B O O K S I N R E V I E W

Universal lessons every manager can learnfrom Andy Grove's paranoia.

by Adam M. Brandenburger and Barry J. Nalebuff

Only the Paranoid Survive:How to Exploit the Crisis PointsThat Challenge Every Companyand CareerAndrew S. GroveNew York: Currency/Doubleday,1996

In 1965, Gordon Moore, chairmanof Intel, put forth a law of technol-ogy: The number of transistors ona chip doubles every 18 months.Moore's law has held up. Indeed,chips are now 500 times faster than

It's all too eosy to forgetthat Intel's success vs^asn't

so assured.

they were when IBM brought out itsfirst PC, in 1981. While the increasein chip processing power is an engi-neering marvel, the increase in In-tel's market value has been an equiv-alent financial marvel. A $1,000investment in Intel stock made inthe days of the IBM PC and the Intel

8086 chip would be worth more than$30,000 today.

Intel's track record is due in nosmall part to the performance of itsCEO and chief visionary, AndrewGrove. In his new book. Grove re-flects on his experiences as theleader of Intel. The title is Grove'snow famous management dictum:Only the Paranoid Survive. Youmight call this Grove's law of man-agement. Intel's history of close callsand spectacular successes suggeststhat there's much to learn from

Grove's paranoia.It's all too easy to for-

get that Intel's successwasn't always so assured.In 1985, the company'searnings were a merepenny per share; in 1986,Intel lost $173 million.

Those two years were marked bylayoffs, plant closings, salary cuts,and time off without pay. In fact,Intel is one of the few survivorsfrom the early days of computing.Remember Unisem, Mostek, or Ad-vanced Memory Systems? Grovedoes: "If you don't recognize the

names, it's because these companiesare long gone." Reason enough forGrove to be a touch paranoid.

At Intel, Grove, now 60, has expe-rienced a series of crises and up-heavals-and survived. He took Intelout of the memory chip business andinto microprocessors. Along theway, he bet on a chip technology(sticking with CISC versus switch-ing to RISC), preempted the com-moditization of chips |with the IntelInside campaign), suffered a publicrelations disaster (the Pentiumflaw), and survived a personal cri-sis (prostate cancer). Time for abreather? Hardly. For Grove, success

Adam M. Brandenburger is a profes-sor at the Harvard Business Schoolin Boston, Massachusetts. Barry f.Nalebuff is the Milton SteinbachProfessor at the Yale School of Man-agement in New Haven, Connect-icut. Brandenburger and Nalebuffare the coauthors of Co-opetition(Currency/Doubleday. 1996). TheirE-mail addresses are abranden-burger@hbs.edu and barry.nalebuff@y ale.edu.

168 HARVARD BUSINESS REVIEW November-December 1996

B O O K S I N R E V I E W

is "only a disaster away," and poten-tial disasters are always looming onthe horizon.

Today Intel, like many otber com-panies, faces new opportunities-aswell as threats-created by the Inter-net. And there's another, even moreimmediate, challenge. Intel's Pen-tium chips are almost too powerfulfor their own good. Given the powerof the Pentium, why should peopleupgrade to Intel's latest chip, thePentium Pro? And if they don't, bowwill Intel continue to grow? (Don'tworry. Grove has a plan.)

Only the Paranoid Survive offerspractical advice on bow to bridgetbat narrow line between catastro-pbe and opportunity, and seize tbeopportunities. Tbat's why this bookisn't just for managers of high-tecbbusinesses. Altbougb tbe illustra-tions come primarily from Intel, it'struly a general management book.Grove focuses on bis decision-mak-ing and tboiigbt processes-and thelessons apply hroadly.

As for tbe decision making pro-cess. Grove analyzes his own trackrecord with disarming honesty.In spite of all his accomplishments,be doesn't take tbe line that be hasgotten things right because be's sosmart. Quite the contrary. Grove'sapproach is. Here's how 1 almostgot it wrong, how I sometimes didget it wrong, what saved me, andwhat 1 learned. Only the humblesurvive? Or, at least, tbe complacentget killed.

Many a SIP Twixt the Cupand the Lip

Grove bas a two-step approacb tomaking strategic decisions. The firststep is to identify what he calls stra-tegic inflection points, or SlPs. ASIP arises when there's an order-of-magnitude ehange in a company'senvironment. These are times wbenmanagement must act. Tbese are op-portunities to go way u p - o r waydown - in the world. How do youspot a SIP?

