Managing without authority:

Notes on the romance and reality of “boss-less” firms

Phanish Puranam1

INSEAD

This draft March 3, 2014

1 I thank Krithiga Sankaran for valuable research assistance, INSEAD R&D committee for
funding the “Organizational Zoo” project, and participants of the Dartmouth Roundtable on
Collaboration 2013 for a stimulating discussion on these issues.

Electronic copy available at: http://ssrn.com/abstract=2495910

Corporate leaders around the world today are likely to violently reject any suggestions

that their organizations are “authoritarian” or “hierarchical”, even though their

organizations very likely are authority hierarchies. More intriguing however is the

fact that a prominent handful of firms appear to operate today without any formally

designated bosses at all. These are of particular interest, as they appear to eschew the

use of formal authority hierarchies in their organization- a puzzle from the perspective

of those who see the defining feature of a firm as its reliance on authoritative

coordination of activity (Coase, 1937; Barnard, 1938; Simon, 1947; Williamson,

1975). In this essay I examine what these instances tell us about the continued

viability of authority hierarchies, the currently dominant model of organizing.

Authority hierarchies- and what we don’t like about them

Organizations are multi-agent systems with goals (March and Simon, 1958). For an

organization to exist, it is necessary and sufficient that solutions exist to four

fundamental problems of organizing – task division, task allocation, reward

distribution and information provision (Puranam, Alexy and Reitzig, 2013). The

business firm represents a form of organizing in which formal authority – the

legitimate power arising from the employment contract, of a superior to dictate the

actions and overrule the decisions of a subordinate within a particular domain - is a

basis for selecting and implementing solutions to these problems (Barnard, 1938;

Simon, 1947; Williamson, 1975). Thus in an organization that relies on formal

authority, which tasks are necessary to achieve the goals of the organization, who

does those tasks, how they are rewarded for doing it, and how they access the

information needed to do it is determined by individuals who have been explicitly

granted the power to make their decisions on these matters binding on others. When

Electronic copy available at: http://ssrn.com/abstract=2495910

exceptions and disputes arise due to the imperfection of these solutions, formal

authority is also used to resolve them by proposing new solutions.

Because authority is an ordering, even a dyad consisting of one superior and one

subordinate represents a hierarchy. We may be more interested though in

understanding why authority hierarchies in organizations tend to be branched (i.e.

typically each superior has multiple subordinates) and multi-layered.2 There are two

known mechanisms that generate these properties. The first involves goal

decomposition. If the tasks that contribute to an organizations goals can be treated as

sub-goals that can be decomposed further, then effectively a multi-layered, branching

task hierarchy exists. The assignation to specialists of responsibility for sub-goals as

well as authority over those they in turn assign sub-sub-goals to, can then produce a

branching, multi-layered authority hierarchy. The second mechanism involves limits

on branching in the authority hierarchy- the maximum span of control. If there are

limits to the number of subordinates over whom a supervisor can effectively exercise

authority, then introducing an intermediate layer of individuals with authority over

those below them can help overcome this limit, and also produced a branching, multi-

layered authority hierarchy. Through the authority hierarchy, a single individual can

control large numbers of people, albeit imperfectly. Either mechanism can exist

independent of the other, or in conjunction.

In authority hierarchies, it is necessarily true that only individuals at the top of

the hierarchy will have no superiors. For all others, authority over one’s subordinates

is limited by and must lie within the scope of one’s superior’s authority over oneself.

In this sense, authority everywhere except at the top of a hierarchy is delegated: if A

2 A third property, containment or “nesting” (i.e. different teams may report to the same
department head, different departments may report to the same division head etc.) is
common but not universal; matrix organizations represent overlapping hierarchies which are
therefore not nested.

Electronic copy available at: http://ssrn.com/abstract=2495910

has authority over B, then A can overturn a decision by B, and the lower the

likelihood of A exercising this right, the greater is the extent of delegation in this

dyadic relationship.

Several attributes and consequences of authority hierarchies have come in for censure.

