No Excuses Management

If everyone in our company made ordinary business decisions in a commonsense way, we would be unstoppable. It turns out that very few people, at our company or anywhere else, make ordinary business decisions in a commonsense way. Most companies don’t fail for lack of talent or strategic vision. They fail for lack of execution—the mundane blocking and tackling that the great companies consistently do well and strive to do better. At Cypress, our management systems track corporate, departmental, and individual performance so regularly and in such detail that no manager, including me, can plausibly claim to be in the dark about critical problems. Our systems give managers the capacity to monitor what’s happening at all levels of the organization, to anticipate problems or conflicts, to intervene when appropriate, and to identify best practices—without creating layers of bureaucracy that bog down decisions and sap morale. Lots of companies espouse a “no surprises” philosophy. At Cypress, “no surprises” is a way of life. We operate in a treacherous and unforgiving business. An integrated circuit is the end result of a thousand multidisciplinary tasks; doing 999 of them right guarantees failure, not success. Last year, we shipped 159 different chips (56 of which were new in 1989) in 7 distinct product categories using 26 different process-technology variations. This year, we plan to add another 50 products to our portfolio. Our watchwords are discipline, accountability, and relentless attention to detail—at every level of the organization. How do we measure success at Cypress? By doing what we say we are going to do. We meet sales projections within a percentage point or two every quarter. We don’t go over budget—ever. Our silicon wafer manufacturing plant in Round Rock, Texas, which now accounts for 70 % of our sales, shipped its first revenue wafer eight months after ground breaking. That performance tied the industry record, which had been set in 1984 by our manufacturing facility in San Jose. We meet sales projections within a percentage point or two. We don’t go over budget—ever. Some who learn about our systems raise the specter of Big Brother or use terms like “electronic treadmill.” This is nonsense. Our management systems are not designed to punish or pressure; people put enough pressure on themselves without any help from me. The systems are designed to encourage collective thinking and to force each of us to face reality every day. At Cypress, we collect information in such detail and share it so widely that the company is virtually transparent. This works against the political infighting and bureaucratic obfuscation that cripple so many organizations. People at Cypress disagree vigorously on many things; in our management meetings, people of any rank are free (and eager) to challenge senior executives. But we disagree over substance and the best ways to make the hardheaded trade-offs that are the essence of management. Nobody gets ahead by using information selectively to win internal battles. Technological change drives the semiconductor industry, and several of our management systems help us make decisions about technology assessment and product development that are quite specific to our business. But most of our management systems (and certainly the most important ones) address the universal challenges of business—motivating people to perform effectively and allocating resources productively. To succeed over the long term, we have to do at least four things better than our competitors: 1. Hire outstanding people and hold on to them. 2. Encourage everyone in the organization to set challenging goals and meet them. 3. Allocate key resources (people, capital, operating expenses) so as to maximize productivity. 4. Reward people in ways that encourage superior performance rather than demotivate superior performers. Before I describe several of our most important systems, let me make three points. First, I don’t use the word systems loosely. Each one involves holding specific meetings, maintaining databases, submitting specific reports, and securing approvals. For example, managers grant raises by using a software package that takes them step-by-step through the required performance-evaluation procedure, recommends salary increases that they can modify within limits, and alerts them if their final decisions violate company policy. A few of our systems virtually run themselves. We’ve developed a set of computer applications that automatically shuts down a manufacturing operation if it detects that performance has violated a critical procedure. Dubbed “killer software” by our MIS group, these applications have inspired dramatic performance improvements in our factories, and we are transferring the concept to administrative functions such as order entry and accounts receivable. Our “killer software” automatically shuts down a manufacturing operation if it violates critical procedures. Second, we don’t confuse systems with bureaucratic planning. In our business (and, I suspect, most businesses), five-year plans, even one-year plans, are obsolete 60 seconds after they are written. So we have plans, but we don’t become enslaved by them at an unrealistic level of precision. Our annual marketing forecasts are probably 95 % accurate, with respect to total revenues for each of our product lines. For groups of products within each line, the plans may be 70 % accurate. For individual products within groups, it’s not that unusual to be off by 100 % . At the same time, we are absolute sticklers about meeting quarterly revenue and profit targets. Precisely because we understand that detailed long-term forecasts are so unreliable, we track product shipments and revenues on a daily basis, evaluate how they measure up against the plan, and identify what adjustments we have to make to meet the plan in the aggregate. We worry less about meeting product-by-product forecasts than about reacting instantaneously to unforeseen competitive developments. Finally, top management cannot manage without a thorough mastery of the details of its business. To my mind, no CEO can claim to be in charge of the organization unless within 15 minutes—and I mean this literally—he or she can answer the following questions. What are the company’s revenues per employee? How do the figures compare with the competition’s? What are the revenue-per-employee figures for each of the company’s leading product lines? What explains recent trends in each line? What is the average outgoing quality level in each product line? How many orders are delinquent? Which of the company’s top 20 executives are standouts, which are low performers, and why? Which departments could recover from a major competitive shock, and which are vulnerable to change? What are the yields, costs, and cycle times at every manufacturing operation? What explains the company’s stock market valuation relative to its competitors’? By now, I suspect, readers are ready to protest: Isn’t this a formula for “micromanagement” by the CEO? Won’t top executives get lost in the thicket of details and lose sight of broader strategic imperatives? Being in command of detail doesn’t mean interfering where you don’t belong. Collecting information, reviewing it regularly, and sharing it widely allows me to practice management by exception in the truest sense. So long as we stick to our systems, this organization virtually runs itself. I intervene only to solve problems and champion urgent projects. How to Hire Outstanding People Great people alone don’t guarantee corporate success—but no company can succeed without them. Sounds like a truism, right? Yet how many companies are as scientific about hiring as they are about designing