In this week’s feature book, High Output Management, the great CEO Andrew S. Grove dedicates a lengthy portion of text to an idea that I didn’t like for a long time. It involves what he refers to as “hybrid organizations”, aka “matrix organizations”, aka “places where you probably have two or more bosses.” Can you imagine such a thing? Two bosses? Or, as a manager, can you imagine sharing your staff with others? Horrible, right?

Indeed. That’s how I’ve viewed the idea for a very long time. Past research on the topic had me fixated on an classic article from HBR on the topic, titled Problems of Matrix Organizations”. More recently, there are concerns that Luis Goncalves articulates in this article from under the not-exactly-subtle title: Why a matrix organizational structure will destroy your company that make a lot of sense, too. It’s a good article. It points to real issues.

And more importantly, a lot of the research on matrix organizations points to a fundamental truth: the organization’s structure is supremely important. If you get it wrong, you have a whole host of issues. Many of which you won’t be able to diagnose because, once the organization is structured, we tend to forget about it. We don’t look back. We just move on, putting more System 2 thinking into our Starbucks order than we in the emergent after-effects of the org chart.

A Bad System Beats A Good Person Every Time

When I spent a week examining Systems Thinking, I wrote an article on the different journeys Sony and Apple took to launch their digital music players. You can find it at this link.

I’ve always loved that story because it articulates the critical, inescapable truth that the organizational structure determines the limitations of the company. Regardless of leadership. Sony tried to straddle a number of divisional needs (hardware, software, record licensing, design) in making its Walkman of the Future. Because they’d always done it that way. Apple, meanwhile, consolidated their teams differently, without competing divisions, and delivered something far more integrated, convenient, elegant, and attractive in the iPod.

A great exploration into the pros and cons of Apple’s organizational approach is provided by Ben Thompson at the always-fantastic Stratechery.

Suffice to say, a few of the reasons Apple won the mP3 market with a better, more-integrated product comes down to a distinction between Apple’s single profit and loss and Sony’s classic P&L across multiple individual, standalone divisions and the drag it creates. That said, there are clearly times when Sony’s example of structure is better. For example, the economies of scale are really nice in those discrete verticals. This gets to an early point made by our book’s author:

Good management rests on a reconciliation of centralization and decentralization.

When To Centralize, When To Decentralize?

It’s always a good idea to think of centralization vs decentralization as a first principle. Whether its for the sake of management or for product and service design. Because as other great business leaders have said, you make money by either “bundling or unbundling”. It works for software design, too. Connor Shorten has more thoughts on this in his article Designing Features for your App: Bundling and Unbundling.

Anyway, back to Grove: when it comes to organizations and the question of centralization versus decentralization, the critical component of the thinking here is the balance between mission-oriented work and function-oriented work. When the organization is mission-oriented, it is much more centralized a’la Apple and it’s single P&L. When the organization is function-oriented, it is much more decentralized a’la Sony and its multiple, competing P&Ls.

Famously, Apple tries to act like a startup. Startups are much more mission-oriented, much more integrated, and also much smaller. Startups operate in areas of uncertainty with significant disadvantages against the status quo and so a more guerilla-like force is needed. The troops are always rallied, so to speak.

Ordinarily, though, a company like Apple would decentralize because they’ve grown to such a degree, with such a prominence in the market, that they can now rest easy and settle down into a functional organization. The dynamics of their business are well-known. They dominate. Uncertainty feels a little less prevalent. Sony was in the same position in the 70s and 80s and thus their decentralized structure served them quite well. They had a model (multiple ones, in fact) and they let it run, decentralizing themselves across multiple markets, products, and services so that they were more a conglomerate than a singular entity.

So we have a distinction here: in arenas of high uncertainty, emerging markets, and ever-changing trends, centralization makes a lot of sense. It echoes the natural growth cycle of a company from startup to mature company.

