Rethinking the Decision Factory

Companies everywhere struggle with the management of knowledge workers. They compete fiercely to find and retain the best talent, often accumulating thousands of managers in the process. For a while this is fine, but inevitably, usually when economic conditions turn less favorable, they realize that these expensive workers are not as productive as hoped, and in an effort to manage costs they lay off a large swath of them. Soon after, though, they’re out recruiting again.

This cycle is highly destructive. Aside from the human and social costs involved, it is extremely inefficient for a company to manage any resource this way, let alone one that is widely recognized as the engine of growth in the modern age. What’s especially puzzling is that the companies that engage in this cycle include some of America’s most revered role models. General Electric, for example, conducted extensive management layoffs in the 1980s and early 1990s. After a gradual regrowth in its ranks, the company announced another round of layoffs in 2001. By 2007 the numbers were back up again—until the recession forced cuts once more. Colgate-Palmolive, MetLife, Hewlett-Packard, and PepsiCo, among others, have all recently gone through the same process.

Why do these companies struggle so much with what ought to be their most productive assets? The answer, I think, is rooted in a profound misunderstanding—despite decades of research and debate about the knowledge economy—of how knowledge work does and does not differ from the manual work we have come to understand so well. In particular, most companies make two big mistakes in managing knowledge workers. The first is to think that they should structure this workforce as they do a manual workforce—with each employee doing the same tasks day in and day out. The second (which derives in part from the first) is to assume that knowledge is necessarily bundled with the workers and, unlike manual labor, can’t readily be codified and transferred to others.

In the following pages I’ll demonstrate how destructive, if understandable, these two beliefs are and describe an emerging paradigm for the management of knowledge workers that is being developed at Procter & Gamble. The initial results from this approach are extremely encouraging; if it is applied more widely, we may finally be able to bid farewell to the current, perverse management cycle.

The Rise of the Decision Factory

What do knowledge workers actually do? Clearly they don’t manufacture products or perform basic services. But they do produce something, and it is perfectly reasonable to characterize their work as the production of decisions: decisions about what to sell, at what price, to whom, with what advertising strategy, through what logistics system, in what location, and with what staffing levels.

At desks and in meeting rooms, every day of their working lives, knowledge workers hammer away in decision factories. Their raw materials are data, either from their own information systems or from outside providers. They produce lots of memos and presentations full of analyses and recommendations. They engage in production processes—called meetings—that convert this work to finished goods in the form of decisions. Or they generate rework: another meeting to reach the decision that wasn’t made in the first meeting. And they participate in postproduction services: following up on decisions.

Decision factories have arguably become corporate America’s largest cost, even at big manufacturers like P&G, because the salaries of these factory workers far exceed those of workers in physical factories. In pursuit of the twin goals of efficiency and growth, companies in the latter half of the 20th century spent ever-greater amounts on R&D, branding, information technology systems, and automation—all investments that necessitated hiring an army of knowledge workers.

I vividly remember working with the CEO of one of North America’s largest bread manufacturers in 1990–1991. He had just replaced a labor-intensive and antiquated plant with the most advanced bread bakery on the continent. He proudly told me that the new computerized ovens and packaging machinery had reduced direct labor costs by 60%. But meanwhile, a throng of new and expensive knowledge workers had been added at both the head office and the plant—engineers, computer technicians, and managers—to take care of the sophisticated computer systems and state-of-the-art equipment. The new plant wasn’t quite the unalloyed good that it appeared at first sight. Variable costs of manual labor fell, but the fixed cost of knowledge workers rose, making it critical to keep capacity utilization high—which was possible in some years but not in others.

The bread company was representative of many businesses. They swapped direct costs for indirect costs, which meant fewer but more productive manual workers and greater numbers of more expensive knowledge workers. (See the exhibit “The Rising Share of Knowledge Work.”)

The Rising Share of Knowledge Work

In the half century since Peter Drucker coined the term “knowledge workers,” these employees have become not just an important part of the workforce but the dominant part. And as China and other low-cost jurisdictions bring more and more manual workers onstream, the developed economies will become ever more reliant on knowledge workers, whose productivity may therefore be themanagement challenge of our times.

Productivity in the Decision Factory

The two critical drivers of productivity in any production process are the way the work is structured and the company’s ability to capture the lessons of experience. These drivers are of course interdependent: How you structure the work influences your ability to learn from it. In decision factories, a mismatch between the reality of work and the way it is structured leads directly to inefficiencies in allocating knowledge work. People being people, this mismatch weakens incentives for sharing knowledge. Let’s look at why.