new products or perfecting the latest market-research techniques? Hiring is one of the most bureaucratic, passive, and arbitrary parts of corporate life. The day we founded Cypress we understood that our greatest proprietary advantage would be to do a better job of hiring than the big companies against which we would compete. We think we’ve sustained that advantage. Cypress now employs more than 1,400 people. Our philosophy of hiring (and the system that puts the philosophy into practice) hasn’t changed much since we founded the company in 1982. I know the entire organization is implementing the system because no one can bring in a new employee, no matter the rank, without submitting a “hiring book” that documents the entire process and gives comprehensive results of interviews and reference checks. Until we reached 400 employees, I read every one of those books before we made a job offer. I now share that task with six other senior managers. In no more than 15 minutes, with no verbal communication whatsoever, one of us can determine whether or not a hiring manager has followed our procedures. When the system has been followed, we make the hire. When it hasn’t, we don’t. Recently, for example, one of our vice presidents submitted a hiring book that violated a cardinal rule not to offer big raises to job candidates. His explanation was familiar: “At my old company, we had to give big raises to convince people to sign on.” My response was unequivocal: “We’ve hired 1,400 people with this system. Is there any reason we can’t hire number 1,401 the same way?” The vice president resubmitted the hiring book two weeks later, and it still contained an out-of-policy raise. To emphasize my point, I tracked down a big pair of scissors, cut the book in half, and mailed it back to him. He now does it right the first time. The hiring book serves another purpose. When people don’t work out, it helps us go back and evaluate what went wrong. Hiring the wrong people is a very expensive mistake. We give no raises to about 3 % of our employees every year, a sign that they are low performers. This company would die if 3 % of the products we shipped were defective. We strive for 100 % quality in manufacturing; we have the same goal for hiring. In many cases, when an employee leaves because of poor performance, we send his or her manager the hiring book, ask how we might have spotted the problem earlier, and use the departure as an opportunity to improve our hiring procedures. Three strongly held beliefs drive our hiring system. First, the only way to hire outstanding people is for managers themselves to find the people they need. Hiring, like most everything else in business, is an acquired skill. The more people you interview, evaluate, and select, the better you get at interviewing, evaluating, and selecting. Thus if 1 of my 12 direct reports (8 vice presidents and 4 subsidiary presidents) is authorized to hire a new manager, or 1 of our 72 managers is authorized to hire an individual contributor, it is up to that manager, not the human resource staff, to locate desirable candidates, prescreen them before they enter the formal evaluation process, and monitor the process as it unfolds. We make almost no use of executive recruiters. The human resource department plays a modest role in suggesting candidates but no role in evaluating them. Indeed, prospective employees don’t see anyone in human resources until they report to work and fill out their insurance applications. The only way to hire great people is for managers—not personnel or headhunters—to find the people they need. Most companies—and certainly most big companies—do just the opposite. Managers sit behind their desks and wait for personnel to parade candidates through their office. Of course, personnel is never as motivated as the hiring manager is to fill an open slot. As the hiring schedule falls behind, the manager grows increasingly desperate and makes an offer to the first warm body that meets rudimentary requirements. This approach guarantees that the quality of the company’s work force will nicely (and disastrously) mirror the quality of the available labor pool. The organization drifts toward average. In our system, managers have only themselves to blame if they can’t fill open slots. After eight years, we have a good feel for the employment “hit rate.” It takes ten prescreen telephone interviews to find one person good enough to qualify for the formal evaluation process. We make offers to about 25 % of the people who go through our evaluation, and about 85 % of the people to whom we make offers accept. If our managers do their math, they quickly understand that quality hiring takes a lot of work. They’ll need to make about 100 telephone calls to find two great people. By the way, if they don’t fill their open slots by the end of the quarter, they lose the hiring requisition and have to justify the new position again. Another weakness with most companies’ passive approach to hiring is that executives lose touch with the job market. Regular interaction with the job market provides insights on marketplace dynamics. A few years ago, in a memo explaining our hiring system to a new top manager, I wrote down what I had learned as a result of my own job-related calls and interviews. For example, people didn’t like Intel’s 8 a.m. sign-in policy, which the company has since eliminated. Intel and National Semiconductor were paying 15 % premiums to layout designers on the night shift while Cypress paid none. But most layout designers cared less about shift differentials than about the kinds of chips they were designing. The market value of plasma engineers ranged from $ 20,000 a year to $ 50,000 a year, with little variation of price with quality. The list could have gone on for several pages. We have a second inviolate principle of hiring: We don’t buy employees. This is uncommon in Silicon Valley, but it serves as an excellent screen. Someone who will join Cypress for a few percentage points more in salary or a better dental plan is not the kind of career-oriented person we want. Good people want to be paid fairly, in a way that reflects the organization’s success over the long term. They are rarely concerned with earning an excessive amount of money relative to their peers. What drives them is the desire to win. Until 1986, therefore, we had a very simple policy: A candidate who had received a raise from a previous employer within the last four months came to Cypress without a raise. We now offer an 8 % prorated increase, but the basic premise holds. We do not bid for great talent; people with great talent come to Cypress because they want to win. Members of our team are rewarded with stock options and the highest percentage raises in the industry. Everyone at the company, from the receptionists to the CEO, participates in an aggressive stock-option program. (We recently repurchased two million shares to fund that option program.) As for raises, our formal policy is to survey the competition, calculate the average of the three companies that grant the highest raises and add one-half a percentage point or more to determine the Cypress raise budget. We have a third principle: The interview and evaluation process should be tough, fair, and expeditious. Quality interviewing is the most important part of the hiring process—and something most managers are terrible at. We’ve developed several interview techniques to keep these sessions productive. (See the insert “The Science of Interviewing.”) Although the interview and evaluation process is demanding, it need not take forever. There’s no reason a manager can’t prescreen a candidate by telephone, bring him or her in for two rounds of interviews, and make an offer within one week. We’re not there yet; we meet the one-week target about 75 % of the time. We also expect the hiring manager to be in daily contact with the candidate as the process unfolds. The Science of Interviewing You can’t hire quality people without a systematic approach to interviewing. Four basic rules guide our interview and evaluation process. 