Seminar recap, session 6, 12-9
Does Institutional Rationality differ by size of institution?
At our sixth session, we talked about how large organizations use bureaucratic mechanisms to operate rationally. We talked about how a small, aggressive startup differs from a large established business. Examples of large enterprises that came up during the discussion included Microsoft, Amazon, Freddie Mac, J.P. Morgan, the U.S. military, and the Church of Latter Day Saints.
The characteristics of information technology are important. I long ago observed that an organization gets the information technology that it deserves. A small start-up or highly independent unit will tend to have an elegant but undocumented system that exists in one talented engineer’s head. I recalled that in the late 1980s Freddie Mac had one mission-critical system written in APL, a concise language that only two employees were familiar with. About 15 years later, when my web business acquired an equivalent business, we found that their application was written in PERL, a computer language aptly described as looking like the characters used in a comic strip to denote cursing.
An established business needs reliable, well-maintained systems. That means that it needs code that is readable and documented. By 1990, Freddie Mac was engaged in a major systems redevelopment project, which took APL out of the picture. My web business did a total rewrite of the software used by the firm we acquired.
I have also observed that the adoption of formal bureaucratic processes in place of informal norms takes place around the point where a business crosses the Dunbar Number of about 150 employees. When a business is sub-Dunbar, every employee knows how what he does affects other employees. A single conversation among a few employees can be sufficient to generate a major change of direction for the business.
When the business is super-Dunbar, employees do not know one another. They can inadvertently work at cross purposes. A major change of direction cannot occur without following a process that ensures that every affected department is aware and on board with the new approach.
The bureaucratic processes that super-Dunbar firms employ, such as organizational charts, employee manuals, training, compensation systems, and formal project management, can help to deal with information problems and principal-agent problems. Employees in a large organization will only work in a coordinated fashion if the incentives lead them to do so and if they have sufficient information about what the business needs from them. But information must flow up as well as down, or else executives will make bad decisions.
Consider the fiasco of the Obamacare web site launch. When a business undertakes a major initiative, it appoints a project executive who has the authority to make decisions to alter the project to make it workable and who has the clout to get recalcitrant departments to provide timely participation. This was evidently not the case with the Obamacare web site, where the project was overseen by middle managers who could not possibly knit together the legacy systems running on mainframe computers into a real-time web application.
A small start-up can change direction quickly. It can “move fast and break things.” A large, successful business is more inclined to operate according to the maxim, “If it ain’t broke, don’t fix it.” It needs rules in place that do not allow middle managers to undertake initiatives without first going through an approval process.
In finance, traders and small sales units need autonomy and freedom of action. But the firm also has to protect itself from rogue traders or from units taking on too much risk. This requires strict rules and enforcement mechanisms.
Microsoft had very strict rules for sales teams that kept them from “poaching” on one another’s territory. I speculate that this has to do with the fact that software has very low marginal cost to distribute. When marginal costs are low and fixed costs are high, price discrimination is a key strategy. So you don’t want one sales team expanding by undercutting the prices charged by another sales team.
Microsoft’s operating system software is a large, integrated system. It needs bureaucratic procedures in place to manage changes, so that an “improvement” in one application does not cause something else to break.
In contrast, Amazon tries to keep its software projects small enough so that they can be operated by two-pizza teams. But this requires highly disciplined development with careful documentation.
All firms must manage a trade-off between reliability and agility. Small start-ups have more to gain from agility and less to lose from unreliability, so they can have fewer rules and fewer restraints on individual employees. Large businesses or government agencies must instead aim for reliability, necessary sacrificing agility.
Amazon has an interesting “Day 1” culture. The idea is to act as if it is a startup, and the biggest need is to earn customer loyalty, as opposed to squeezing the most revenue out of existing customers. Will this culture persist now that its founder is no longer running the company? Founders tend to “report to the dream,” as Capital One’s founder once put it. So Amazon under its founder could be focused on customers. But a successor is likely to be more inwardly focused, since getting to the top of an existing company requires navigating internally.