Work structure in the decision factory. The basic unit of labor in the decision factory is the job. In this respect, decision factories follow the product factory model, whereby managers typically identify a specific activity that makes up an individual’s job and needs to be repeated more or less daily. If you know how much output you want, you can estimate how many of these “jobs” you need and hire accordingly. Of course, output is always somewhat variable, and to the extent that this is predictable, you can build it into job contracts. Some employees work fewer or shorter shifts than others. But on the whole, the assumption implicit in this structure is that the output of product factory work is steady.

Decision factory jobs are based on the same assumption. The vice president of marketing, for example, is implicitly assumed to produce the same amount of output every day. So job descriptions are written as a collection of ongoing tasks that add up to one full-time job. In the prototypical VP of marketing job, the incumbent is responsible for product branding, promotional activities, market research, and so forth—all described as if they needed to be done day after day, week after week, and month after month.

But here the analogy between decision and product factories breaks down. Knowledge work actually comes primarily in the form of projects, not routine daily tasks. Knowledge workers, therefore, experience big swings between peaks and valleys of decision-making intensity. That VP of marketing will be busy during the launch of an important product or when a competitive threat arises—and really, really busy if the two overlap. Between these spells, however, she will have few or even no decisions to make, and she may have little to do but catch up on e-mails. Yet no one suggests that she vacation during these periods, let alone that the company should stop paying her salary.

Binge-and-purge cycles of hiring and firing knowledge workers are the unfortunate consequence of this approach to knowledge work. When entire workforces are organized around permanent, full-time jobs, it is difficult to redeploy resources to extremely busy areas to deal with peak demand. Typically, the HR department has to create a new position, write a job description, and then fill the position either by transferring someone from another full-time job or by running an external search.

All managers in all areas tend to staff for what they perceive as the peak demand for knowledge work in their area of responsibility. This institutionalizes a significant level of excess capacity spread in small increments throughout decision factories. That is why decision factory productivity is a persistent modern challenge.

Of course, it is most certainly not in the interest of knowledge workers to go to their bosses and declare that they have “spare capacity.” At best, they might then be judged in performance reviews as having an easy job and being not very productive. At worst, the bosses might decide that these employees could be cut. Thus it is to every knowledge worker’s benefit to look busy all the time. There is always a report to write, a memo to generate, a consultation to run, a new idea to explore. And it is in support of this perceived survival imperative that the second driver of productivity—knowledge transfer—gets perverted.

Knowledge in the decision factory. As I describe in my book The Design of Business,knowledge development goes through three stages. When a new manufacturing or service operation is created—for example, Intel’s first microprocessor chip fabrication facility, in 1983, or Disney’s first theme park, in Anaheim, California, in 1955—the task is a mystery. What is the optimal process flow through the fabrication plant? How should queues be structured at Disneyland? The pioneering experimental work tends to be inefficient and error-filled, as for any mystery.

In due course, with lots of practice, a body of wisdom is created—what can be called a heuristic—that guides how the process is carried out. Intel’s next dozen fabrication plants were no longer hit-or-miss, because knowledgeable masters who had worked on the first one designed them. And as Disney opened its four theme parks at Walt Disney World in Orlando, Florida, it was able to use the Anaheim heuristic.

In product factories the advancement of knowledge doesn’t stop with a heuristic. The culture at large-scale manufacturing and service operations is to keep pushing until the knowledge becomes an algorithm—a formula for guaranteed success. An operating manual replaces the knowledgeable master. Less experienced managers can use the algorithm to get the job done. That culture lies behind the success of McDonald’s, Wells Fargo, and FedEx. And the work doesn’t stop with the existing algorithm: It is honed and refined in a continuous improvement process.

In the decision factory, however, knowledge tends to remain stubbornly at the heuristic level, where experience and judgment are the key requirements for competent decision making. A big part of the explanation, of course, is that the knowledge challenge is simply tougher in decision factories. Many decisions have to be made for the first time and are thus in the mystery category. For example, how should a company enter Nigeria, its first developing market? And what about the next country? The proper entry strategy will be different there. Even after making entry decisions for 10 countries, the company may not have a heuristic, let alone an algorithm.

But the job-based work structure creates a significant hazard. If experienced knowledge workers turn a skill-based heuristic into an algorithm, they are inviting the company to replace them with lower-skilled, less expensive workers. That is why many organizations find it difficult to get master knowledge workers to spend time teaching apprentices even the heuristics: Something else always seems to be more pressing.

