Paper 26 - No Plagiarism - URGENT

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 17 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerServ…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

been one of the most stable industries. But in 1996, when Jean-Bernard Lévy and I were discussing Cisco Systems, the structure of the computing and communications industry had suddenly become "uid and turbulent. Of the visible trends, the rise of personal computing and data networking were on everyone’s radar. The deregulation of telecommunications and its shift to digital technology had been long anticipated. The two more mysterious shifts were the rise of software as a source of competitive advantage and the deconstruction of the traditional computer industry.

It seems obvious in hindsight. Both the rise of software’s importance and the computer industry’s deconstruction had a common cause: the microprocessor. Yet these connections were far from obvious in the beginning. Everyone in high tech could see the microprocessor, but understanding its implications was a much more di$cult

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 18 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

proposition. Let me share with you a personal view of some of the steps along that path.

Software’s Advantage

How had software become such a sharp source of advantage? The answer is that millions of microprocessors, in everything from PCs to thermostats, bread machines to cruise missiles, meant that their programming determined the performance of these devices.

When I was an undergraduate studying at UC Berkeley in 1963, two of the main areas of excitement within electrical engineering were integrated circuits and computing. The !rst integrated circuits had been demonstrated in 1958, and devices created for the Minuteman missile project had integrated hundreds of transistors onto a single chip. With regard to computing, there was nothing mysterious about making

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 19 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

a computer’s central processor—the circuit patterns had been common knowledge since the 1950s.

All of my Berkeley classmates understood that if one could push integration from hundreds of transistors to thousands of transistors on a single chip, it would be possible to have a one-chip computer processor. Silicon Valley historians and patent attorneys love to argue over who “invented” the !rst microprocessor, but the idea of putting all the components of a processing unit onto one chip was in the air as soon as integrated circuits appeared. In any event, the !rst microprocessor o#ered for sale was Intel’s 4-bit 4004 in 1971, containing 2,300 transistors.

At that time, the chip market had two segments. Standardized chips such as logic gates and memory were produced in high volume but were commodities. Specialized proprietary chips or chipsets provided much higher margins but were only

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 20 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

produced to order in low volumes. And the rights to a complex proprietary chip belonged to the customer who had ordered it and sometimes designed it.

Signi!cantly, and often sadly, many crucial decisions do not appear to be decisions at the time. Instead, managers simply apply standard operating procedures to the situation. Within Intel, the 4004 microprocessor was classi!ed as a specialized proprietary design, so the rights to it belonged to customer Busicom, which intended to build it into its line of calculators. As luck would have it, Busicom came under pro!t pressure and asked Intel for a price reduction. Intel lowered the price in return for the right to sell the chip to others. Then, amazingly, this pattern was repeated with Intel’s next microprocessor, the 8-bit 8008. This microprocessor had been created as a proprietary chip for CTC, the maker of Datapoint computer terminals. Running out of money, CTC traded the

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 21 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

rights to the 8008’s design for the chips it needed. In this case, CTC had actually designed the chip’s instruction set—you can still see echoes of that instruction set in Intel’s most advanced x86 processors.

It took years for Intel’s management, and the rest of the industry, to fully appreciate the implications of a general-purpose chip that could be specialized with software. Instead of each customer paying to develop a specialized proprietary chip, many could use the same general-purpose microprocessor, creating proprietary behavior with software. Thus, the microprocessor could be produced in high volume. Instead of being a job shop for other companies’ designs, Intel could be a product-based technology company. Speaking about the 4004, and microprocessors in general, then chairman Andy Grove said, “I think it gave Intel its future, and for the !rst !fteen years we didn’t realize it. It has become Intel’s

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 22 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

de!ning business area. But for … maybe the !rst ten years, we looked at it as a sideshow.”1

Intel cofounder Gordon Moore became famous for his “law” predicting the rate of progress in speed and manufacturing costs in integrated circuits. He was less well known for his observation that design costs for custom chips were becoming larger than fabrication costs and were escalating at an unsupportable rate. He wrote, “Two things broke the crisis for semiconductor component manufacturers … the development of the calculator [microprocessor] and the advent of semiconductor memory devices.” To Moore, the beauty of these devices was that, although complex, they could be sold in high volumes.