Grove offers a "six forces" frame-work for identifying SIPs. He startswitb Micbael E. Porter's five-forcesmodel: customers, suppliers, com-petitors, potential competitors, andproviders of substitutes. He then in-

corporates some of our own work,which adds complementarity to thestrategy map. After all, what is Mi-crosoft to Intel? It isn't a customer,a supplier, or a competitor, and yetthere's clearly a critical interdepen-dence between the two companies.Intel and Microsoft are what we'vetermed complementors. Microsoft'sWindows 95 operating-system soft-ware complements Intel's Pentiumchips, and vice versa-tbat is, each prod-uct makes the othermore valuable. Com-plementors, then, areGrove's sixtb force. Toidentify a SIP, Grovesuggests scanning tbeenvironment to lookfor order-of-magni-tude, or "lOX," changes in any oftbese six forces that affect a busi-ness's fortunes.

So you do a six-forces analysis anddiscover that something in theworld has fundamentally changed.You're at a SIP. It's time to makea decision. That's the seeond step.

Grove's approach to decision mak-ing is highly analytieal, almost sci-entific. His case studies illustratehis method, whicb also relies on abealthy attitude. Grove is remark-ably good at not letting pride or egoget in the way. He recognizes tbeneed to step outside Intel, so tospeak, and analyze a situation fromthe perspective of someone whodoesn't have a vested interest intbe status quo. He regularly triesto prove bimself wrong by lookingfor counterexamples to bis currenttbinking. Trained as an engineer.Grove positively enjoys immersinghimself in the data and being thescientist again. For example, whendiagnosed witb prostate cancer, heused this approach to challenge theconventional wisdom of treatment.[See "Taking On Prostate Gancer,"Fortune. May \3, 1996.)

Grove doesn't agree with manage-ment guru W. Edwards Deming's callto end fear in organizations. Indeed,Grove believes tbat at least somefear is healthy-especially in organi-zations that have bad a bistory ofsuccess. Fear can be a bealtby anti-dote to tbe complacency tbat suc-

cess often breeds. A touch of para-noia- a suspicion tbat tbe worldis changing against you - is whatGrove prescribes.

Grove realizes that anotber kindof fear often rules organizations-onethat is unhealthy and counterpro-ductive to good decision making. Heis well aware of the meltdowns tbatoccur when employees know thatsomething big is happening out

Some fear is healthy-especially In organizationsthat have had a historyof success.

there but arc too cowed to tell man-agement ["1 don't think they want tohear that"). Grove doesn't want In-tel's management to he among themany that failed to respond to chal-lenges because "had news neverreacbed them." He favors open dia-logue at Intel and proudly reports aconversation in whicb be was told,"Hey, Grove, you're out of yourdepth here; let me teach you a fewthings." (In aeademia, professors saythat all tbe time to tbeir deans, butthey have tenure.)

Forget MemoriesProbably the most critical deci-

sion Grove has had to make at Intelwas one prompted by a dramatic in-crease in competition from Japanesememory-chip makers in the mid-1980s. Memory chips were Intel'soriginal business.

Life in the memory business wastough. As Grove describes it, tbeJapanese semiconductor manufac-turers decided to win tbis businesswith the 10% rule: They would un-dercut Intel by 10% to get the cus-tomer. If Intel matched, tbey wouldundercut hy anotber 10% -and so onuntil they got the customer.

This policy created a crisis, or SIP,for Intel. A decision had to be made,and various options were proposed.One suggestion was to build a giantnew dedicated factory in tbe bopesof establishing a cost advantage overIntel's Japanese competitors. An-

HARVARD BUSINESS REVIEW November-December 1996 169

B O O K S I N R E V I E W

Other was to push the technologyenvelope and develop a new and su-perior memory ehip. Anotber was totarget niche markets.

After almost a year of frustratingdebates and much dithering. Grovecame to realize that none of thoseoptions were the answer to Intel'sproblems. In fact, there was no solu-

Grove tries to provehimself wrong. If there's o

problem v̂ rith his mentalmodel/ he ^^ants to find it.

tion to be found within the memorybusiness. Even tben, Grove didn'tfind it easy to imagine abandoningthe business in which Intel bad got-ten its start: "Intel equaled memo-ries in all of our minds." To eome togrips witb the crisis, be had to takehimself and his own baggage out ofthe picture.

Here's how he did it. In the midstof the doldrums of 1985, Groveposed a hypothetical question to hiscolleague Gordon Moore: "If we gotkicked out and tbe board brought ina new CEO, wbat do you tbink bewould do?" Moore answered witb-out besitation, "He would get us outof memories." To which Grove re-sponded, "Why shouldn't you andI walk out the door, come back anddo it ourselves?"