Of course, this hazard exists in the blue-collar world as well. But there, knowledge is advanced through the observation of physical processes. From the time of Frederick Winslow Taylor and his infernal stopwatch, blue-collar workers have understood that their work can and will be observed and optimized. The jobs of knowledge workers, however, are hidden between their ears.

Senior executives in modern corporations know that they have more knowledge workers than they need, but they don’t know where the excess lies. Thus when they face a dip in sales or some other tough patch, they reflexively chop knowledge workers, trusting that some portion of the excess will disappear without particularly negative consequences.

There is a better way to run the world’s expensive decision factories. It has two central attributes: It adopts the method that successful professional services firms use to manage their human resources, and it embraces the ethic of knowledge advancement found in the best blue-collar factories.

Redefining the Job Contract

The key to breaking the binge-and-purge cycle in knowledge work is to use the project rather than the job as the organizing principle. In this model, full-time employees are seen not as tethered to certain specified functions but as flowing to projects where their capabilities are needed. Companies can cut the numbers of knowledge workers they have on the payroll because they can move the ones they have around. The result is a lot less downtime and make-work.

Think of a freshly hired assistant brand manager for Olay at P&G. She may initially view her role as pretty standard: helping her boss guide the brand. However, she will quickly learn that the job is ever-changing. This month she may be working on the pricing and positioning of a brand extension. Two months later she may be totally absorbed in managing production glitches that are causing shipment delays on the biggest-selling item in the Olay lineup. Then all is quiet until the boss approaches her desk with yet another project. Within months she will figure out that her job is a series of projects that come and go, sometimes in convenient ways and sometimes not.

This characterization of knowledge work is gaining traction among management thinkers. In “The Rise of the Supertemp” (HBR May 2012), Jody Greenstone Miller and Matt Miller describe an emerging class of managers who are focused on short-term, high-value-added, knowledge-based projects. Similarly, the Silicon Valley legend Reid Hoffman, with Ben Casnocha and Chris Yeh, suggests in “Tours of Duty: The New Employer-Employee Compact” (HBR June 2013) that organizing knowledge workers’ employment into time-bound “tours of duty” can help companies retain these workers and keep them happy. And although actually organizing knowledge work around projects may seem a radical idea in mainstream business, it is very familiar to professional services firms, some of which have become as large as manufacturing corporations. In 25 years Accenture has grown from its inception as the “systems integration practice” of Arthur Andersen into an independent firm with revenue equivalent to 3M’s. The iconic consultancy McKinsey & Company has about as much revenue as a typical Fortune 500 company.

These companies are made up almost exclusively of knowledge workers. When a project comes in, a team is assembled to carry it out. When the project is finished, the team is disassembled and its members are put on other projects. They don’t have permanent assignments; they have established skill levels that qualify them to work in certain capacities on certain projects.

This ability to channel resources flexibly and seamlessly to projects as they arise enables these consulting firms to do something that their clients cannot—that is, to staff projects that the clients can’t handle themselves because the necessary personnel are lodged in permanent assignments. True, for some projects a professional services firm has unique expertise. But often its ability to flow bodies quickly to the task at hand is the reason it was hired. Indeed, professional services firms have grown so quickly in part because they are organized around projects, whereas their clients are organized around permanent jobs.

This approach is not limited to professional services firms. Hollywood studios, for example, have always organized around film projects. A team comes together to plan, shoot, edit, market, and distribute a film. As individual team members finish their tasks, they are assigned to other projects.

Mainstream companies are catching on as well. In 1998 P&G carried out a major operational reorganization. The centerpiece was a shift from four integrated regional profit centers to seven global business units (GBUs)—including baby care, fabric care, and beauty care—along with market development organizations that were responsible for distributing the products of all seven GBUs within their given regions.

A feature of the reorganization was the creation of Global Business Services in order to share information technology and employee services. Shared service organizations had become popular, so the fact that P&G took this step was not in itself notable. But how GBS operated was.

Under the leadership of Filippo Passerini, who is now the president of GBS, P&G in 2003 engaged in what was then the biggest outsourcing deal in corporate history: It sent approximately 3,300 jobs to IBM, HP, and Jones Lang LaSalle. Passerini transferred to those organizations the GBS employees who were performing the most-routine, least-project-oriented work. This allowed him to think more innovatively about the jobs that remained within GBS. The classic move would have been to structure them as flat jobs, assuming a consistent stream of similar work for each one.

