfeasant popularism.
You just earned enemies and downvotes from the entire political spectrum. I'm onboard, let the martyrdom commence.
Likewise, a leaderless group can work very well for a while until someone joins the group who will put their own interests before that of the group. Then the group has much less capable defenses than a traditional hierarchy since by definition there is no leader to expel the new person. The chance a group meets such a person is clearly dependent on both the size of the group and the lifetime of the group.
In physical terms you might say that such groups are in an unstable equilibrium: any pertubation away from perfect harmony leads to a total collapse because there is no more benefit to cooperating if there are non-cooperators present.
I think the terminology we use may be letting us down.
Rather than talk about flat or horizontal organization, think it would be more useful to think in terms of encouraging freedom, autonomy and reducing bureaucracy while achieving {x} purpose.
How can one design for this?
To summarize, diffuse authority, have authority be ultimately responsible to the group (aka the employees as a whole), and avoid informal networks and concentration of resources by rotating responsibilities and equal access to equipment.
An employee-owned co-op with some or all of these systems in place could achieve the freedom and autonomy you seek. It has to be proved in practice, however. Many co-ops have more traditional authority structures still. Some bureaucracy is inevitable when designing a decision-making process for a large group.
yes, but these are measures designed to prevent an accumulation of power and information so that leaders remain interchangeable (like rotation); this does work against building individual competence. These are hard to maintain as a movement/organization gets bigger - for example the green party in Germany abandoned rotation and direct democracy at some stage.
I think that creating an organization that works well (any form of organization) is really a form of art - there seems to be some secret sauce not described in any books that makes it all tick. (for example even a dysfunctional system like this one somehow works for Valve, they seem to be doing fine from the business side of things)
If we were to make distributed systems more like humans, every 1000 systems would have a greedy psychopath amongst them. Unlike a chaos monkey type system, such bad actors could reprogram and take over the whole system. The chaos monkey is useful in software because they're ultimately controlled and can be shut down. Greedy psychopaths in meat space can pursue high value relationships and weave narratives to shut down systemic protections (I.e. WMDs, too big to fail). Why not perform some group screening to filter out excess greed, assuming it's a trait visible at the genetic level?
Disclaimer: these are my wandering thoughts likely more appropriate for sci fi that any society in my lifetime.
My takeaway. At organizations that pretend to be egalitarian, you're golden if you're in the in-group. But if you're outside of that group, you matter zilch.
Basic primate social behavior.
Exactly. As we all know, these people all much making much more than the upper management at Microsoft or Google, who haven't got any real personal benefit from their position.
There's this false dicotomy between "freedom for people to self-organize" and "formal structures to maintain a system", where people see what pointy-haired suits do and conclude that informal orgs will naturally be better. But _something_ needs to be in place to maintain a good environment.
You see this with people talking about Toyota. So many people see Toyota and think "oh I should do kanban". But kanban isn't the thing, the processes that led Toyota to do kanban is the thing.
Similarly, the objective isn't "have a flat org chart", but "maintain an environment where a formal power structure doesn't weigh down progress for the company". That requires acknowledging the informal power structures and making sure that the players involved are willing to maintain that environment _through_ this informal power structure.
Informal power dynamics do not need to be dysfunctional, if there's a principle driving the players
It was the result of a study by several car companies and several governments to learn what made the Toyota Production System different from the Mass Production systems used in the US.
- "We seem to have a problem with people seeking power and privilege for their own gain." - "To prevent it, we should let some people exercise power over others in high-status positions."
IE, Firefighters will stop the spread of fires by setting controlled fires.
see also: literally every company in existence.
I realize people tend to become invested in reinforcing the status quo but I am still baffled at how every time Steam comes up 99 out of a hundred comments act as if office politics aren't exactly as powerful and unaccountable as every anecdote about Steam and their cliques.
...Or that it worked just well enough while they were creating a platform that set them up as a gatekeeper and quasi-manopoly, and works just well enough to maintain it.
You report to your boss who reports to their boss and so forth. Authority is decided based on your title which is largely determined once at interview time. As is your compensation structure which almost never has bonuses being a significant percentage of the total package.
It is useful to be able to talk about those differences. What people have in mind here is that Valve is way more anxiety inducing then average company.