First, there is an irrefutable concentration of power (even if it is legal) at the apex of

the hierarchy, which the denizens of an egalitarian ethos naturally dislike. As

egalitarian values become more widespread and entrenched, so does the dislike for

authority hierarchies. Another implication of the concentration of power at the apex is

that decisions about the goals of the organization are likely to be taken there. Yet the

knowledge needed to make these decisions wisely may not filter up effectively

through a hierarchy. Worse, poorly made decisions at the top of the hierarchy are

more immune to questioning than those made at the bottom, even though the former

probably impact the organization more significantly.

Second, because delegation necessarily involves some loss of control, those at the top

of formal authority hierarchies also tend to use standardization of work and

procedures as an additional means of control. To those at the receiving end of

standardization, the local costs of forcing “one size fits all solutions” tend to be more

visible than the global benefits of coordination, quite independent of their relative

magnitudes; accordingly ‘red tape’ and “bureaucracy” are two commonly cited
3
negative attributes of authority hierarchies. Relatedly, the results of purely

3
A related problem with authority hierarchies may be the systematic tendency to under-

delegate; see the “authority game” of Ernst Fehr and colleagues

managerial work- such as the role of middle managers in an authority hierarchy who

exist purely as a reflection of span of control constraints (and not of the task hierarchy)

are less easy to see, compared to the tangible outputs of employees engaged in

production activity. Thus, independent of their relative value, the ease of perceiving

their value is not identical, often leading to angst about the “bloat” introduced by

middle management layers.

Third, when authority hierarchies arise as a consequence of goal decomposition, it is

inevitable that sub-goals are defined more and more narrowly as one travels down the

hierarchy, producing uber-specialization (and its discontents). To the extent that

variety in work is valued, extensive authority hierarchies may give rise to

dissatisfaction on this account particularly towards the bottom of the hierarchy.

Fourth, while the existence of “silos” – barriers to free communication- is not

stipulated by an authority hierarchy, their emergence is a natural consequence of

greater emphasis on communicating among those who are jointly subject to the same

source of authority (and are possibly working towards the same sub-goal). Yet in a

world with imperfectly decomposable task structures (Simon, 1965), silos prevent

attention to and discovery of important interdependencies that cut across groupings.

Fifth, the increased importance of knowledge based work – in which employees are

hired primarily for their intellectual capital rather than their physical capacity for

effort, raises doubts about some of the traditional arguments about the benefits of

authority. These have included a) the ability to motivate, monitor and reward efforts,

b) the benefits of exploiting superior knowledge without engaging in knowledge

transfer to another by making the knowledgeable individual the boss (Demsetz, 1982),

and c) the function of authority as a means for resolving disputes when bargaining

breaks down (Mintzberg, 1979; Tushman & Nadler, 1978). With knowledge based

work, the ability of a boss to monitor, or indeed to systematically have a knowledge

advantage over the employees seems less credible (though the dispute resolution

function may still remain) 4.

So what alternatives do firms have to the formal authority hierarchy? Alternatives to

formal authority as a basis for organizing in general are of course quite well known.5

These include informal authority based on expertise, reputation and status (e.g. the

role of Linus Torvalds in the Linux community) as well as bargaining power (e.g.

Toyota in its supplier network). These can result in fairly hierarchical organizations

even without any formal authority at work (for instance see Gulati, Puranam and

Tushman, 2012 on stratification in meta-organizations). Perhaps the more interesting

question, and the one that this essay focuses on, is what alternatives to formal

authority exist within the legal framework of the firm, and how they co-exist with the

formal authority that is automatically created through the contractual structure linking

a firm to its employees.

How “boss-less firms” work: An example

4 There are degrees to which authority may be exercised even in its dispute resolution function- with
the right to only resolve disputes when they arise vs. the right to make the final decision being the two
extremes, with the right to veto subordinates decisions being somewhere in between.
5
We could treat a market as a special kind of organization, whose goal is to maximize the utilities of
buyers and sellers using prices as the key mechanism through which task, division, task allocation,
reard and information distribution occur; the conditions under which markets can achieve this goal are
very well known. Evolution through natural selection represents another approach to finding solutions
to the basic problems of organizing; in the ant colony, natural selection writes these solutions into
genetic material, which produces inter-locking patterns of behaviour.
 Consider Valve, a Seattle based software company that is behind such highly

successful video games such as Half-Life and Counter Strike, the world’s largest

online gaming portal Steam, and the widely used game programming environment,

Source through which it allows users to modify, or “mod” its games. In 2013, it was

privately held, and was estimated to be worth upwards of 2 billion USD; its estimated

revenue per employee was higher than that of Google, Amazon or Microsoft. Valve

was founded in 1996 by two ex-Microsoft employees and had grown to about 400
6
employees in 2013.