In a discussion with a group of managers at Qualcomm, a San Diego maker of mobile phone chips, I reviewed Moore’s point about the escalating costs of designing

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 23 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

more and more complex special-purpose chips. One manager was puzzled and asked if it wasn’t also expensive to create software. He went on to rhetorically ask “Are software engineers less expensive than hardware engineers?”

It was a good question. None of us had an instant answer. I sharpened the question with an example from my own experience. Rolls-Royce had wanted to create a sophisticated fuel monitoring and control unit to enhance the e$ciency of its jet engines. It could have accomplished this through proprietary hardware embodying the necessary logic, or it could have used a general-purpose microprocessor, expressing its proprietary ideas by writing software to program it. Whether it chose to use a microprocessor plus software or proprietary hardware chips, it would have had to do a lot of engineering work for only a few thousand installations. Why prefer software?

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 24 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

As is often the case, restating a general question in speci!c terms helped. We quickly developed an answer that has since stood up to scrutiny by a number of other technical groups: Good hardware and software engineers are both expensive. The big di#erence lies in the cost of prototyping, upgrading, and, especially, the cost of !xing a mistake. Design always involves a certain amount of trial and error, and hardware trials and errors are much more costly. If a hardware design doesn’t work correctly, it can mean months of expensive redesign. If software doesn’t work, a software engineer !xes the problem by typing new instructions into a !le, recompiling, and trying again in a few minutes or a few days. And software can be quickly !xed and upgraded even after the product has shipped.

Thus, software’s advantage comes from the rapidity of the software development cycle—the process of moving from concept

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 25 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

to prototype and the process of !nding and correcting errors. If engineers never made mistakes, the costs of achieving a complex design in hardware and software might be comparable. But given that they do make mistakes, software became the much- preferred medium (unless the cutting-edge speed of pure hardware was required).

WHY COMPUTING DECONSTRUCTED

In 1996, soon after my conversation with Jean-Bernard Lévy in Paris, Intel chairman Andy Grove published his insightful book Only the Paranoid Survive. Grove drew on his expertise in both business and technology to forcefully describe how “in"ection points” can disrupt whole industries. In particular, he described the “in"ection” that had transformed the computer industry from a “vertical” to a “horizontal” structure.

In the old vertical structure, each

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 26 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

computer maker made its own processors, memory, hard drives, keyboards, and monitors, and each wrote its own systems and applications software. The buyer signed up with a maker and bought everything from that manufacturer. You couldn’t plug an HP disk drive into a DEC computer. By contrast, in the new horizontal structure, each of those activities had become an industry in its own right. Intel made processors, other !rms made memory, others made hard drives, Microsoft made systems software, and so on. Computers were assembled by mixing and matching parts from competing manufacturers.

Grove was dead-on in understanding that “not only had the basis of computing changed, the basis of competition had changed too.”2 Still, as a strategist, I wanted to know more. Why had the computer industry deconstructed itself and become horizontal? Andy Grove wrote, “Even in retrospect, I can’t put my !nger on

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 27 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

exactly where the in"ection point took place in the computer industry. Was it in the early eighties, when PCs started to emerge? Was it in the second half of the decade, when networks based on PC technology started to grow in number? It’s hard to say.”3

I was puzzled over the cause of the computer industry’s deconstruction. Then, about a year later, the reason snapped into focus. I was interviewing technical managers at a client !rm and one said that he had formerly been a systems engineer at IBM. He then explained that he had lost that job because modern computers didn’t need much systems engineering. “Why not?” I asked without thinking.

“Because now the individual components are all smart,” he answered. Then I saw it.

The origin of Andy Grove’s in"ection point was Intel’s own product—the microprocessor. The modularization of the computer industry came about as each

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 28 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

major component was able to contain its own microprocessor—each part became “smart.”