With a healthy corporate culture, being "in" means respecting your co-workers and meeting clearly-laid out expectations.
In a toxic one, it corresponds with social connection, cultural identity, obscuring expectations (and rewards) and zero-sum empire building.
* beginner or crappy
* moderate or senior-delusional
* seniors/experts
Most of the time the actual seniors are not the people trying to prove themselves or validate their existence. They honestly know how they perform. If the current job doesn't work out they know they can get another easily enough. If anything this camp may even try to hide some of their skills to avoid getting pegged into those crappy tasks you cannot seem to hire anybody else for.
The middle group is composed of people who are honest about their station in life. These honest people are happy where they are and aren't dicking around with office politics, but they do play the game just enough to avoid rocking the boat.
Then there are the dishonest people in the middle group. Sometimes this dishonesty is intentional because they are an imposter... to the point of fraud. Usually, the dishonesty is unintentional. These people are often disillusioned into thinking they are some sort of senior level rock star but curiously wonder why their management completely disagrees. These people are easy to spot because they talk a big game and their work is hilariously out of sync with the hot air.
Warning. Dishonest people are dangerous to your career and will run you over because they are too busy looking out for themselves. If you challenge their dishonesty they may put up a fight that pits you are as the evil aggressor. If you don't challenge the stupidity you can easily get stuck cleaning up their code failure. Pick your battles wisely.
Likewise there are honest beginners who understand their station in life and are eagerly learning to improve their skills. These are the people you dream of mentoring. Then there are also really crappy developers who can't figure out why nobody wants to mentor them or why they always get the worst assignments.
In a hierarchical organization I can usually hide from this stupidity well enough. I can only imagine the horrid levels of back-stabbing that occurs in a flat organization.
quality people: early, midterm, late
full shitbirds: early, midterm, late
In organizations of chaotic leaderlessness, to keep one's wits sharp, usually the bellwether of choice is nerd signaling.
This is where knowledge of relevant trivia is used as a guideline for experience level, trading in quips, name dropping and advertisements of awareness in exchange for actual measures of skill. Actual measures of skill can prove extremely expensive, and sometimes require live fire incidents before a skill will show itself.
This is where the dishonesty starts to brew. It's easy to hear experienced people trade quips and advertise their qualities, and then become a poser somewhere else, and imitate the signals in a totally different group, while lacking actual skill.
Sometimes, in a leaderless organization, that's enough, because democratically nerd posers still add to a voting block, and doofy imitators know they are imposters and, are usually incapable of doing anything but vote a straight ticket, so adding a nerd poser to your team still ensures strength in numbers, and potentially amps up the current of buzz and gossip information channels.
But to stay alive, you need more than party members, because ultimately it still boils down to whataboutism, in the form of what-have-you-done-for-me-lately? Party members are that fickle, and core experts are still required. So now the back-stabbing, and boys clubs form.
But who really matters? The truth is in a leaderless system, there are still leaders, but the convention is to avoid signaling. It's all expressed by waiting games. You have to sit around and wait for things to become evident. So lurkers are rewarded, over time, based on familiarity with a deeper awareness of organizational hierarchy. It's just a different system of control and signaling, that integrates literal time via in-group awareness of social histories, instead of titles awarded to people who just walked in off the street.
This sucks if you were cool somewhere else, because you'll be starting over with at least half of your significance and relevance rolled back to beginner status. It's a mechanism for only rewarding years of mediocre service (loyal nerd signaling party members), or clear-cut high performance (core experts that can prove their worth), or some mixture of the two. But the core tenets remain: proof of loyalty, or actual rock stardom.
The side effect here is flame-outs. People who have social currency or expect redemption of experience levels and recognition for past external work at a third party, or expect to level up on nerd signalling alone, find frustration at the pace of advancement, and if they can't find an opening to perform, wind up lumped in with the imposters.
Some of these are valid grievances, but if you want to make an omelette, something, something...
[1] https://en.wikipedia.org/wiki/Dunning%E2%80%93Kruger_effect
I want to say something like Aspergers* but without the emotional or decision impairment that otherwise comes with Autism. I am thinking a similarly lacking self-awareness with regard to the immediate social environment, but a deflated sense of self-worth opposed to inflated.