Valve quite self-consciously describes itself as non-hierarchical. As one employee

noted in a blog:

“If most of the value is now in the initial creative act, there’s little benefit to
traditional hierarchical organization that’s designed to deliver the same thing
over and over, making only incremental changes over time. What matters is
being first and bootstrapping your product into a positive feedback spiral with
a constant stream of creative innovation. Hierarchical management doesn’t
help with that, because it bottlenecks innovation through the people at the top
of the hierarchy, and there’s no reason to expect that those people would be
particularly creative about coming up with new products that are dramatically
different from existing ones – quite the opposite, in fact. So Valve was
designed as a company that would attract the sort of people capable of taking
the initial creative step, leave them free to do creative work, and make them
want to stay. Consequently, Valve has no formal management or hierarchy at
all.”

There are no employees called “bosses” in Valve. Every employee can initiate

projects, and choose which projects to work on. Employees are free to choose how to

use their time and talents. There is no manager or system architect to oversee or

control these choices. Instead employees are encouraged to work on “what interests

them and what brings value to Valve”. The official employee handbook is subtitled:

6 The description of Valve is based on public sources only.

“A fearless adventure in knowing what to do when no one’s there telling you what to

do.” Self-selected teams of individuals form spontaneously around topics of interest.

As a consequence team size and composition is constantly in flux- reflecting which

projects are believed to be “hot”. At Valve employees are highly empowered to the

extent that they can ship their own products (provided 2 more employees agree to

such decisions). There is no separate marketing or quality assurance department in

Valve.

The company does not have any formal top down and lateral communication channels.

It is upto the individual employees to talk to others in the company to find out what is

happening; thus most communication within the company is largely informal. To

coordinate with each other, employees simply move their wheeled workstations to be

physically proximate to team members; in this sense Valve employees “vote with

their wheels” on which projects are attractive to them. Code on any project can be

accessed by any employee within Valve.

Employee’s performance is assessed by means of a peer reviewed performance

system where peers review each other performance and rank them. Top performers

receive generous bonuses and raises. But since employees are rewarded for their

contributions, there is an incentive for employees to be part of successful projects

which makes such project teams increasingly popular and well-staffed at Valve.

Projects perceived as risky may not be able to attract talent and thus may not be

adequately staffed. Pay is very high by industry standards, and the perquisites include

a gym, massage rooms. The company also sponsors annual paid vacation for all its

employees and their families.

Dispute resolution at Valve is largely through consensus. An employee stated on a

blog:

We’re all human, so teams sometimes argue (and sometimes passionately)

about what to do and how to do it, but people are respectful of each other, and
eventually get to a consensus that works. There are stresses and more rigid
processes when products are close to shipping, especially when there are hard
deadlines for console certification (although shipping for the PC is much more
flexible, thanks to Steam). Sometimes people or teams wander down paths that
are clearly not working, and then it’s up to their peers to point that out and
get them back on track.

This process seems well described by the term “lateral escalation”. Decision making

is initially attempted within the team, then the next circle of peers and so radiates

outwards till the dispute is resolved. The possibly adverse consequences to reputation

(and therefore to the likelihood of participating in future projects) may thus act as a

source of pressure against adopting unreasonable positions, and may thus serve as a

dispute resolution mechanism.

Projects do have leads chosen by informal consensus, but there is no prestige

or money attached to the label, and it’s only temporary – a lead is likely to be
an individual contributor on their next project. Leads have no authority other
than that everyone agrees it will help the project to have them doing
coordination. Each project decides for itself about testing, check-in rules, how
often to meet (not very), and what the goal is and when and how to get there.
And each project is different.”