So to centralize or decentralize is based largely on two additional factors:

  1. The nature of the endeavor – is it mission-oriented or function-oriented? In Apple’s case, iPod was mission-oriented and everything from the glass used on the device to the service used to provide music was deeply integrated under a single P&L. In Sony’s case, the Walkman mP3 was function-oriented. Every division contributed according to their discrete divisional interests, fighting for the sake of their own multi-divisional P&Ls.
  2. The nature of the endeavor and the environment that surrounds it. In Apple’s case, the creation of an iPod was uncharted waters. They saw themselves sailing a Blue Ocean. In Sony’s case, the Walkman was just a digital version of the old thing they’d already had. They were still riding the waves of a Red Ocean. They weren’t rebuilding the company around this device; they were simply retooling the factory.

If the work is mission-oriented and involves more speculative, high-risk work (i.e., a startup), centralization is essential. And thus a more “matrixed” organization is what is needed. Grove explains it well:

What are some of the advantages of organizing much of a company in a mission-oriented form? There is only one. It is that the individual units can stay in touch with the needs of their business or product areas and initiate changes rapidly when those needs change. That is it.

Which is to say that, for all other combinations, including instances of mission-oriented work in predictable, non-speculative environments, a more conventional structure is warranted. As Grove writes:

All other considerations favor the functional-type of organization.

This leads to the following supremely-amateur diagram offered by yours truly:

So Should You Have A Matrix Organization?

Wrong question! The very discussion of “matrix” organizations has been faulty from the start. This is the point where I try to make an artful pivot by way of deliberate misdirection. First, consider the following from our author:

But the business of any business is to respond to the demands and needs of its environment, and the need to be responsive is so important that it always leads to much of any organization being grouped in mission-oriented units.

Along with this line:

Here I would like to propose Grove’s Law: All large organizations with a common business purpose end up in a hybrid organizational form.

Andy is right. All large organizations eventually coalesce on some amount of centralization. For some amount of the work. For reasons that are built on the axis that I just illustrated.

The point, I think, is that the question of whether or not to be a matrix organization is, again, the wrong question to ask. We will have some element of “matrix” organizations in every large company. It happens naturally. It’s the stuff of the informal structures that underlie our formal org charts. Sure, Joe might be “the boss” in title but his secretary runs the show.

In fact, I think our author knew a lot more than me about matrix organizations. Yet, he never once used the term in his book. He always stuck to the term “hybrid organizations” and I now understand why. Consider these final quotes for today’s article:

… the most important consideration should be this: the shift back and forth between the two types of organizations can and should be initiated to match the operational styles and aptitudes of the managers running the individual units.

and

Like it or not, the hybrid organization is a fundamental phenomenon of organizational life.

A matrix organization is, in some ways, an overcompensation against the limits of the traditional functional design of an old-school Sony. And as our systems thinking will tell us, overcompensations to one extreme leads to overcompensations on another extreme and we oscillate back and forth until we settle, at some point, on equilibrium. In this case, to our author’s salient point, we eventually balance ourselves onto the point of a “hybrid” that features some ad hoc and informal matrix characteristics and some formal divisional styles.

Embrace the Ambiguity

The actual org chart, as a document, can never capture this complexity in motion. Worse still, it can make the less-experienced think this behavior can be controlled. It can’t. So “Grove’s Law” holds firm.

We managers have a faulty desire to overdetermine. The minute we realize that we need staff to work in more than one vertical (i.e. silo, department, division) on a topic, we decide to formalize that idea with something like a matrix. Because we think we’re either matrix or we’re not. With nothing in-between.

Andy transcended that tendency at Intel. He instituted a third way. The space in-between. The hybrid organization that modulates between centralization and decentralization on the nature of the work (Mission critical?) and the nature of the challenge (New markets? High risk?).

He’s the first to admit, in his book, that it creates ambiguity. But ambiguity will exist regardless. We just need a way to operate within it properly. His model of the hybrid organization does this to terrific effect.


Photo by Katie Moum on Unsplash