The result of such inverted bias is judging one's self in reflection to the performance of that person's peers only to be frustrated that certain performance or results are seemingly incomplete or of noticeably weaker quality. Since the biased person is judging themselves in reflection to the performance of their peers they will incorrectly view this incompetence as a reflection of their own performance or potential while the people who actually performed the supposedly incompetent work don't see the ascribed faults resulting in the biased individual over-compensating for faults that are generally not valued. This reminds me of OCD and perfectionism.
What you described is also covered by the Dunning-Kruger effect. From the Wikipedia page:
"the miscalibration of the incompetent stems from an error about the self, whereas the miscalibration of the highly competent stems from an error about others."
Seeing the effect in visual presentation (as a graph) might also aid you in understanding the above.
I got an autism diagnosis in 2017 (and my partner a week ago at age 35 and 36 respectively). According to DSM-5 (in effect since 2013 IIRC), asperger syndrome and all other previous specific autism disorders are superseded as it falls under the larger umbrella of ASD (autism spectrum disorder).
You described: "but without the emotional or decision impairment that otherwise comes with Autism." Someone who has all the traits of what was previously known as asperger syndrome sans what you described can still get an ASD diagnosis.
> Any sufficiently complicated company [without] management contains an ad hoc, informally-specified, bug-ridden, slow implementation of management.
What a flat structure entails is a star-like structure where everyone is just 1 degree of separation removed from everyone else. This high degree of coupling is bad in design and IRL because it creates high interdependency like the ex-valver mentioned. As others have mentioned, every action you do effects practically everyone else since people are so closely connected.
There's a reason why most structures are tree-like such that people are sufficiently insulated from others. Creating a comfortable degree of separation will allow better separation of concerns in the things you do.
Value created the 2nd greatest lie in tech about "not having a boss is cool" because it attracts people who think they're too smart to be working for someone else. But if you're working for no-one in the company, you're practically working for everyone.
I'll remember "if you're working for no-one in the company, you're practically working for everyone" as being in the vein of "if everyone is responsible for something, no one is responsible for that something."
Even in a company that small I’ll wager your boss knows and sets your salary and has the power to fire you, but not vice versa
Shallow orgs can move quicker, but that isn’t free. Because the designated leaders cannot actually manage up to dozens of direct reports, you end up with what I call a “circle” org structure. People get voted on/off the inner circle, and the downstream leaders get disempowered.
I’ve never worked in a place with no explicit command structure. Perhaps I lack imagination, but I cannot see that ever working. Fundamentally it’s a lie, because some individuals have to control the money.
In my experience, shallow orgs also have a half life. They need to be purged every 18-24 months.
Valve hadn't had a half life since 2007. They're surely due for another one :)
Where the hell is it Gabe?!?
That's not intrinsic to shallow organizations. One common failure mode when scaling shallow orgs is consensus-based decision making that can grind everything to a halt. As you say, a flat structure with a few designated leaders as bottlenecks for all decision making barely gets you off the starting blocks. Conversely, a deep org can move quickly as long as each level of the management chain is given appropriate scope for autonomous decision making.
> This is a haphazard process. The payment mechanism is to a very large extent bonus-based. So the contracts usually have a minimum pay segment in it, which is more or less established by tradition. And then the interesting part in this contract is how much is left to the peer review process, which is very complicated. It involves various layers of mutual assessment.
Yanis Veroufakis. Greece's former finance minister, known for his ability to describe Greece's financial collapse and negotiations with the EU in terms that the average Greek citizen could grasp-- called the Valve's bonus system "complicated."
This in a piece where his position as Valve's economist-in-residence ostensibly meant he was trying his best to paint Valve in a positive light.
Did anyone here decide to work for Valve after reading his interview at Gamasutra[1]? If so I would love to have a glimpse at the decision tree that led down that path. I can only imagine three nodes in that tree, one of which is, "Must eat."
[1] (linked from the article) https://www.gamasutra.com/view/news/187296/How_Valve_hires_h...
Here the same person was much more positive about the system (easy to understand he is not, at least not to me), wonder were he was telling the truth.
http://blogs.valvesoftware.com/abrash/valve-how-i-got-here-w... Also Michael Abrash thinks it is a swell place to work for - I guess if you have something to say within such a company then the experience is way better.