In many traditional computers, and early personal computers, the CPU—the active heart of the machine—did almost everything itself. It scanned the keyboard, looking for a keystroke. When it sensed one, it analyzed the row and column of the keystroke on the keyboard to determine the letter or number that had been pressed. To read a tape drive, the CPU constantly controlled the speed of the tape reels and the tension of the tape, stopping and starting the drive as it interpreted the data coming in and storing it in memory. To typewrite with a “daisy-wheel” printer, the CPU controlled the spin of the wheel and the timing of each separate hammer strike. In some cases, designers created custom mini-CPUs to manage these peripherals, but the integration among these devices remained complex and unstandardized and

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 29 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

consumed a great deal of systems engineering e#ort.

After the arrival of cheap microprocessors, all that changed. The modern keyboard has a small microprocessor built into it. It knows when a key has been hit and sends a simple standardized message to the computer saying, in e#ect, “The letter X was pressed.” The hard-disk drive is also smart so that the CPU doesn’t need to know how it works. It simply sends a message to the hard disk saying “Get sector 2032,” and the hard-disk subsystem returns the data in that sector, all in one slug. In addition, separate microprocessors control the video screens, memory, graphics processors, tape drives, USB ports, modems, Ethernet ports, game controllers, tape drives, backup power supplies, printers, scanners, and mouse controllers that make up modern computers.

Smart components operating within a de

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 30 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

facto standard operating system meant that the job of systems integration became almost trivially simple. The skills at systems integration that IBM and DEC had built up over decades were no longer needed. That was why my engineer-informant had lost his job.

With the glue of proprietary systems engineering no longer so important, the industry deconstructed itself. Modules did not have to be custom designed to work with every other part. To get a working system, customers did not have to buy everything from a single supplier. Specialist !rms began to appear that made and sold only memory. Others made and sold only hard drives or keyboards or displays. Still others made and sold video cards or game controllers or other devices.

Today, there are many academic researchers who look at the computer industry and see a network of relationships, each one a channel whereby one !rm

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 31 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

coordinates with another. The idea of a network especially enchants modern reductionist sociologists who count connections among people, skipping over the old-fashioned hard-to-quantify questions about content and meaning. However, dwelling on this network of weak relationships confuses the background with the absent foreground. What is actually surprising about the modern computer industry is not the network of relationships but the absence of the massively integrated !rm doing all the systems engineering—all of the coordination—internally. The current web of “relationships” is the ghostly remnant of the old IBM’s nerve, muscle, and bone.

CISCO SYSTEMS RIDES THE WAVE

As I mentioned to Jean-Bernard Lévy that day in 1996, Cisco Systems was a recent start-up that had “grabbed the inter-

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 32 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

networking equipment market right out from under the nose” of industry giants. The story of how Cisco Systems came into being and how it came to beat the giants vividly demonstrates the power of using waves of change to advantage. The particular waves Cisco used were the rise of software as a critical skill, the growth in corporate data networking, the shift to IP networks, and the explosion of the public Internet.

In the early 1980s, Ralph Gorin, the director of Stanford University’s computer facilities, wanted a way of hooking together separate networks of Apple, Alto, and DEC computers as well as various printers. Each type of computer network used di#erent kinds of wires, plugs, timing, and, most important, di#erent proprietary protocols for encoding information. Gorin’s request was for a way to interconnect these proprietary networks. The solution, called the blue box, was engineered by Stanford

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 33 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

sta# members Andy Bechtolsheim and William Yeager.4

Two other Stanford sta# members, Len Bosack and Sandy Lerner, began to re!ne the blue box, moving the improved designs into their newly formed company, Cisco Systems. After an acrimonious split with Stanford in 1987, Cisco Systems received full legal rights to the software in return for a payment to Stanford of $167,000 and the promise of discounts on Cisco products. It was thought that Cisco would sell some of its boxes, now called routers, to other research universities, but there was no expectation that Cisco would make much money. There was certainly no expectation that Cisco would become, for a brief moment in 2000, the most valuable corporation in the world.