That's just what they did. Theyredirected Intel's resources awayfrom memories and into its micro-processor business-which, at thetime, played second fiddle to Intel'smemory business. True, Intel hadinvented the microprocessor back in1971, but in tbe mid-1980s, secondsourcing had created multiple com-petitors, and a downturn had led toexcess capacity. Microprocessorsweren't an obviously attractive busi-ness. The bold decision to commitIntel's future to microprocessorsnevertheless saved the company andstarted it on the road to today'sgreatness.

Aware of his own struggle over thememory chip crisis. Grove arguesthat, very often̂ it takes fresh blood

to overcome management's attach-ment to the status quo. It's not tbatfresh blood is neeessarily smarterthan existing management. Rather,the practice of replacing corporateheads is motivated by tbe need tobring in someone who is not invest-ed in the past. For Grove, this pro-cess is almost a biological impera-

tive. People succeed ina certain corporate en-vironment precisely be-cause tbeir mind-set fitstbat environment. Thosesame people are thereforeunlikely to be quick atadapting to a changed en-vironment. Grove issuesthe following challenge:

"If existing management want tokeep tbeir jobs when the basics ofthe business are undergoing pro-found cbange, they must adopt anoutsider's intellectual objectivity."

Self-ParanoiaAccording to the American Her-

itage Dictionary, the definition ofparanoid is "showing an unreason-able distrust or suspicion." Groveoften distrusts his own ideas, sus-pects himself of being wrong. Tbat'sanother source of strength.

It's easy to fall into tbe trap oflooking for confirming evidence ofone's views. This is one of severaldecision traps identified by Max H.Bazerman, a professor at Northwest-ern University's Kellogg School ofManagement, in bis book fudgmentin Managerial Decision Making(Wiley & Sons, 1994). When peoplebelieve something to be true, theytend to look for more evidence insupport of their theory. Prior suc-cesses often reinforce this bebavior.Almost by definition, a track recordof success means there is lots of evi-dence confirming what one beheves.

However, collecting confirmingevidence-no matter bow much-will never prove a hypothesis. Onthe other hand, we can disprove ahypothesis by finding a counter-example. To test a hypothesis, weneed to aetively seek data thatwould disprove it. That is alwayshard to do because we don't like tobe proved wrong-it's simply humannature. But at Intel, Grove goes out

of his way to try to prove himselfwrong. He runs experiments with aview to challenging bis currentthinking. If tbere's a problem withhis mental model, be wants to findit. (See tbe insert "See a Pattern?")

Back in tbe late 1980s, there was asplit at Intel. Some people tbougbttbe future lay in the more powerfulRISG [reduced instruction set com-puting) chip architecture. Tbeythought RISC could be a "lOX"cbange. Otbers argued tbat tberewas still plenty of room for improve-ment to the existing CISC (complexinstruction set computing) chips.Grove didn't believe that RISGwould replace CISC in PCs, primar-ily because it was backward ineom-patible, so he thought Intel shouldcontinue with the CISC architec-ture. But he was willing to run an ex-periment just in case he was wrong.Accordingly, he supported a projectto develop an Intel RISC ebip. Asthings turned out, Intel went on tohave a huge success with its next-generation 486 CISC chip and hasstayed witb GISG ever since. But theRISC cbip it developed was a sensi-ble insurance policy and even turnedinto a moneymaker.

Today tbe big debate in the com-puter industry is, of course, aboutthe Internet. Some people-most no-tably Larry Ellison of Oracle-aretalking about a $.S00 network com-puter (NG) that will replace today'smore expensive Intel-based ma-chines. Grove doesn't tbink such achange will happen. Indeed, his viewis that the Internet will result in"new players on the scene to be sure,but they are just as likely to play therole of complementors as competi-tors." Still, just in case he's wrongabout NCs, the ever paranoid Groveis running anotber experiment. Intelitself is testing out tbe NC concept.As Grove puts it, "You can't sudden-ly start experimenting wben you re-alize you're in trouble unless you'vebeen experimenting all along."

Flawed LogicEvery now and then, even the best

of us miscalculate. Andrew Grove-and the Pentium chip-are no excep-tions. In fact, this embarrassment isthe first story in Grove's book.