I don't think I understand what you mean.
Here's the relevant passage from your link wrt bonuses:
> [Before writing more in this, especially regarding Valve, I shall need to become better acquainted with the peer-review based process of determining bonuses. Watch this space!]
That's from August, 2012 on an internal blog. The external interview he gave where he calls the review process "very complicated" is from February, 2013.
So let's recap:
* A reasonable person joining Valve after reading the Aug. 2012 blog would be worried because their "economist-in-residence" is publicly reserving judgment on the peer-reviewed bonus system
* A reasonable person joining Valve after reading the Feb. 2013 interview would be worried because their "economist-in-residence" is characterizing the peer-reviewed bonus system as "very complicated" to a reporter
* A reasonable person joining Valve after reading the Aug. 2012 blog and the Feb. 2013 interview would be very worried because their "economist-in-residence" moved from reserving judgment on an internal blog to criticizing the bonus system in a public interview
At the very least, I see no discrepancies in what Veroufakis said in Aug and Feb that would lead you to conclude he wasn't telling the truth in one of those links.
What am I missing?
to quote:
"The tantalising thought arose, during my musings, that this organisational structure may be as scalable as a market mechanism (assuming that the right technologies are in hand, ensuring transparency and low communications’ costs within the company).
There is one important aspect of Valve that I did not focus on: the link between its horizontal management structure and its ‘vertical’ ownership structure. Valve is a private company owned mostly by few individuals. In that sense, it is an enlightened oligarchy: an oligarchy in that it is owned by a few and enlightened in that those few are not using their property rights to boss people around ... "
blah blah blah.
If there is still a games division in Valve it is exceedingly dysfunctional.
Saliens
Trying to avoid press, only to have said press knocking on his door and asking for an interview. Classy.
That being said, it doesn't surprise me a bit. I worked on a startup for a year or so that touted a flat structure. I was lucky to be in the in-group and in the end, our decisions (A couple other senior engineers + me, plus the founders themselves) were the only ones that mattered. Not gonna lie, it was a rather enlightening and profitable period in my life, and when I parted ways I made sure I did so in good terms, but it still leaves me with a slight bad taste in my mouth. Had I not been a childhood friend of one of the co-founders, I'd probably have gotten treated like nothing, or worse yet, not hired at all.
Later on I found out the company had gone bankrupt and the founders had parted ways, probably fishing for that buyout or that VC money they never got. Every once in a while I correspond with the friend that let me in, but we try not to talk about our time in the company.
Clearly a military like hierarchy doesn't work well in systems development. Most aspects of systems development are fundamentally creative processes that are somewhat diametrically opposed to being micromanaged. At least this is what I have found in my experience.
As jofreeman.com notes there are fundamental constructs of organizational structure that are critical. Ie: one of the most important (IMHO) is the top down delegation of duties. A good leader ensures critical things like cross training and staff development are achieved.
Bottom line: every organization has the same human dynamic - no matter how it's organized - if you get face time with the leaders, you wield power through association.
The military understands quite well that most tasks cannot be planned or managed at all. It is never a matter of giving people scripts and having them execute scripts. The world is chaotic and unpredictable. It's also very hard for any one person to get anything done. Most work requires a team which introduces even more chaos, failure modes and unknown variables. Therefore the military introduces the concept of 'command'. Commanders are not there to micro-manage anybody. Good commanders understand quite quickly that they cannot control everything. Commanders are there to work with the chaos.
Once you accept the concept of command the next step is the basic understanding that there are some people who are good commanders and others who are not just bad at it but they may actively dislike it. The commanders may indeed thrive off a certain amount of chaos and structurelessness may despise being "micro-managed." Give them a big goal, the necessary resources, and set them off. As long as feedback mechanisms are in place they'll get there. The specialists want routine and stability. They want to be part of a team where roles and responsibilities are clearly laid out. They want to be able to focus and specialize on a single problem and excel in a particular domain.
BTW, this division, like most binaries, is not written in stone. Excellent results can be achieved by rotating commands over time. Specialists who claim to dislike command should sometimes be given command responsibilities simply to edge them out of their comfort zone and broaden their perspective. Sometimes commanders need to get to their hands dirty to grain a deeper tactical perspective. Avoiding a rigid hierarchy and command structure is how good organizations scale and adapt.