Soon after receiving its !rst injection of venture capital in 1988, Cisco’s management was turned over to professionals. John Morgridge was CEO

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 34 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

from 1988 to 1995, followed by John Chambers. Both CEOs skillfully guided Cisco to take advantage of the powerful forces at work in its industry. During the 1988–93 period, Cisco rode three simultaneous waves. The !rst wave was the microprocessor and its key implication—the centrality of software. Cisco outsourced the manufacture of its hardware, concentrating on software, sales, and service. Ralph Gorin remarked that “Cisco cleverly sold software that plugged into the wall, had a fan, and got warm.”

The second wave lifting Cisco in its early years was the rise of corporate networking. Just as at Stanford, corporations were discovering the need to connect mainframes, personal computers, and printers that used di#erent network protocols. The Cisco router’s ability to handle multiple protocols was in growing demand.

The third wave was IP (Internet Protocol)

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 35 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

networking. In 1990, most network protocols had corporate owners and sponsors. IBM had SNA (Systems Network Architecture), Digital Equipment Corporation had DECnet, Microsoft had NetBIOS, Apple had AppleTalk, Xerox had developed the Ethernet, and so on. By contrast, IP was created in the late 1970s to handle tra$c on ARPANET, the precursor to the Internet. IP was pure logic—it had no wires or connectors, no timing speci!cations or hardware. Plus, it was free and was no company’s proprietary product. As corporations began to hook disparate computers into networks, corporate IT departments began to see the value in a protocol that was vendor neutral. Increasingly, corporations made IP their backbone network protocol and, at the same time, Cisco began to make IP the central hub protocol on its routers.

Critically, none of the industry incumbents jumped to forcefully occupy

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 36 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

this space. Each had its own proprietary network protocol and each was loath to totally abandon it. And each was even more loath to produce equipment that would help a competitor’s equipment onto the network.

If three waves were not enough, Cisco literally exploded under the force of a fourth wave that hit in 1993: the rise of Internet use by the general public. Inside corporations, computer users suddenly wanted Internet access. Not just dial-up access over a modem, but a direct always- on connection to the IP backbone. As universities and corporations scrambled to make this happen, IP won the battle for internal network standards, and Cisco routers captured two-thirds of the corporate networking market. At the same time, tra$c on the Internet backbone was skyrocketing and Cisco was there, providing the high-speed routers to handle the "ow of Internet data on a continental

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 37 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

scale. Where there were competitive barriers, Cisco maneuvered around them. In 1992–94, Cisco worked with IBM to add support for IBM’s proprietary SNA protocol to its routers and worked with AT&T and others to make sure that its equipment supported telecommunications industry protocols (for example, Asynchronous Transfer Mode and Frame Relay).

When faced with a corporate success story, many people ask, “How much of the success was skill and how much was luck?” The saga of Cisco Systems vividly illustrates that the mix of forces is richer than just skill and luck. Absent the powerful waves of change sweeping through computing and telecommunications, Cisco would have remained a small niche player. Cisco’s managers and technologists were very skillful at identifying and exploiting these waves of change, and were lucky enough to make no fateful blunders. Key competitor IBM was pulling its punches after thirteen

11/7/19, 9'26 PMGood Strategy Bad Strategy: The Difference and Why It Matters

Page 38 of 59https://pdc-evs-ebscohost-com.ezproxy.liberty.edu/EbscoViewerSer…onAndAbstract&isEvsViewerCssEnabled=True&isHoldModalEnabled=True

years of antitrust litigation. The Internet came along at just the right time to accelerate Cisco’s upward climb. And various incumbents were held in check by their own inertia, their strategies of supporting a single or a proprietary protocol, and by the very rapidity of change.

SOME GUIDEPOSTS

It is hard to show your skill as a sailor when there is no wind. Similarly, it is in moments of industry transition that skills at strategy are most valuable. During the relatively stable periods between episodic transitions, it is di$cult for followers to catch the leader, just as it is di$cult for one of the two or three leaders to pull far ahead of the others. But in moments of transition, the old pecking order of competitors may be upset and a new order becomes possible.