Technically, Valve is not hierarchy-free because the founder owner, Gabe Newell

clearly has formal authority over his employees; he can fire his employees, not the

other way round. The employee handbook appears to acknowledge this, tongue in

cheek, when it defines the founder as follows: “Gabe Newell—Of all the people at

this company who aren’t your boss, Gabe is the MOST not your boss, if you get what

we’re saying.” Newell himself allegedly offered a less subtle statement about where

all authority ultimately lies in this organization, after a recent round of layoffs.
Another domain in which the founders exercise considerable authority is in hiring-

extreme care with multiple rounds of interviewing and due diligence precedes any

hire, and the founders make the final decision. However Valve does have a very

shallow formal authority hierarchy, and Newell also seems to have delegated a lot of

his authority to enable employees to make their own decision on how to organize.

There are several instances of non-hierarchical organizations like Valve today.

Perhaps the oldest is W.L. Gore, the maker of Gore-Tex fabric. Morning Star makes

tomato paste using fairly low skilled workers. Both eschew a formal managerial

hierarchy. But it is in the world of software start-ups, inspired as they are by Valve

and the open-source ethos, that there appears to be a lot momentum building towards

non-hierarchical organizational models. GitHub, which offers a sophisticated system

for managing distributed software development, claims its objective is to “maximize

happiness” and was valued at $750 million in its Series A valuation. Menlo

Innovations, a software development and consulting business has its developers work

in pairs two to a keyboard, writing code together and switching partners every

fortnight or so, and says it is in the “business of joy”.

It appears that the organizing principle that has supplanted formal hierarchy in these

firms is that of self-organized teams. By “self-organized”, I mean teams that are

formed through individuals selecting what to work on, with whom and how. The task

division and allocation at the level of the firm thus arises spontaneously as each

employee chooses what to work on; within a project it may be negotiated among

project members. The division of labour through self-selection may be the distinctive

feature here compared to self –managed teams- and closely resembles division of
labour in open-source projects. More like conventional self-managed teams, self-

organized teams may also be self-managed in the sense that reward distribution,

information provision and exception management within teams occur without

recourse to a formal authority structure outside or inside the teams. Like self-managed

teams, eschewing formal authority does not preclude a role for informal authority.

While there may be no bosses, leaders, it appear, often “emerge” in these teams - who

can “gain the respect of their peers and can attract followers”, in the words of the

CEO of W.L. Gore. However, self-selection may also impact how such teams are

self-governed; self-organized teams may be different from traditional self-managed

teams because they are formed through self-selection which may qualitatively alter

the way in which integration of effort works- by making self-governance easier.

The virtues and vices of division of labour through self-selection

Self-organized teams may enjoy certain advantages over systems of organizing that

involve formal authority in decisions about task division or task allocation (as self-

managed teams often do). These are in addition to the benefits of escaping some of

the negative aspects of formal authority hierarchies.

We can make some conjectures in this regard: For instance, decisions about task

division may be made by those more knowledgeable about making them, and may

result in a better match between tasks and individual skills/ interests, leading to

improved productivity. This in turn may confer advantages in the integration of effort.

Because of intrinsic motivation from work the problem of reward distribution may be

easier to solve. Cooperation and coordination may be more likely because (common

knowledge of) self-selection suggests (common knowledge of) a higher level of

similarity of interests and knowledge so that disputes may be less likely to arise in the

first place.

We can also conjecture about the possible vices of organizing using self-organized

teams. Again, I mean problems beyond those that would arise for any system of

organizing that rejects formal authority hierarchies, such as the missing ladders of

status, opportunity and mentoring, and interfacing with a hierarchy infested labour

market.

First, there is the problem of possible under and over provision of effort (not within

the teams but at the level of the firm). Since rewards are peer driven, individuals may

want to work on visible rather than truly promising projects; this has an increasing

return dynamic that may crowd out truly useful projects.

Second, there is a question about scalability. It is worth noting that all start-ups look

like self-selected teams, but switch to conventional authority hierarchies as they grow.