1. Use the big guns. If you want job prospects to know that you are serious, high-ranking executives should take the time to get involved in the interview process. At Cypress, all candidates for exempt positions, whether or not they are senior enough to report to vice presidents, interview with two vice presidents. I interview all candidates who would report to vice presidents as well as many important individual contributors who report to managers. This is the technique Jerry Sanders used on me when I was considering joining Advanced Micro Devices. The first day I walked in the door, the receptionist in a building of 2,000 people smiled and said, “Oh yes, Dr. Rodgers, Jerry is waiting for you.” She took me right upstairs—no waiting in the lobby—and our session began. That’s the way to communicate to job prospects how valuable they are. 2. Make interviews tough and technically demanding—even for people you know you want. We are a hard-charging company in a tough business and have a no-nonsense way of communicating. People should know that before they sign on. At the beginning of the evaluation process, candidates receive a form that lists the technical skills the position requires, with whom they’ll be interviewing, and the questions they will be asked. This focuses the interviews and alerts the candidates to how rigorous the sessions will be. After several “technical” interviews and the interviews with two vice presidents, candidates go through what we call the “pack of wolves” session. The applicant sits in a conference room where senior technical people pose difficult questions that the candidate does not know in advance. Some of the questions are virtually impossible to answer, and they come in rapid succession. When the candidate makes a mistake, the interviewers point it out and give the correct answer. The tone of this session is aggressive but not abusive. It is an excellent way to weed out qualified managers and engineers who can’t take the pressures of our business. 3. Interviews should lead to detailed assessments of strengths and weaknesses, not vague impressions. Our interview evaluation forms include numerical scores (on a scale of zero to five) that mirror the technical qualifications on the requisition. Our people are tough graders. I also insist that the hiring vice president write an interview strategy before my session with a managerial candidate. That strategy highlights the specific strengths and weaknesses of the candidate, particular concerns he or she has expressed about the job or the company, and other critical issues. I’m not reluctant to share the numerical evaluations with a candidate, especially one who has been assessed as weak in a particular area. This candid feedback is usually a positive experience. Good people know their weaknesses and are eager to improve them. 4. Check for cultural fit. Most companies claim to do this, but few are very systematic. We probe work attitudes and career goals through a questionnaire that requires brief but direct answers to open-ended questions. The questionnaire forces candidates to be as specific as possible about hard-to-quantify issues that are addressed only obliquely, if at all, in most evaluation processes. Among the questions are: How is the morale in your company or department? Why? What do you expect Cypress has to offer you in the way of a work environment that your employer doesn’t offer? What would your boss say is your best attribute? What would the “needs improvement” section of your performance review address? Can you describe your personal experience with a difficult boss, peer, or subordinate? Set Goals—Then Measure and Meet Them All of Cypress’s 1,400 employees have goals, which, in theory, makes them no different from employees at most other companies. What does make our people different is that every week they set their own goals, commit to achieving them by a specific date, enter them into a database, and report whether or not they completed prior goals. Cypress’s computerized goal system is an important part of our managerial infrastructure. It is a detailed guide to the future and an objective record of the past. In any given week, some 6,000 goals in the database come due. Our ability to meet those goals ultimately determines our success or failure. Most of the work in our company is organized by project rather than along strict functional lines. Members of a project team may be (and usually are) from different parts of the organization. Project managers need not be (and often aren’t) the highest ranking member of the group. Likewise, the goal system is organized by project and function. In Monday project meetings, employees set short-term goals and rank them in priority order. Short-term goals take from one to six weeks to complete, and different employees have different numbers of goals. At the beginning of a typical week, for example, a member of our production-control staff initiated seven new goals in connection with three different projects. He said he would, among other things, report on progress with certain minicomputer problems (two weeks), monitor and report on quality rejection rates for certain products (three weeks), update killer software for the assembly department (two weeks), and assist a marketing executive with a forecasting software enhancement (four weeks). On Monday night, the project goals are fed back into a central computer. On Tuesday mornings, functional managers receive a printout of their direct reports’ new and pending project goals. These printouts are the basis of Tuesday afternoon meetings in which managers work with their people to anticipate overload and conflicting goals, sort out priorities, organize work, and make mutual commitments about what’s going to get done. This is a critical step. The failure mode in our company (and I suspect in most growing companies) is that people overcommit themselves rather than establish unchallenging goals. By 5 p.m. Tuesday, the revised schedule is fed back into the central database. This “two pass” system generates the work program that coordinates the mostly self-imposed activities of every Cypress employee. It allows the organization to be project driven, which helps us emphasize speed and agility, as well as functionally accurate, which works against burnout and failure to execute. On Wednesday morning, our eight vice presidents receive goal printouts for their people and the people below them—another conflict-resolution mechanism. In the early days of the company, until we hit about 100 employees, I read every employee’s goals every week. I knew what every person in the company was doing on a week-to-week basis. That’s not the role of the CEO in a company with more than 1,400 people, so my approach to the goal system has changed. Today my job is to anticipate problems, largely by sorting through the goal system looking for patterns. I use the system as a kind of organizational speedometer that not only tells me how fast we’re traveling but also helps explain what’s holding us back. On Wednesday afternoons at my weekly staff meeting, I review various database reports with my vice presidents. We talk about what’s going wrong and how to help managers who are running into problems. The following reports typically serve as the basis for discussion: progress with goals on critical projects; percentage of delinquent goals sorted by managers (their goals plus those