In 1989, the collapse of the Berlin Wall brought the crumble of the defense industry in the United States. The end of Soviet containment requirements ended the demand for large-scale weapons, and noted industry pioneer General Dynamics took a large share of the brunt. By 1991, the company was losing money and saddled with $600 million in debt. Vice-chairman Bill Anders was named CEO of a company that appeared headed for sale or bankruptcy.

Anders’ first order of duty was to improve the culture at General Dynamics. Upon taking the reins, Anders found that lower-level managers were increasingly focused on developing larger and faster weapons with no regard for profitability. He wasted no time in replacing twenty-one of the company’s twenty-five managers and aligned each with profitability metrics that created value for shareholders. When that task was complete, Anders began focusing on his main role – allocating company resources.

He promptly sold-off overvalued parts of the company and began looking to purchase new businesses. When no opportunities presented themselves, he decided to return cash to the company’s shareholders. When all was said and done, Anders had sold-off over half the company, watched the lower-level managers improve profitability, and generated $5 billion in proceeds for shareholders. Of those shareholders, Anders’ track-record had attracted a famous name – Warren Buffett, CEO and Chairman of Berkshire Hathaway.

$1 invested in Berkshire Hathaway when Buffett took over in 1964 had turned into $27,442 by January 2020. Buffett’s main strategy consisted of using the cash from insurance operations to purchase companies with — among other things — competent management teams. Regarding those management teams, in 1986 he wrote:

“At Berkshire, we feel that telling outstanding CEOs, such as Tony [Nicely of GEICO] how to run their companies would be the height of foolishness. Most of our managers wouldn’t work for us if they got a lot of backseat driving.”

For Berkshire however, not all management teams are worthy of attention. Of those that Buffett feels fortunate to observe at a distance, the Oracle of Omaha makes an interesting note:

“The supreme irony of business management is that it is far easier for an inadequate CEO to keep his job than an inadequate subordinate.”

William Thorndike’s book The Outsiders profiles both Buffett and Anders (in addition to six others) and concludes that the best executives provide autonomy to competent lower-level managers and make the management of company resources their top priority. I consider Thorndike’s thesis an effective one, and I think it has insights that can help us understand the situation around Judd Brackett and the role of management in the NHL.

Is this oversimplified? Absolutely.

However, sometimes first principals are easiest to forget. To that end, simple mental models can help us think about why a decentralized strategy actually works, and how to determine which level of an organization should perform various functions.

There are two general categories of organizational processes that occur in the construction of a contending roster. First, the team has to allocate assets (draft picks, internal budgeting, salary cap space, and roster players — within the constraints set forth by the CBA) in such a way that the collective roster peaks over some “window” that optimizes the probability of winning the Stanley Cup. This process inherently requires establishing an overall strategic direction and maintaining organizational buy-in.

Second, when an asset is allocated in a specific way, the value of that asset needs to be maximized. This happens through multiple individual processes, including scouting, coaching and player development. It is important to note that value is dependent on the stage of a team’s cycle — this is why trades happen. For example, a team who feels they can compete in the future is more than happy to part with assets whose output is maximized in the present.

Astute readers will point out that there is some overlap. To that end, the person (or people) responsible for making decisions around asset allocation and strategic direction need an understanding of all the individual processes supporting asset maximization. For our purposes, an asset-allocation decision would be “We’re not going to compete for a few years, so we best trade roster player x for a prospect,” or “The market at this year’s deadline is high. We should consider forgoing a year of competing to extract value for our roster players.” Asset value-maximization involves determinations like: “Prospect y isn’t as strong as prospect z. We should look at trading for him instead,” or “It’s best for this player’s development if we give him drills xyz.”

Before determining how these processes are best executed, we need to examine the domain in which hockey operations decisions take place. In a memo to clients, hedge-fund manager Howard Marks quoted Jackson Grayson’s book Decisions Under Uncertainty: Drilling Decisions by Oil and Gas Operators. Marks noted that you can make decisions based on what you know, but the outcome may still be guided by 1) knowable factors you missed, or 2) unknowable/random factors that manifested themselves. When a decision-maker operates in a domain in which 2) is numerous and prominent, the domain is conducive to high-degrees of luck. Decisions are made on a probabilistic basis, and a strong forecasting process may create a poor result.

Effectively, almost all professional hockey decisions are made in probabilistic domains. As a result, the focus of every organizational function has to be improving process, with the hope that increased trials eventually create strong results. To that end, there are three activities undertaken by decision-makers at each trial:

  1. They incorporate as many knowable factors that could significantly drive the outcome as possible.
  2. They analyze the factors by linking them to an eventual outcome.
  3. They gauge the probability of the manifestation of each factor, and therefore the probability of each outcome.

Knowing better information and improving the analysis of that information is what teams seek through process improvement. The meaningful long-term outperformance of any organizational function in a probabilistic domain is the result of conducting these activities better than peers.

Making decisions with many and sometimes unknowable variables is theoretically best accomplished through collectives. James Surowiecki’s book The Wisdom of Crowds provides a compelling example. In the 19th century, scientist Francis Galton polled a series of random Scottish villagers on the weight of an ox. None of the villagers had any previous experience as farmers or butchers, but in the exercise, each brought some unique perspective to the problem. The average of the guesses culminated to 1197 lbs, against an actual ox weight of 1198 lbs. The irrationalities in the guesses cancelled out, and the cognitive diversity of the group allowed the average to incorporate more factors than could be captured by an individual.