Does that mean there are natural limits on scale for this kind of organization? Gore

splits off divisions when they reach 300 employees; Valve has run into some troubles

and fired people at around this number. Relatedly, most of the examples considered

here pertain to privately held young companies - could this work in a PLC? In a

brownfield context? Consider this blog post from another Valve employee

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. The question arises: what happens to the alternative spontaneous
order within Valve if some or all of the owners decide to sell up? Granted that
Valve’s owners do not intend to do this, the question remains, at least at the
theoretical level.
Third, while leaders in these systems are those who acquire popularity- this may lead

to “in-crowd” phenomena and give rise to cliques. Relatedly, social intertia may lead

to team composition and even division of labour being determined by history rather

than what is appropriate for the task at hand (Raveendran, Puranam and Warglien,

2013). Finally, being over ruled (or even ostracized) by a community of peers may not

necessarily feel better than being overruled by one boss. 7

Fourth, the speed of decision making may be slower as consensus takes time and

lateral escalation has no obvious stopping point other than the boundary of the firm

(unlike vertical escalation, which does). Consider the following quote from a GitHub

employee’s blog:

“Talented people with a lot of experience have strong opinions. With

experience comes baggage, and we all have baggage. We swear off tools
or techniques because they failed in one situation, without realizing it
simply was not a good fit for that situation. Now imagine 80 experienced
people all trying to build amazing things together. There are a lot of
intense discussions about what tools we should or should not use and
how features should or should not work. Most of it is productive, but it
can become tiring. I have learned that nothing solves an argument like
a pull request with working code. Working code moves the conversation
forward, and changes are made from there. If you don’t like someone’s
pull request, then go create your own and lobby for its acceptance.
Compared to what people have to deal with in other work environments,
these are all first world problems.”Error! Bookmark not defined.

-
“The vast majority of the time that's it. Everyone is happy, everything is
great. We don't run into a lot of paralyzing problems all too often. When
we do, we turn back towards the open source context I mentioned at the
start of this. Open source may not have "managers" in the traditional

77Does the boss not serve a symbolic common focal point of dislike that unites the
subordinates?
 sense, but it does have leaders… people who have been in the codebase
for years and understand context beyond just the immediate problem at
hand. I wouldn't really call Linus a (traditional) manager, but he
definitely plays the leadership role when he needs to on Linux.
Accordingly, we have a couple people in charge of every product or
team that take that position. If something is really crazy, we usually
defer to their expertise. Everyone's opinion doesn't matter equally. My
opinion as a bystander just floating by a project's pull request matters
much less than the team that's been working day-in, day-out on that
particular pull for three months now. The team's opinion defers to the
leadership of the project itself. It's not some weird power scramble,
either: almost in all cases it just comes down to experience, and if you've
been living and breathing something for four years, there's a good
chance you have a deep context that others new to the project don't have
That said, if you're at the point where you need someone else to come in
and mediate a massive fight on a pull request, it's almost always a sign
that both sides are idiots and you should go back to the drawing board
and figure out a different way. It's one of those cases where whatever
problem you're facing is probably just a really hard problem with valid
concerns from both parties. Take a breather, and, in a lot of our cases,
just scrap the entire project. We've had a number of projects staff-
shipped for six months (or even twelve months) and ultimately we
decided no one was happy so we scrapped the whole thing and
unshipped it.”8

The role of authority in non-hierarchical firms

In conventional authority hierarchies, given the goal of the organization, how to factor

it into constituent tasks, how to distribute responsibilities, how to motivate and how to

coordinate are the primary responsibility of authority. The notion of division of labour

based on self-selection is at the heart of so called boss-less firms. In these

organizations, the overall goal of the organization is decomposed into sub-goals not

by the decision of an authoritative superior, but rather emerges spontaneously through

individuals recognizing valid sub-goals, and coalescing around them. The role of

authority, to the extent it is exercised, is thus limited by a recognition that many of the

8 https://github.com/holman/feedback/issues/430
interactions necessary for the organization cannot be foreseen. Nonetheless, authority

may still play an important role in creating a context that allows such interactions to

arise as and when required. To make this concrete, consider the following:

- The founders of open source software development communities exert a

strong influence on them, in particular in shaping the “architecture of

participation”; they can define the project mission and architecture in a

way that makes it easier or harder for others (whom the founders do not

know) to contribute at a time and in a manner which the founder cannot

really anticipate.