of their subordinates); percentage of delinquent pending goals that are delinquent for all people reporting up the chain of command to each vice president); all employees without goals (something I do not tolerate); all goals five or more weeks delinquent; and all employees with two or more delinquent goals, sorted by manager. As we’ve refined the goal system and used it more extensively, I’ve developed some general principles. First, people are going to have goals they don’t achieve on time; the key is to sense when a vice president or a manager is losing control of the operation. My rule of thumb is that vice presidents should not have delinquency rates above 20 % , and managers should not let more than 30 % of their goals become delinquent. When managers do have a delinquency problem, I usually intervene with a short note: “Your delinquency rate is running at 35 % , what can I do to help?” I often get back requests for specific assistance. Part of my role is to hold people accountable. But it is also to identify problems before they become crises and to provide help in getting them fixed. Second, people need positive feedback. Every month we issue a Completed Goal Report for every person in the company. The report lists all goals completed over the past four weeks as well as those that have yet to come due. Individual Monthly Goal Report This excerpt lists all the goals one production-control staffer completed in workweek 45. The entire report consisted of 49 completed and pending goals. The completed goal report is also a valuable tool for performance evaluation. Like most companies, we use annual reviews to set salary increases. The trouble with annual reviews is that managers succumb to the “proximity effect.” An employee who performs outstandingly for the first ten months of the year but has a subpar two months just before the review is more likely to get a poor evaluation than is a colleague who had a lousy ten months but did a great job in the two months just before the evaluation. At Cypress, the completed goal report triggers a performance minireview; every month managers read through their people’s printouts and prepare brief, factual evaluations. At year end, managers have a dozen such objective reviews to refresh their memories and fight the proximity effect. It’s important to note that the goal system does not require big investments in computer hardware and software. Indeed, we have no equipment dedicated exclusively to the goal system. We use personal computers and a DEC minicomputer that also handles the company’s administrative tasks. The software is Lotus 1-2-3 and Paradox—two of the most common microcomputer applications. Much of the system still runs on “sneakernet”—people pass disks back and forth rather than use networking systems, although we are in the process of networking the entire company more effectively. The goal system doesn’t require big computer investments. I used to manage it with a blackboard. Indeed, I developed the goal system long before personal computers existed. It has its roots in management-by-objectives techniques I learned in the mid-1970s at American Microsystems (AMI), where I ran the random access memory (RAM) group.1 Back then my hardware was a blackboard, and my software was chalk. In a typical year, we would develop as many as ten new chips, and tracking each project became a nightmare. So I got a blackboard, wrote down everything that had to be done before each product could ship, and attached names and due dates to each task. My update routine was an eraser. After I left for Advanced Micro Devices (AMD), where I was responsible for RAM manufacturing as well as design, it became a bit harder to keep score. So I covered my walls with blackboards, used masking tape to divide them into project panels, and transported the system from AMI. Once a week, my assistant would wheel in an electric typewriter, record the schedule, and send it out to the troops. If people in my organization wanted to know where things stood, all they had to do was come into my office and look at the blackboards. The Cypress system is just an enhanced, electronic version of what I’ve been doing for 15 years. The computer record on each goal includes a description of the task, when the goal was set, when it is due, what priority it is, who has agreed to complete it, to what manager that person reports, and to what vice president the manager reports. We are now enhancing the system by recording long-term goals (which we call “strategic” goals) and all quality-oriented goals. Of course, recording that information means I can sort by each of those fields and create special-exception reports. With a few keystrokes, I can check on the performance of any one of my vice presidents, see how a manager is relating to subordinates, or check on the progress of a particular project. No one can accuse me of not “knowing the details” of my business. Yet I don’t have to intervene where I don’t belong or try to be in a hundred places at once. The goal system provides warnings when something goes wrong and offers instant access to data in any area that I am concerned about. I’m a big advocate of management by walking around, and I regularly block out time on my calendar to visit our facilities. The goal system lets me practice MBWA all the more effectively. By sorting through the database and following up with telephone calls, I can get up to speed on an operation before I arrive. I don’t want to give the impression that the goal system is strictly a support mechanism. It does give me the ammunition I need to cut through bureaucratic obfuscation. My access to the details means vice presidents and managers know they can’t snow me. For example, we recently shafted a valued customer by delivering an emergency shipment of parts one day late. (I learned about the problem because every officer at Cypress is “godfather” to one strategic customer and I happened to be godfather to this customer.) I spent 15 minutes at my computer screen reviewing relevant goal reports and data from a few other systems. With a follow-up telephone call or two, I discovered what went wrong. Shipping was overloaded on the day the parts were scheduled to go out; the parts were not marked as a “JIT order,” so they weren’t shipped immediately. By the time the department shipped the parts, they were a day late. With another call, I learned that certain people in marketing did not understand the shipping department’s priority system. I called the relevant people into my office and asked them to change our procedures so that the problem would never happen again. Now think what might have happened without the goal system. After I learned about the problem, I would have called in the marketing manager, who would have made plausible excuses, produced nice-looking graphs showing a 99 % on-time delivery performance—and blamed the jerks in shipping. Then I’d have called shipping, a manager there would have complained about having been overloaded, about how marketing always expects miracles, and the political infighting would have been hot and heavy. I would have been virtually powerless to cut through to the truth. People at Cypress know I don’t tolerate bureaucratic politics—it’s one of the items on a very short list of what can get you fired. They also know I have the ability to peer down into the bowels of the organization, drag up relevant data, make a few telephone calls, and find out what really happened in any situation. The knowledge that I can call on so much detail so easily means I seldom have to. No vice president or manager wants to be in the awkward position where the CEO knows more about a situation than he or she does. Allocate Resources for Maximum Productivity Middle managers