Although large firms try to avoid large failures, Apple had several in its history—the Lisa, the Newton. It managed to survive these, and the bigger successes made up for it.
The military and the LDS church both use rotation as a method to try to reduce staleness and improve information flows. Capital One also uses rotation of young managers. The military has made a point of studying management and trying to improve methods. Private firms often copy methods from the military.
Can the process of going from a small start-up to a large established business be taught? Elad Gil’s book addresses the topic. But I think that it cannot be taught. I joined Freddie Mac in late 1986, when it had just scaled up above the Dunbar number and was scaling up further, so that I could witness the whole process. This was very helpful to me when in the 1990s I started my own business. In particular, I could recognize when the need for formal processes was arising. My joke was that my karma was that I spent so much time berating information technology bureaucrats at Freddie that I came back as one when I was in my own business.
Will distributed autonomous organizations (DAOs) provide a new solution to the trade-off between reliability and agility? I don’t know enough about the concept, but my inclination is to be skeptical.
It’s “The Church of Jesus Christ of Latter-day Saints.” You left out the important part.
"I have also observed that the adoption of formal bureaucratic processes in place of informal norms takes place around the point where a business crosses the Dunbar Number of about 150 employees."
Some observations:
Even when many things become officially formalized and bureaucratic, those de jure rules and procedures are a map that only bears slight resemblance to the actual territory. Actual practices and incentives deviate - sometimes quite substantially - from the picture one would get by taking the black letter rules at face value. New members of teams often need time to figure out the real deal. On the one hand, this state of affairs creates a lot of needed flexibility to bend and adapt in order to get things done without too many potential points of veto or having to make a federal case out of it. On the other hand, leaders can easily get so used to bending the rules that they take it for granted and become complacent about rubber-stamping whatever new official rules or policies come across their desk for approval. "Why scrutinize carefully? If it turns out to be a problem, we'll just bend informally like we always do." The predictable result is a premature and unwieldy accretion of often incoherent bureaucratization which - as soon as the culture of 'they are more like advisory guidelines, bend whenever you want' disappears with turnover or becomes otherwise untenable - traps the spider in its own web and gums up the works.
Where I work, the large organization does not have one character but two distinct ones.
On the one hand, there is a sub-Dunbar-sized inner circle and various cliques at or close to the executive / top staff officer level, and things in that circle often operate in a de facto sub-Dunbar manner. Discussions, decisions, and plans in that circle are extremely secretive and compartmentalized, with as many matters handled in a personal, relationship-based, confidential, face-to-face manner as possible. Whether their fellow insiders are worthy of it or not, things operate by necessity along the lines of "high trust equilibrium", that is, with mostly voluntary cooperation without mechanisms for monitoring, verification, or discipline. People out of this circle, even just below it, almost never have any good idea of what is being done or said or planned within it.
Outside the sub-Dunbar inner circle there is the vast super-Dunbar workforce. And, for any particular position in that workforce, one's experience of degree of formal regulation and bureaucratization seems to be correlated with one's overall distance in the hierarchy from the inner circle (in terms of status, rank, prestige, pay, skill-level, autonomy, decision-making power, etc.), the numbers of other people doing the same job, and the degree to which one is externally or internally-focused, that is, 'front-line', interacting with people and objects in the real world outside one's own organization, instead of processing, analyzing, and manipulating abstract ideas, numbers, words, and symbols.
Still, even though it's a spectrum with many levels and flavors of effective bureaucratization, there is a uniquely clear and stark discontinuity between the inner circle and everything else.
One interesting aspect of this dual-charactered organization, at least where I work, is that there is a severe amount of friction and impairment of communication, direction, guidance, and trust at the border between the two worlds, which in turn generates a lot of angst, drama, and general dysfunction. My impression is that elite managers still have not quite discovered how to lead, manage, and administer effectively in such organizationally discontinuous circumstances, or else, it simply takes a combination of great skill, experience, a highly beneficial incentive structure, and independence to avoid these issues.