- The headquarters of multi-business firms, instead of looking for and

pursuing synergies between divisions, could instead create an “architecture

of collaboration”- a procedure that helps divisions agree how to depute

individuals, share resources and revenues to pursue opportunities that the

divisions themselves discover and surface to HQ. The role of HQ is to

help divisions form these “internal alliances” more easily than they could

with external entities.

- Physical location of activity remains an extremely important (but often

invisible) organization design variable. Formal reorganizations may be

undermined or strengthened by decisions about physical relocation.

Choices about location in physical and organizational space can be made

in such a way as to reduce the silos and barriers to interaction each alone

create, in order to allow self-selection based informal teams to emerge.

Conclusion
The examples of boss-less firms discussed here are diverse organisations within vastly

different industries, but there are important commonalities. In its own way, each of

these experiments in organising has replaced the conventional corporate hierarchy

with self-organised teams, i.e. teams that are formed through individuals selecting

what to work on, with whom, how and when. Because these team members have a

greater hand in shaping their own work conditions and expectations, they require less

hand-holding, training, inducement and discipline from managers compared to

employees motivated purely by desire for a pay rise or the threat of termination. This

is the mouth-watering promise held forth by these attempts at radical decentralisation

- a dramatic reduction in the traditional costs of management.

It is however important to note that the “boss-less” model may have a significant

scalability problem, and the successes we see may be rather special cases. To see

why, it is useful to ask why the multi-layered authority hierarchy came into existence

in the first place. When it takes the expensive contributions of many to produce

something, and their contributions have complex linkages between them, bosses play

critical roles in motivating and coordinating contributors, as well as resolving disputes

between them. But they have limited spans of control, which gives rise to layers in

the hierarchy. It follows that to dispense with the hierarchy of bosses for the

production of complex products, peer-to-peer dispute resolution through consensus

will be critical. That’s where scalability constraints begin to bite, because consensus

decision-making is very hard to scale. It is worth noting that all start-ups look like

self-selected teams, but most switch to conventional authority hierarchies as they

grow.
There are other potential problems too. Consensus wisdom may be preferable to

conventional leadership in many ways, but time efficiency isn’t one of them.

Hierarchy provides a useful “stopping rule” – which can terminate debate and reach

decisions fast (though these decisions may not necessarily be the best ones). Without

task assignment being handled from above, talent may migrate toward the most

conspicuous project options available, not necessarily the most critical. This is a

problem that Nature, in the design of ant-colonies, has solved beautifully. But it took

millions of years of evolution to do it.

The right question may not be whether one should dismantle the hierarchy or not, but

instead where to do so (and how much), as the classic organisation theorists Burns

and Stalker observed several decades ago (1958).

In certain parts of the organisation, where projects require complex large scale

interactions, critical projects may not necessarily be attractive, and consensus within

projects is difficult to achieve, the hierarchy should continue to reign. In others,

where the inputs of a few key individuals are sufficient for a project, where such

individuals may have a better sense of which projects are critical than their bosses,

and where peer-to-peer consensus within projects can be relatively easily achieved

because of professional norms and reputation concerns, the principle of self-selection

into teams may be very powerful indeed. This is where the traditional hierarchy may

offer very little advantage. Yet, even here there is an implicit role for authority; it

creates the context within which the spontaneous interactions necessary for self-

selected teams can occur.

The best of both worlds may belong to companies that tear down the hierarchy around

certain departments and leave it intact elsewhere, essentially creating “boss-less”

firms within a hierarchical structure. After all, most would agree that manufacturing

departments might be bogged down by the freedom a “flat” structure affords, but that

same freedom may prove to be a wellspring of innovation for an R&D department.

The opportunities and challenges lie in taking these ideas beyond the traditional

R&D-manufacturing dichotomy to unusual contexts. For instance, rather than impose

its views on how synergies must be realised between divisions in a multi-business

firm, what if headquarters instead encouraged divisions to self-select into synergy

projects and form internal alliances to manage them, based on their own estimation of

mutual benefit?
Table 1: Comparing traditional hierarchies and boss-less firms