can be an organization’s most enduring strength. They are more aware of the company’s day-to-day business realities than any other group, and they are earnest, committed, and creative. Middle managers can also cause companies to grow fat and uncompetitive, not because they don’t do their jobs, but because they think their jobs are the most important in the world and thus lose sight of the broader corporate imperative. I call this phenomenon middle management myopia. In organizations that suffer from this disease (and it afflicts the majority of large companies), middle managers clamor for resources while top managers are chartered to hold the line. Usually, top managers are forced to cave in because middle managers can call on so much more information and functional expertise. How can a senior executive turn down a request for resources (people, equipment, expenses) when a well-respected middle manager makes a plausible argument that the department will unravel without them—probably taking the company with it? These “gun-to-the-head” stories are a staple of corporate life. If you don’t think any good middle manager can come up with several of them, you are fooling yourself. Consider just two: From a financial manager: “We have $ 40 million of excess receivables on which we pay $ 140,000 of interest every year. I could get the receivables down to $ 20 million, but I need another clerk to do it. The basic problem is paperwork; sometimes our invoices don’t match our customers’ records, and we have to work the phones to get them to agree to pay. I know I told you seven clerks was enough. But if we hire one more at $ 30,000 a year, we can cut the uncollected receivables in half, saving us $ 70,000. It’s foolish not to make the hire.” From the engineering manager: “We generate annual revenues of $ 20 million with this particular chip. Our final-test yield is 90 % , which means we’re throwing away $ 2 million a year. If you let me hire two more engineers (they will cost us less than $ 100,000 a year combined), we can get yields up to 94 % , saving $ 400,000—a four-month payback. It’s foolish not to make the hires.” The moment senior executives buy into the tunnel vision of their middle managers, they’ve lost control of the company. If that happens, it’s not the middle managers’ fault; they’re simply doing their jobs as they understand them. Our resource-allocation systems are designed to prevent middle management myopia by forcing collective thinking at all levels. Every quarter, based on our revenue projections, we establish total corporate allowances for new hires, capital investment, and operating expenses. We then allocate this overall target by departments and force critical trade-off decisions down to the middle management level. We get these people into the same room at the same time, with a computer model that allows us to conduct resource negotiations on-line, and cut the deals that would otherwise lead to memo wars and special pleading. Our resource-allocation systems also address a second source of waste and inefficiency: the false confidence created by prosperity. The seeds of business failure are sewn in good times, not bad. Economic reversals have a wonderful way of concentrating minds and encouraging groups to overcome long-standing differences. During times of prosperity, however, danger lurks everywhere. Growth masks waste, extravagance, and inefficiency. The moment growth slows, the accumulated sins of the past are revealed all the way to the bottom line. Prosperity is dangerous. Growth masks waste, extravagance, and inefficiency. At Cypress, our annual revenues have grown 40 % or more per year since we founded the company in 1982. We’re proud of that record, but we’re even more proud of our consistent quarter-to-quarter profitability. Even in 1986, when the semiconductor industry was in the midst of a crippling downturn, we maintained industry-leading operating margins of 20 % . And we did it on revenues of only $ 50 million. This was no accident. The performance measures that drive the company adjust our thinking to control the prosperity illusion. Sure we want aggressive revenue growth. But we demand ever-increasing revenue per employee, ever-higher capital productivity (revenue dollars per dollar of undepreciated fixed assets), and ever-lower expense ratios. And we constantly check our performance against the best of our competition. The graphs in “Tracking Performance at Cypress” review our record on these measures. Tracking Performance at Cypress Our system for controlling head count is one of the three most important systems we have at Cypress. Here’s how it works. At the beginning of every quarter, we hold a head-count meeting for the entire company. Our eight vice presidents attend the meeting along with middle managers. Together we review the last quarter’s total revenue, total head count, and revenue per employee. We then use forecast revenues to determine the allowable head count at the end of the quarter and thus the total number of requisitions we can afford to open in the new quarter. There are two things to keep in mind. First, we are almost always within 1 % or 2 % of forecast revenue—this is not a hypothetical exercise. Second, our target head count is based on continual improvement in revenue per employee. We expect this critical productivity measure to improve every quarter or, in difficult economic times, to hold the line.2 Thus, quickly and directly, everyone in the room (which means virtually everyone in the company with hiring authority) understands how many new employees will join Cypress over the next three months. This is not a negotiable figure. The negotiations start once we total up the vice presidents’ head-count requests and start adjusting them to meet the new total. The power of our system is that it presents the real trade-offs to everyone at the same time. Everyone in the room understands that we are playing a zero-sum game. A vice president who insists on all his or her requisitions understands (as does everyone else) that these slots come at the expense of other requests. Such stark trade-offs create incentives for everyone else to think creatively about solving that vice president’s problems without hiring more people. In other words, I enlist many allies in my effort to hold the line on resources. Moreover, vice presidents who can’t make the number of hires they want don’t feel arbitrarily denied. Even if they’re unhappy, they understand the logic behind the final decisions. The negotiations are conducted in great detail—they’re not for the faint of heart. Every department has a microperformance index that divides some useful measure of output by head count. These indexes, which we flash on the screen for everyone to see, become the basis of the trade-offs between departments. Let’s say shipping and receiving has requested an additional clerk. We flash a graph that tracks the department’s productivity index—line items shipped per week per clerk—and ask the manager and vice president to justify the request. Or say the vice president of manufacturing needs ten more operators in our San Jose fab. The group reviews weekly wafer output per employee per week, builds in further progress on this performance index, and weighs the vice president’s request accordingly. In each negotiation, the group knows the number of requisitions we must eliminate to meet the corporate target, the productivity trends for each department making requests, and which departments won’t be able to hire new people if other departments make all the hires they’ve requested. Our system for allocating capital works much the same way. We track total revenues as a function of undepreciated capital, insist on continual improvement in the