The problem with the completely flat model is that it may serve commanders very well (who again, can thrive off chaos) but specialists will experience a lot anxiety and frustration.
I'm not sure this is accurate. The Space Shuttle systems software, likely among the highest quality software systems ever developed, does not appear to have been a creative process--at least not in the artistic sense of creative self-expression (thanks to a previous commenter for the link: https://www.fastcompany.com/28121/they-write-right-stuff). If the Shuttle example were representative of software development as a whole then we might posit that software quality is inversely proportional to creativity. But I don't think it is.
Instead I would describe the creativity required as: improvisational problem-solving within the bounds of a known form. In warfare, for example, there are standard tactics (the form) for reacting to an ambush, however, every ambush is unique and the local commander must improvise within those tactics to solve the unique problems. Likewise in classical music, the symphony had a standard form in which a composer like Mozart could express creativity. Software development should work the same way.
The problem is that we don't have a sufficient catalog of proven software designs (forms) to refer to when solving common problems. We end up solving the same problems over and over again in unique ways. This allows developers a lot of creativity but does it result in high quality software at a reasonable cost? I don't think it does. I lay a large part of the blame for this state of affairs on the culture of neophilia so prevalent among developers. In the culture of other professionals who make things, they seem to cultivate a tradition of what works, why don't software developers?
Triggered. Are they trying to create an office politics cluster-fuck on purpose?
"various layers of mutual assessment" sounds like the ultimate tit-for-tat game where compensation has no longer any relation to the real-world and is strictly determined by internal politics, especially in the face of a limited total bonus fund.
This self-organizing company fad rears it's ugly head under different names every decade or so, and it usually ends in utter disaster.
A hierarchicy will always form, and if you don’t deliberately form it, the result will be, as described, suboptimal.
Ah... The Lord of the Flies management structure!
For instance, in any organization structure a guy like Steve Ballmer is going to bully his way into a leadership role.
...is not what I would have imagined popping up in this comments section.
Sounds good on paper, and execs love to brag about how this brings people together to recognize excellence and talent and blah blah blah, but what really happens is that people start to condition their work to your likeness to recognize them. If you are known to send bonuses often, your requests are magically fulfilled. If you are not into it, prepare to have support tickets forgotten for months.
In the end, you can't expect much from increasingly large groups of people. Eventually interests will diverge and self interest will prevail, politics will emerge and abuses will happen. Some kind of governance is required to counterbalance that, even if it usually fails catastrophically.
I've seen various people organize their own hierarchy where certain people take control of the ability to push the code or access to the founder. This control is premised upon getting everyone to maintain the party line, and shunning anyone who deviates from the party line.
Self-organizing structures are not necessarily the most effective structures because they are given form by the information technology that the agents are using and are shaped by the constraints placed by the environment.
When there is no mold (the information technology is too shapeless - human speech or something like forum software) then the whole thing will essentially implode once the information grows out of bounds for every individual to keep up with all of it (delegation and separation of concerns is not explicit so it doesn't happen so readily).
Coming up with a sensible communication system that gives shape to organization without restricting the freedom of the agents is an interesting 21st century problem that should get more attention.
When people say that no structure works with few agents it has to do with the fact that humans, when in small groups (tribes), establish informal hierarchies that are very fluid and natural. However at larger scales the group needs to be fragmented into several factions which are also ranked, sometimes explicitly, it becomes important for individuals to signal a visual indication of rank or somehow show which group they belong to (this can be seen throughout history, I recommend a book called "The Dominant Man: The Pecking order in Human Society"). I think the trick is to try and create an "unstructured" network out of units that are small enough that they don't reach the threshold of needing formal structure, these cells can be informal and comfortable (tribes) but compose to create the organisation.
Effectively this means thinking of 'two pizza teams' as individuals when organisations are at scale and abstract out the individual as just a member of his team. Giving the team complete autonomy over how it handles its internal affairs and only judging it based on the work it produces i.e. as an "individual".
I've run into this problem a couple of times. Ultimately it means you can't really discuss problems and you can't place responsibility.
Is there any organization where this is not true? Every single organization has stories of talented people who lost out because they could not convince the right people of their value to the organization.