Instead Passerini decided to embrace the project nature inherent in the work still at GBS. He made the part of his enterprise that remained within P&G what he called a “flow-to-the-work organization.” Of course, some of his employees were still working in flat, permanent jobs, but a large proportion were assigned to whatever projects had high urgency and high payoff. These knowledge workers didn’t expect to stay in one business unit in one region; they understood that they would be working in teams organized specifically to tackle pressing assignments in succession.

The integration of Gillette was one such assignment. The 2005 acquisition of Gillette was by far the biggest P&G had ever made, adding 30,000 employees and costing $57 billion. The most challenging aspect lay in the GBS area: integrating all the back-office functions—finance, sales, logistics, manufacturing, marketing—and information technology systems. Thanks to GBS’s flow-to-the-work structure, Passerini could quickly channel extensive resources to the integration. As a result, it was accomplished in a mere 15 months—less than half the time normally required for an acquisition of this size. With synergy savings from integration estimated at $4 million a day, this translated into a saving of close to $2 billion.

The project-based approach to knowledge work is currently being rolled out across P&G. In 2012 the company announced an initiative to eliminate excess white-collar costs and manage the remaining costs more effectively. Each part of the P&G organization is defining what proportion of its knowledge workforce should be in permanent, flat jobs and what proportion should be in flow-to-the-work jobs. The flow proportion may vary by unit, but it is required to be greater than zero.

Toward the Knowledge Algorithm

Switching to a flow-to-the-work structure will do much to improve the productivity of a company’s knowledge workers and to remove obstacles to codifying and transferring knowledge. But it will not guarantee that the codification and transfer actually take place.

For that to happen, knowledge workers must be persuaded to go the extra mile. P&G has become a leader in this respect as well, putting key executives in charge of codifying its knowledge. Since 1837 the company has been a model brand builder, but for a long time it left brand building as a heuristic to be developed in the heads of experienced and expensive executives. Learning the heuristic traditionally involved apprenticing with one or more of them to slowly absorb the unwritten rules.

P&G eventually decided that this approach was no longer acceptable. In 1999 Deborah Henretta, then the general manager for fabric conditioners, sponsored a project to codify the company’s brand-building heuristic—and thereby move it in the direction of an algorithm. The brand-building framework (dubbed BBF 1.0) was intended to enable young marketers in the organization to learn the techniques for brand building more quickly, thus lowering the time and cost required for this task. BBF was found sufficiently valuable to be further refined, producing BBF 2.0 (2003), BBF 3.0 (2006), and BBF 4.0 (2012).

GBS has actively moved in the same direction. An example of its efforts involves the labor-intensive preparatory work that finance and accounting managers in each of 20-plus categories across P&G carried out in advance of the annual strategic-planning exercise. Traditionally, a manager would rely on experience to determine what sorts of information would be helpful to the category team in preparing for the strategy work, would collect that information from a variety of sources, and would organize it in some form.

GBS, whose information systems were tapped to provide much of the data, noticed a pattern of requests for certain kinds of data at a certain time of year by certain kinds of managers. In due course it ascertained that the preparatory materials of all these F&A managers were very similar in content and could easily be assembled by GBS from an algorithm; in fact, most of them could be assembled and spit out by a piece of software that GBS had built. Rather than spend hundreds of hours putting together a data package, each manager could simply e-mail GBS and ask for a preparatory deck for the upcoming strategy process.

Obviously, not all knowledge work can be reduced to algorithms. But it’s possible to go further than most corporations realize. Philip Parker, a marketing professor at Insead, has strikingly demonstrated how much the envelope can be pushed. He developed and patented an algorithm that enables a computer to write a research report on almost any topic, drawing on databases and automatic internet searches. Parker reports sales of more than 200,000 copies of the books his programs have authored. (More recently he has been trying out algorithms for poetry and fiction.)

No organization the size of P&G can become project-based overnight or reduce every heuristic to an algorithm. Nor should it; that would be overkill and very disruptive. But a company with 100% flat jobs is almost certainly obsolete. Likewise, knowledge in the modern corporation can be advanced only so fast, and a large share of employees will continue to be invested in running current heuristics. But some people are clearly needed to solve the next new mystery. The key is to invest significant knowledge worker resources in projects that move knowledge forward. Only then can organizations avoid cycles of binge and purge while improving the productivity of their knowledge workers.

Source: Harvard Business School (http://goo.gl/vRASXQ)