macroindex, approve overall corporate targets on this basis, and use the targets to drive capital spending down through the departments. In our capital-intensive business, having just enough capital just in time is even more important than controlling head count. Our expense-control system goes one step farther. The CEO of any startup company signs every purchase order for the first few years. This is a deadly efficient way to control spending. But there comes a time (at about $ 40 million of annual revenues for a semiconductor company) when the CEO faces a moment of truth: he or she can spend every waking minute signing purchase orders or develop a new approach. This is where the extravagance usually begins or where a new purchasing bureaucracy chokes off necessary expenditures and creates frustration throughout the company. We’ve developed a system that gives senior management ironclad control of spending and that requires no more than five working days for a purchase order to be approved or sent back to the requester for cause. As with head count and capital, allowable expenses in a given quarter start from the top line. We project revenues, insist on improvement in the corporate expense-to-revenue ratio, and then allocate allowable purchases by function and department through real-time negotiation and adjustment. Then comes a second critical step. We divide each manager’s quarterly expenses by 13 and create a weekly budget that governs spending for the quarter. Of course, some spending comes in clumps; a manager may need to make one purchase that exceeds the weekly budget. That’s allowed—so long as the manager has accrued enough “credit” in previous weeks so that the big purchase does not bump overall spending for the quarter off track. This system is rigorous. It also works. If senior management controls the quarterly spending meter, and the meter determines how fast money flows into the corporate pipeline each week, we can’t be surprised by what comes out at the end of 13 weeks. Controlling the meter does not mean I abdicate my role in monitoring expenses. With few “emergency P.O.” exceptions, the spending window at Cypress is open for two hours a week. Every Thursday from 8 a.m. to 10 a.m. , we hold a purchase-order commit meeting with the vice presidents. They summarize the purchase orders that need my approval that week; I look for questionable expenditures and probe for extravagance. Even more important, the weekly meetings become a way of transferring best practices on spending throughout the organization. I invariably come out of them full of ideas about cost reduction that I immediately share with people who can benefit from them. A visitor unfamiliar with our system could sit in on a purchase-order commit meeting and conclude that the CEO of this $ 200 million company still signs every purchase order above the petty cash limit. That visitor would be correct in the literal sense but dead wrong about how the system really works. Indeed, our approach to expense control has cascaded down through all four levels of the company, to a point where I don’t worry about missing a purchase-order commit meeting. (I miss one out of three.) Vice presidents routinely take their weekly expense budgets and divide them appropriately among their managers to guide weekly budgets. In turn, managers take their weekly expenses and divide them appropriately among their subordinates, who take their weekly budgets and divide them among their contributors. Today our expense-control system virtually runs itself. It drives us every week of every month of every quarter. It’s no accident we have never exceeded a quarterly budget by more than 1 % . Take the Mystery out of Performance Appraisal Managers shouldn’t expect outstanding performance unless they’re prepared to reward outstanding performers. Yet evaluation and reward systems remain an organizational black hole for three reasons. First, managers aren’t very scientific about rating their people. They may be able to identify the real stars and the worst laggards, but the vast majority of people (who must still be ranked) get lost somewhere in the middle. Second, even if they evaluate people correctly, managers like to spread raises around evenly to keep the troops happy. This is a deadly policy that saps the morale of standouts who deserve more and sends the wrong signal to weak performers. Third, managers are totally incapable of distinguishing between “merit” and “equity” when awarding increases. Merit refers to that portion of a raise awarded for the quality of past performance. Equity refers to adjustments in that raise to more closely align salaries of equally ranked peers. Merit and equity both have a place in the incentive mix, but confusing the two makes for mushy logic, counterproductive results, and dissatisfied people. Cypress is not immune to these problems; I’d say a third of our vice presidents still don’t have a good feel for how to review performance and award raises. But I’m not too concerned about that. Our focal-review system, one of the most efficient systems we’ve ever developed, controls the three big problems I just identified. Our managers follow the system because the only way to grant raises in the company is to use the series of computer templates that walks managers through a series of review-process steps and alerts them to violations against company policy. As with all our resource-allocation systems, the focal-review system starts with policies at the top and forces middle management decisions to be consistent with that thinking. Senior management and the board of directors review our annual revenue forecasts, survey compensation trends among our competitors, and settle on a total corporate allowance for raises. The “raise budget” is not negotiable, and it drives raises throughout the company. If the corporate budget is 8 % , then every department must meet a weighted-average salary increase of 8 % . It’s up to managers to distribute the 8 % pool, which is where the focal-review system comes in. There are four stages to the process. Everyone at Cypress is part of a “focal group” (there are 132 groups in the company) composed of peers with comparable responsibilities. For example, all vice presidents are in one focal group, all weekend night-shift assembly operators are in another, all RAM circuit designers are in a third. The ranking committee for each focal group includes the members’ supervisors as well as managers from other parts of the company in a position to judge each member’s performance and service. The ranking committee for our shipping clerks might include representatives from sales or manufacturing. The ranking committee for a group of circuit designers might include representatives from marketing or manufacturing. Before evaluations take place, we post rosters of focal groups and ranking committees to allow people to question their assignments. Much of the discussion in the ranking committee centers on the monthly reviews stemming from the goal system. This controls the proximity effect I’ve already discussed and provides an objective record of performance based on goals employees have set for themselves. The ranking procedure helps control a second problem that undermines so many performance reviews; I call it the mayonnaise effect. Suppose you’re running a food company that wants to enter the mayonnaise market. You can make your mayonnaise with olive oil, sunflower oil, or corn oil. You can use vinegar, lemon, both, or neither. You can emulsify with egg yokes, dry eggs, or whole eggs. You can use lots of salt or not much salt. If you put ten different mayonnaise formulas on a table and ask people to rate each one, chances are pretty good they’ll make some big errors. They’ll