How Apple Is Organized for Innovation

When debates reach an impasse, as some inevitably do, higher-level managers weigh in as tiebreakers, including at times the CEO and the senior VPs. To do this at speed with sufficient attention to detail is challenging for even the best of leaders, making it all the more important that the company fill many senior positions from within the ranks of its VPs, who have experience in Apple’s way of operating.

However, given Apple’s size and scope, even the executive team can resolve only a limited number of stalemates. The many horizontal dependencies mean that ineffective peer relationships at the VP and director levels have the potential to undermine not only particular projects but the entire company. Consequently, for people to attain and remain in a leadership position within a function, they must be highly effective collaborators.

That doesn’t mean people can’t express their points of view. Leaders are expected to hold strong, well-grounded views and advocate forcefully for them, yet also be willing to change their minds when presented with evidence that others’ views are better. Doing so is not always easy, of course. A leader’s ability to be both partisan and open-minded is facilitated by two things: deep understanding of and devotion to the company’s values and common purpose, and a commitment to separating how right from how hard a particular path is so that the difficulty of executing a decision doesn’t prevent its being selected.

The development of the iPhone’s portrait mode illustrates a fanatical attention to detail at the leadership level, intense collaborative debate among teams, and the power of a shared purpose to shape and ultimately resolve debates. In 2009 Hubel had the idea of developing an iPhone feature that would allow people to take portrait photos with bokeh—a Japanese term that refers to the pleasing blurring of a background—which photography experts generally consider to be of the highest quality. At that time only expensive single-lens reflex cameras could take such photos, but Hubel thought that with a dual-lens design and advanced computational-photography techniques, Apple could add the capability in the iPhone. His idea aligned well with the camera team’s stated purpose: “More people taking better images more of the time.”

As the team worked to turn this idea into reality, several challenges emerged. The first attempts produced some amazing portrait pictures but also a number of “failure cases” in which the algorithm was unable to distinguish between the central object in sharp relief (a face, for instance) and the background being blurred. For example, if a person’s face was to be photographed from behind chicken wire, it was not possible to construct an algorithm that would capture the chicken wire to the side of the face with the same sharpness as the chicken wire in front of it. The wire to the side would be as blurred as the background.

One might say, “Who cares about the chicken wire case? That’s exceedingly rare.” But for the team, sidestepping rare or extreme situations—what engineers call corner cases—would violate Apple’s strict engineering standard of zero “artifacts,” meaning “any undesired or unintended alteration in data introduced in a digital process by an involved technique and/or technology.” Corner cases sparked “many tough discussions” between the camera team and other teams involved, recalls Myra Haggerty, the VP of sensor software and UX prototyping, who oversaw the firmware and algorithm teams. Sebastien Marineau-Mes, the VP to whom the camera software team ultimately reported, decided to defer the release of the feature until the following year to give the team time to better address failure cases—“a hard pill to swallow,” Hubel admits.

To get some agreement on quality standards, the engineering teams invited senior design and marketing leaders to meet, figuring that they would offer a new perspective. The design leaders brought an additional artistic sensibility to the debate, asking, “What makes a beautiful portrait?” To help reassess the zero-artifacts standard, they collected images from great portrait photographers. They noted, among other things, that these photos often had blurring at the edges of a face but sharpness on the eyes. So they charged the algorithm teams with achieving the same effect. When the teams succeeded, they knew they had an acceptable standard.

Another issue that emerged was the ability to preview a portrait photo with a blurred background. The camera team had designed the feature so that users could see its effect on their photos only after they had been taken, but the human interface (HI) design team pushed back, insisting that users should be able to see a “live preview” and get some guidance about how to make adjustments before taking the photo. Johnnie Manzari, a member of the HI team, gave the camera team a demo. “When we saw the demo, we realized that this is what we needed to do,” Townsend told us. The members of his camera hardware team weren’t sure they could do it, but difficulty was not an acceptable excuse for failing to deliver what would clearly be a superior user experience. After months of engineering effort, a key stakeholder, the video engineering team (responsible for the low-level software that controls sensor and camera operations) found a way, and the collaboration paid off. Portrait mode was central to Apple’s marketing of the iPhone 7 Plus. It proved a major reason for users’ choosing to buy and delighting in the use of the phone.

As this example shows, Apple’s collaborative debate involves people from various functions who disagree, push back, promote or reject ideas, and build on one another’s ideas to come up with the best solutions. It requires open-mindedness from senior leaders. It also requires those leaders to inspire, prod, or influence colleagues in other areas to contribute toward achieving their goals.

While Townsend is accountable for how great the camera is, he needed dozens of other teams—each of which had a long list of its own commitments—to contribute their time and effort to the portrait mode project. At Apple that’s known as accountability without control: You’re accountable for making the project succeed even though you don’t control all the other teams. This process can be messy yet produce great results. “Good mess” happens when various teams work with a shared purpose, as in the case of the portrait mode project. “Bad mess” occurs when teams push their own agendas ahead of common goals. Those who become associated with bad mess and don’t or can’t change their behavior are removed from leadership positions, if not from Apple altogether.

Leadership at Scale

Apple’s way of organizing has led to tremendous innovation and success over the past two decades. Yet it has not been without challenges, especially with revenues and head count having exploded since 2008.

As the company has grown, entering new markets and moving into new technologies, its functional structure and leadership model have had to evolve. Deciding how to organize areas of expertise to best enable collaboration and rapid decision-making has been an important responsibility of the CEO. The adjustments Tim Cook has implemented in recent years include dividing the hardware function into hardware engineering and hardware technologies; adding artificial intelligence and machine learning as a functional area; and moving human interface out of software to merge it with industrial design, creating an integrated design function.