While tapping collectives for all probabilistic decisions might seem intuitively enticing, there is an important trade-off between team size and coordination costs. A team can grow until the additional coordination cost of a new member is greater than the additional value-add of their perspective. Trades are an interesting example. While the process of valuing incoming and outgoing assets can (and should) be executed by a relatively larger group, the coordination costs of negotiating a trade through a “trade-team” would invariably result in a significant value-loss.

As it pertains to amateur scouting, an asset value-maximization function, a group of cognitively diverse individuals with some expertise is easily able to incorporate more factors than an individual. The lower coordination-costs here relative to the benefits render this the optimal way to perform a broad function. As I write this article, I try and present it with a neutral stance on the Judd Brackett departure and upper-level interference. However, it is interesting to note this reference to the 2016 draft in Thomas Drance’s piece in The Athletic:

“It’s no secret that Finnish defenceman Olli Juolevi was long the apple of Benning’s eye in the lead-up to the draft,” wrote Drance, “In retrospect, organizational leaders worried that the club’s scouting process had been focused too narrowly on one top prospect. That the debates in the meetings and in composing the list hadn’t dug deeper.”

While the trade-off between diversity and coordination cost is apparent, it is also apparent that groups of a certain size require structuring. The opinions of the group need to be aggregated and thought diversity must remain intact. This is best performed by a leader, and the primary question in the Judd Brackett debate is the distance from the problem source the leader must maintain.

To determine this, we must examine the number of variables the leader must analyze and the degree of specialization required for analysis. In the case of amateur scouting, I’d argue that the former is so prominent that it automatically requires a leader who maintains that organizational function as his sole focus. I’d additionally argue that the apparent transferability of the specialization (pro-scouting may be thought similar) makes GMs ignore the scope of responsibility and overstep.

To further drive my point – we never see GMs try and perform surgery on players (luckily), because their skills are clearly non-transferrable. It seems convenient that the organizational functions which apparently require the most interference are those in which the GM apparently has the most expertise.

Laurence Gilman often says that “players play, coaches coach, and managers manage”. We can extend this idiom to include the fact that scouts scout, development guys develop, and capologists cap. Effectively, the drivers of most asset value-maximization functions in a hockey operation are too broad for one person to handle, and that one person certainly can’t handle them all. Groups are often required to incorporate broad perspectives, and these groups usually require their own leaders for structure and cohesion.

So, if a GM ought to keep his role minimal in these functions, what should he actually do?

The allocation of assets is also done on a probabilistic basis and the drivers range through every asset value-maximization function of the organization. This entails determining the direction of the team, ensuring department leaders buy-in, and maintaining a consistent philosophy across the board. At the end of the day, if pro scouts identify a high-end talent, it must be decided that pursuing this acquisition is in the best interests of the team and consistent with the way a coach utilizes the roster.

To that end, the person or people responsible for asset-allocation need the following:

  • A high-level understanding of each department within the organization.
  • A view towards how the team intends to reach contention over a prolonged period.
  • An ability to identify managers for each department able to operate in a consistent manner.
  • A clear understanding of the value of the assets under management.

This undoubtedly should remain the primary role of the GM. While the size of the group that aids in this effort is at the GM’s discretion, the GM should take the lead. Determining priorities, where the organization should aim next, and establishing an overarching culture is a large task, but still smaller than attempting to interfere in every facet of the organization.

In short, I’m presenting the following – a structure in which management focuses on higher-level ideas around strategic direction and asset allocation, and pushes the operation of value-maximizing functions to competent lower-level managers (depending on team size). The executive seeks lower-level managers with similar philosophies to maintain organizational consistency, and as they perform, make adjustments and suggestions as necessary. Should a department be significantly outperforming peers, that department best be left alone, no matter the GM’s expertise.

An Outsider GM and His Radical New Structure

Mike Gillis accomplished a lot after leaving the Canucks. The former GM visited Australian Football Clubs, Soccer teams, and sat on the board of a University of Michigan think-tank in an effort to figure out what drove a winning organization.

At the coaches conference sponsored by TeamSnap, Gillis noted the following:

“One of the biggest issues NHL teams have is that a lot of that responsibility resides with one person,” he said. “That one person over time gets worn out and makes poor decisions.”

How should a team structure itself according to Gillis? From Sportsnet,

“Gillis, who appreciates the Premier League organizational chart, believes an NHL front office could maximize its effectiveness by hiring four assistant GMs, plus a behind-the-scenes cast of problem-solvers devoted to maximizing its individual players through the study of hard evidence and suggesting their ideal linemates and situational usage to the GM and coach.”

Sounds oddly familiar.

Thorndike’s book changed the way many investors thought about management teams. Credit was no longer given to operational expertise, the ability to meet analyst forecasts, or appearances in Davos. Instead, analysis centred on how management was able to think about managing company resources and establishing a strategic direction.

Within the NHL, you wonder if Thorndike’s insights could be applied to achieve similar results.