probably recall their absolute favorites and the recipes that made them sick. But what about all the recipes in the middle? Are you confident your tasters will correctly rank formulas three through seven—none of which made a dramatic impression, but each of which has its unique qualities? How Cypress Awards Raises This condensed report shows how a focal-group leader awards raises among eight SRAM product engineers. The ranking committee rated engineer DD higher than FF, so the group leader awarded DD a larger merit increase. However, the computer’s equity analysis showed that DD was overpaid by $5,500 relative to his peers, while FF was underpaid by $3,600. As a result of equity adjustments, DD received a lower raise than FF. This same problem afflicts performance appraisal. It’s not too hard to identify the two outstanding performers and the two laggards in a group of ten employees. But to the people in the middle, there’s a big difference between being ranked fourth and seventh—and managers can badly muff these rankings. That’s why our system forces comparisons between pairs that leave little room for error. Consider a ten-person focal group with members A through J. The ranking committee’s job is to review every possible two-employee comparison and determine who was the superior performer for the year: Has A performed better than B? Has A performed better than C? Has B performed better than C? And so on. The software records the outcome of each comparison and develops relative rankings as the basis for merit increases. No confusion, no mistakes, no mayonnaise effect. These rankings become the basis of merit increases decided by the focal-group leader, the senior member of the ranking committee. (The focal-group leader alone is chartered to divide up the raise budget once the rankings are completed.) And merit means merit. The software reviews the ranking committee’s individual judgments and recommends appropriate merit increases. At this point in the process, no salaries have been disclosed—everything is expressed in percentages. The computer uses a pure and uncomplicated process to assign a certain percentage raise for a certain ranking. The leader can make adjustments to these recommendations, but the software checks to make sure the final numbers comply with several principles. There are 20 quality checks in all, including these: Monotonic distribution. Any group member ranked higher than another must receive a higher merit raise. This check prevents managers from taking money away from a top performer who happens to be highly paid and giving it to a weak performer. Forced differentiation. There must be a minimum spread (last year it was 8 % ) between the largest and smallest raises in the group. This check prevents the most common managerial cop-out, spreading raises evenly throughout the group. Reasonable adjustments. Although leaders are free to adjust the computer-generated recommendations, there are limits. This check works against clustering the majority of increases around one figure. Budgets. The average raise for the group cannot exceed the corporate target. This check emphasizes the point that giving big raises to people who deserve low raises hurts the entire group. What to Do When a Valued Employee Quits February 24, 1988 To: Vice Presidents and Managers From: T.J. Rodgers I have not been called on to save a valued employee for some time. Last week, I became involved in two such situations—both were successful—but my job was made more difficult because we did not follow some basic rules. I realized I have never formally stated our policy on resignations. Here it is: 1. React immediately. That means within five minutes. Nothing takes priority over working with a valued employee who has resigned. Delays such as “I’ll talk to you after our staff meeting” are unacceptable. Cancel the next activity you have scheduled. This demonstrates to the employee that he or she takes precedence over daily activities. It also gives you the best chance of changing the employee’s mind before he or she makes an irreversible decision. 2. Keep the resignation quiet. This is important for both parties. If other employees don’t know about the resignation, the employee does not face the embarrassment of publicly changing his or her mind. The company also gets more latitude. In one recent case, the resignation was disclosed. After I convinced the employee to stay, there were multiple rumors (all untrue) that we had “bought back” the employee. Cypress does not negotiate the salaries of employees who have resigned. We may communicate information about upcoming raises and evergreen stock options if that information is available at the time. 3. Tell your boss immediately and me within an hour. There is no excuse for not informing me (and everyone in the chain of command between the individual and me) as soon as a resignation occurs. I expect instantaneous communication and can be interrupted in meetings, called out of meetings in outside locations, or called at home. (I am listed in the telephone directory.) 4. Listen carefully to the employee. Once a resignation has occurred and the proper people know, the employee’s manager and vice president should sit down with the employee and listen carefully to the reasons behind the resignation. Any attempt to retain the employee will be severely impaired unless management listens to exactly what the employee says and accepts it. That message should then be transmitted up through the chain of command without any changes, even if it is unflattering to the manager involved. You should also make an exact determination of the employee’s options at the other company. Is he or she looking at a better job, more money, slower pace, faster pace, or fundamental career change? These issues will obviously be imperative in constructing an argument to change the employee’s mind. 5. Construct your arguments. Once you’ve gathered accurate data, sit down with your vice president and put together a plan to convince the employee to stay. The only possibly effective argument is one that validly claims the employee’s best interests are served by staying at Cypress. Once you and the vice president have formulated your employee-retention arguments, you may want me to become involved to help set the overall strategy. This strategy should be defined and refined on the very same day the employee resigns. Typically, an employee will have quit because of a “push” of some sort involving long-standing frustration at Cypress and a “pull” from another company where the grass looks greener. In some cases, realism will dictate that we can’t keep the employee. Ninety percent of the time, however, we can make a good argument that it is in the employee’s bets interest to stay at Cypress. 6. Use all the horsepower at your disposal to win. With a carefully constructed strategy, we can proceed to win back the employee. Think of what we’ve already done on day one. The employee got the message that quitting was a big deal because of our rapid reaction to the resignations. We reminded the employee that the company was truly interested in him or her because we took as long as needed to listen to what was wrong. On the second day, the employee should get the message that quitting was a mistake, that the company knows it was a mistake, and that we will single-mindedly try to rectify that mistake. Cypress will accept only two answers to our proposal to stay: yes, or we’ll talk about it some more. On the second or third day, as we present our position, the employee should continue to understand that this is not business as usual. We will interrupt our schedules. If appropriate, we may meet over meals during off hours. If the employee’s spouse is a major factor in the resignation, the spouse should be involved in the discussions. Bring in any level of management required to get the job done. If it takes the president (and it does in half the cases), then I have nothing more important to do than to sit down with the employee. Many middle managers mistakenly assume that I am too busy to interrupt my schedule to keep a good person in the company. Nothing is further from the truth. 