It sounds like Valve may need some work. It also sounds like the author may need some work. There are two sides to every story, and we're getting the tweetstorm from someone who has built up years of frustration and is now letting it out.
Sitting in a place where you aren't happy, or where you are envious of the positions of others, isn't a good place to be either personally or for your company. All companies are flawed in some way - you need to find a place where you are accepting of the flaws and can be productive and satisfied with your work anyway. If you aren't in such a place, take action - either invoke change to make it better, or leave.
This. I'm a compsci master-student and currently (since 2017/03) work at a big german tax-software company.
This company is not a self-organizing firm, but has a clear hierarchy (from small groups of 5 ppl to several divisions with up to ~300 ppl. Nevertheless, they tried to adapt some aspects of thr american work big players - such as the open office concept.
The situation there is even worse, as students/interns have to dynamically choose a random workplace every day.
In the open office -filled with ~100 ppl in total- I felt surveillanced and could not concentrate on my current projects as every now and then someone passed my desk and glanced into my code etc.
Lucky for me: I do have a pretty exotic status&job at this company and was able to get a company laptop that enabled me to wander from place to place within the company until I found a pretty isolated room for ~20 ppl (full with software testers). I managed to get access rights for this room and am officially allowed to work there.
Although the ambient noise is worse than in open workplace (those software testers are chatty as shit :-) ...), I'm generally more concentrated and prefer wearing hearing protection now and then, instead of working in an panopticon.
Conclusion: after my graduation I'm looking for a small/middle sized company and consider buzzwords like 'flat hierarchy', open office, free fruit, etc. as red flags.
I think it would be an interesting thing to analyze, I wonder if anyone at companies like this track this stuff, looking at what structures get created, key people, how things shift, and who lets go,etc.
I felt the same, until I read “Requisite Organization” by Elliot Jacques. It changed how I run my company. Jacques explains very clearly why hierarchy exists and is needed in any organization of people. It’s instinctual.
If you're a leader who doesn't want to provide leadership, and is fortunate enough to own a near-monopoly service, a self organising team takes a lot of work off your shoulders.
What made Valve go for the 'self organising/flat' option in the first place?
Was this in reaction to the founders experiences of working in technical roles at 'hierarchy/pyramid' style companies?
I think that society currently is quite fearful of the technically able folk who can put people out of work with algorithms and robots. These technically able people can also put processes in place and bring in efficiencies that lead to a bigger business and the need for more people. If you sell twice the amount of stuff you might need twice the amount of people doing customer service, quality control, production, warranty and other things for the new business brought in. However these people need to be thinking people, not the people that did stuff by rote and were automated out of existence by algorithms.
Despite there being a veritable internet based industrial revolution underway society wants to promote people with accounting skills or a law degree to the top of the org chart lest the technically able class end up running the show. In fact techies are only allowed to go into management if they abandon actual tech and keep their hand out of it.
To be honest a lot of office departments could be obliterated with a few well designed website forms and a few API calls. There isn't any need for A.I. or self driving cars for this level of revolution to happen, a lot of people really could be replaced by a few 'if statements'. Then there are the pointless jobs that congeal into existence in any medium sized company that are absolutely not needed.
Yet the people that put the company business logic into code know how the whole thing works. They also have the data. If your tool of business is the phone and your idea of work is shuffling from one meeting to the next then it must be hard to accept that customers prefer the website and even harder to accept that you should be way down the org chart compared to the people that can code. It might not have been cool twenty plus years ago to learn tech so those that didn't can't really accept that they missed the boat on the tech revolution.
I think that this fear of the unknown, this idea that code is voodoo, this insistence on keeping programmers in the basement is the bigger problem than 'flat/hierarchy'. We have ended up with a management class that are about as useful as our politicians and there is real danger that in the Anglo-West we will lose to China bigly like how the Germans and Japanese reamed us out so badly with engineering during the last century.
Good engineers, who love coding, can design a big system, are unlikely to rule over the Design. It'll be bad engineers, now Managers, with good political skills, that got promoted because its their only way to survive, being otherwise useless.
It is totally upside-down, hence the mess, unless you have Bill Gates at the top, true coder.
Is there any mirror of the article without the advertizement nd tracking trap ?