Another challenge posed by organizational growth is the pressure it imposes on the several hundred VPs and directors below the executive team. If Apple were to cap the size or scope of a senior leader’s organization to limit the number and breadth of details that the leader is expected to own, the company would need to hugely expand the number of senior leaders, making the kind of collaboration that has worked so well impossible to preserve.

Cognizant of this problem, Apple has been quite disciplined about limiting the number of senior positions to minimize how many leaders must be involved in any cross-functional activity. In 2006, the year before the iPhone’s launch, the company had some 17,000 employees; by 2019 that number had grown more than eightfold, to 137,000. Meanwhile, the number of VPs approximately doubled, from 50 to 96. The inevitable result is that senior leaders head larger and more diverse teams of experts, meaning more details to oversee and new areas of responsibility that fall outside their core expertise.

In response, many Apple managers over the past five years or so have been evolving the leadership approach described above: experts leading experts, immersion in the details, and collaborative debate. We have codified these adaptions in what we call the discretionary leadership model, which we have incorporated into a new educational program for Apple’s VPs and directors. Its purpose is to address the challenge of getting this leadership approach to drive innovation in all areas of the company, not just product development, at an ever-greater scale.

When Apple was smaller, it may have been reasonable to expect leaders to be experts on and immersed in the details of pretty much everything going on in their organizations. However, they now need to exercise greater discretion regarding where and how they spend their time and efforts. They must decide which activities demand their full attention to detail because those activities create the most value for Apple. Some of those will fall within their existing core expertise (what they still need to own), and some will require them to learn new areas of expertise. Activities that require less attention from the leader can be pushed down to others (and the leaders will either teach others or delegate in cases where they aren’t experts).

Rosner, the VP of applications, provides a good example. Like many other Apple managers, he has had to contend with three challenges arising from Apple’s tremendous growth. First, the size of his function has exploded over the past decade in terms of both head count (from 150 to about 1,000) and the number of projects underway at any given time. Clearly, he cannot dive into all the details of all those projects. Second, the scope of his portfolio has widened: Over the past 10 years he has assumed responsibility for new applications, including News, Clips (video editing), Books, and Final Cut Pro (advanced video editing). Although apps are his core area of expertise, some aspects of these—among them editorial content for News, how book publishing works, and video editing—involve matters in which Rosner is not an expert. Finally, as Apple’s product portfolio and number of projects have expanded, even more coordination with other functions is required, increasing the complexity of collaborating across the many units. For instance, whereas Rosner is responsible for the engineering side of News, other managers oversee the operating system on which it depends, the content, and the business relationships with content creators (such as the New York Times) and advertisers.

To cope, Rosner has adapted his role. As an expert who leads other experts, he had been immersed in details—especially those concerning the top-level aspects of software applications and their architecture that affect how users engage with the software. He also collaborated with managers across the company in projects that involved those areas.

But with the expansion of his responsibilities, he has moved some things from his owning box—including traditional productivity apps such as Keynote and Pages—into his teaching box.
Now he guides and gives feedback to other team members so that they can develop software applications according to Apple’s norms. Being a teacher doesn’t mean that Rosner gives instruction at a whiteboard; rather, he offers strong, often passionate critiques of his team’s work. (Clearly, general managers without his core expertise would find it difficult to teach what they don’t know.)

The second challenge for Rosner involved the addition of activities beyond his original expertise. Six years ago he was given responsibility for the engineering and design of News. Consequently, he had to learn about publishing news content via an app—to understand news publications, digital advertising, machine learning to personalize news content, architecting for privacy, and how to incentivize publishers. Thus some of his work fell into the learning box. Here managers face a steep learning curve to acquire new skills. Given how demanding this is, only critical new activities should fall into this category. Over six years of intense learning, Rosner has mastered some of these areas, which are now in his owning box.

As long as a particular activity remains in the learning box, leaders must adopt a beginner’s mindset, questioning subordinates in a way that suggests they don’t already know the answer (because they don’t). This differs starkly from the way leaders question subordinates about activities in the owning and teaching boxes.

Finally, Rosner has delegated some areas—including iMovie and GarageBand, in which he is not an expert—to people with the requisite capabilities. For activities in the delegating box, he assembles teams, agrees on objectives, monitors and reviews progress, and holds the teams accountable: the stuff of general management.

Whereas Apple’s VPs spend most of their time in the owning and learning boxes, general managers at other companies tend to spend most of their time in the delegating box. Rosner estimates that he spends about 40% of his time on activities he owns (including collaboration with others in a given area), about 30% on learning, about 15% on teaching, and about 15% on delegating. These numbers vary by manager, of course, depending on their business and the needs at a given time.

The discretionary leadership model preserves the fundamental principle of an effective functional organization at scale—aligning expertise and decision rights. Apple can effectively move into new areas when leaders like Rosner take on new responsibilities outside their original expertise, and teams can grow in size when leaders teach others their craft and delegate work. We believe that Apple will continue to innovate and prosper by being organized this way.

. . .

Apple’s functional organization is rare, if not unique, among very large companies. It flies in the face of prevailing management theory that companies should be reorganized into divisions and business units as they become large. But something vital gets lost in a shift to business units: the alignment of decision rights with expertise.

Why do companies so often cling to having general managers in charge of business units? One reason, we believe, is that making the change is difficult. It entails overcoming inertia, reallocating power among managers, changing an individual-oriented incentive system, and learning new ways of collaborating. That is daunting when a company already faces huge external challenges. An intermediate step may be to cultivate the experts-leading-experts model even within a business unit structure. For example, when filling the next senior management role, pick someone with deep expertise in that area as opposed to someone who might make the best general manager. But a full-fledged transformation requires that leaders also transition to a functional organization. Apple’s track record proves that the rewards may justify the risks. Its approach can produce extraordinary results.