7. Solve the employee’s problems. If we correct the problems that caused the employee to start looking around, we will succeed more than 80 % of the time in changing the employee’s mind. Most often, resigning employees like Cypress, its benefits, and the people with whom they work. They usually, however, do not like some of the particulars of their jobs or their direct supervisors. Their resolve to leave is further strengthened because they have found jobs (typically) at companies that are poor seconds to Cypress but that appear to offer some relative benefits. By alleviating the root problem at Cypress, and by stressing the fundamental differences between us and the other company, we can usually persuade the employee to stay. 8. Wipe out the competitor. Two objectives are important here: to shut down the competitor so firmly that it will conduct no further negotiations with our employee, and to shut down the competitor in such a manner that it believes it has wasted its time trying to hire away from Cypress. The employee should call the competition and turn down the offer and, in doing so, make it clear that he or she does not want counter offers and continuing negotiations. A model response goes something like this: “There was no counter offer, I just want to stay at Cypress. I think my long-term interests are served by being here. The same hour that I told my boss I was thinking about leaving, I had meetings with my boss, the vice president, and T.J. Rodgers. When they made the comparisons between my career at Cypress and at your company, it was clear that I made a mistake in thinking about leaving. I really do not want to take any time to come over to talk to you; my mind is made up. It would not be helpful to change your offer monetarily; I am fairly paid and have a good stock-option package. Money is not the issue.” 9. Prevent the next problem. The last step in the process is really the first step: sit down, think about your people, try to anticipate where you might have a problem in the future, and fix it before it starts. Only after they have awarded percentage increases based strictly on merit can managers make adjustments for salary inequities created by personal circumstances and historical accidents. Again, the software plays a central role. For each and every focal group, it generates a simple graph that compares the salaries of each group member (on the vertical axis) with merit rankings (on the horizontal axis). The graph also displays a trend line that describes the desired salary distribution for the department. We operate according to the general principle that the best performer in a department should earn at least 50 % more than the worst performer in that department. Managers adjust raises up or down to move salaries closer to the trend line. Here too the system checks for mistakes and makes sure the overall increase stays within budget. The power of our focal-review system is its simplicity. An easy-to-use software application (it runs on Lotus 1-2-3) rigorously applies a set of principles without usurping managerial discretion. It guides managers on a step-by-step journey through the evaluation and reward process, alerts them instantaneously when their decisions violate company policy, and guarantees that they stay within budget. The system also creates a paper trail that managers can use to explain to employees how raises were determined. We can tell each of our people how the ranking committee evaluated his or her performance (“You rated above average but not in the top 10 % ”), the percentage merit increase awarded by the focal-group leader, and how the manager adjusted the merit increase for equity. Most people will accept an outcome they are unhappy with so long as they understand the logic behind it. As for senior management, the system becomes yet another tool for management by exception. In a single, effective two-hour meeting, I review the raise of every employee at Cypress by reviewing the printouts the system generates. I can compare an individual’s raise with those granted to his or her peers, look at raise distributions in one group versus those in another, or compare this year’s raises for an individual or group with raises from the last two years. Of course, I can also check to make sure the system has been followed. Any manager who submits a printout that violates our policy gets it back with a note to do it right—which means very few managers ever submit faulty printouts. The system is simple, effective, and consistent. How Big Is Too Big? It’s natural to ask whether the systems I’ve described can work in companies much bigger than Cypress. I don’t know. But I do know that if they can’t, then these companies are probably unmanageable. We respect the limitations of size. I am absolutely convinced that any small group of smart, dedicated, hard-working professionals can beat any large group of average professionals with superior resources. There comes a point in any organization’s evolution when adding more people, more buildings, and more equipment reduces overall productivity. I believe that point is somewhere in the neighborhood of $ 100 million in annual revenues for a semiconductor company. Which is why Cypress today is not one company with total revenues of $ 200 million but five distinct companies linked by a common strategy, vision, and management systems. We have chosen to fund our growth with a venture-capital model, seeding new ventures under the Cypress umbrella. Our first startup, Cypress Semiconductor (Texas), was the Round Rock wafer fab. Aspen Semiconductor was founded in 1987 to design and develop new categories of logic chips. Multichip Technology was founded in 1988 to combine multiple chips into memory subsystems. Ross Technology, our fourth new company, has delivered the industry’s highest performance RISC microprocessor since it was formed in 1988. These subsidiaries have true autonomy. But I do expect two major commitments. A subsidiary must use the management systems that have served Cypress so well. It must also send high-level managers (almost always the president) to a few key corporate meetings in which we track and coordinate activities across the subsidiaries. We intend Cypress to be a billion-dollar enterprise before too long—without the waste, bureaucracy, and complacency that afflict so many large organizations. We’ll achieve that goal by seeding new startups, dividing down successful businesses as they get too big to manage tightly, continuing to refine our management systems, and driving them down through our organization and its subsidiaries. 1. Several of Cypress’s systems are enhancements of best practices among our competitors or companies with which our executives have been affiliated. For example, the goal system traces its origins to management-by-objectives techniques I learned from former AMI chief executive Glenn Penisten, who brought them over from Texas Instruments. Our expense-control system draws on techniques practiced at Mostek. 2. It’s not always possible to improve corporate revenue per employee. When our Round Rock fab went on line, corporate productivity “dropped” until the plant began volume shipments. But we still insisted on continual improvement from the rest of the company, and we adjusted our quarterly figures to measure it.

No Excuses Management

Partner Center