172 HARVARD BUSINESS REVIEW November-December 1996

B O O K S I N R E V I E W

In October 1994, Thomas Nicely, amatb professor at Lyncbburg Collegein Virginia, posted a notice on tbe In-ternet describing a flaw in the waythe Pentium chip did division. Infact, Intel had known about the prob-lem since the summer and was wellon the way to fixing it. The companyhadn't told people about tbe flaw,in part because its engineers badestimated tbat tbe cbances of tbeirencountering it were vanishinglysmall. But after CNN picked up thestory, an avalanebe of negative pub-licity started. Intel argued tbat theflaw was essentially irrelevant tousers, though it did offer to replacePentium cbips on a case-by-case ba-sis. Tben, in early Decemher, IBM as-serted that the ebances of encounter-ing a problem with tbe Pentium weremuch greater tban Intel was sayingand that, to protect customers, it washalting sbipments of Pentium-basedcomputers.' Tbat forced Intel's hand.A week later, Intel reversed courseand offered a no-questions-askedreturn policy. In tbe process, it tooka write-off of $475 million.

How could a minor flaw in tbePentium cause a public relations dis-aster and a balf-billion-doUar write-off? Looking back. Grove sees that

tbis time be failed to step outside In-tel and analyze the situation fromtbe outside in. He was still tbinkingof Intel as an engineering companyand as something of tbe underdog toboot. He wanted to debate the Pen-tium issue on its tecbnical meritsand to decide whose chips to replaceon a tecbnical basis. Tbe problemwas that tbis approach didn't takeinto account bow other peopleviewed Intel.

Wbat Grove missed, he now real-izes, was that, as a result of its verysuccessful Intel Inside campaign, In-tel had effectively transformed itselffrom an engineering company intoa consumer products company. Itbad created a direct relationshipwitb consumers-even if consumersdidn't buy from it directly. Peopleidentified tbeir computer by tbe typeof chip [a 486 or a Pentium), not justby who made tbe box. There alsowas tbe fact tbat Intel bad grownto he tbe world's largest semicon-ductor manufacturer, oversbadowingtbe Japanese companies that bad al-most put it out of business a decadeearlier. Tbe public perceived Intelas top dog, not underdog. Tbis newreality didn't sink in fast enough atIntel. As Grove writes, "[We were]

like a kid wbo suddenly looks downat bis fatber, our sizes reversed."

So, in bandling tbe Pentium crisis,the relevant yardstick for Intelsbould have been bow leading con-sumer-products companies bandiedsimilar crises. In fact, the standardbad been set by Johnson St Jobn-son during tbe Tylenol poisoningepisode. Johnson & Johnson wentabove and beyond tbe call of duty, re-placing all Tylenol pills in the coun-try, even in regions wbere there wasno apparent threat. That responseraised the har. From then on, con-sumer products companies have hadno choice but to go to enormouslengths to protect their customersfrom any barm- real or perceived.Anything less isn't good enougb.

Against that yardstick, Intel's ac-tions didn't look very good. First, thecompany didn't announce tbe exis-tence of the problem as soon as itknew there was one. Had it done so,tbere would probahly have been nostory. By letting someone else dis-close the problem first, Intel under-

1. IBM's strategy ultimately backfired. OnceIntel announced its free-exchange program,consumers flocked back to tht Pent mm-basedmachines, Mtanwhilc, IliM cuntinued to li)semarket share with its 486-powered machines.

See a Pattern?

Here's a sbort exercise that il-lustrates tbe idea of trying toprove yourself wrong. We've cbo-sen a rule tbat some sequences oftbree numbers obey. We'll startby telling you that tbe sequence2,4, 6 obeys the rtile. Can youguess wbat the rule is? Youcan produce any sequence ofthree numbers you wisb, andwe'll tell you wbether the se-quence obeys tbe rule. Onceyou think you know tbe rule,make your guess. But be awarethat you get only one guess. If youwant to give the exercise a sbothcfore reading on, you can goto our Web site and do tbe exer-cise interactively; the address

IS http://mayet.som.yale.edu/coopetition.

When people play tbis game,they usually proceed in the fol-lowing manner: Tbey ask, "Does4, 6, 8 satisfy tbe rule?" Yes. Tbe

246adventurous then ask whether1, 2, 3 satisfies tbe rule. Yes,again. How about 5, 10, 15? Onceagain, yes. At this point, peopledecide tbat they've seen enougbdata and can giiess tbe rule: Thetbree numbers bave to increase

by a constant amount. Was thatyour guess?

If so, tben before making yourguess, you would bave done wellto ask wbetber tbe sequence 1, 2,47 satisfies the rule. Suqirise! It

does. How about 3, 2, 1? Itdoesn't. As you've probablyrealized by now, tbe rule is areally simple (okay, dumh)one: Any three increasingnumbers will do. The point is

that tbe vast majority of peoplestop and guess the rule incorrect-ly before tbey've asked a questionthat could discriminate betweentheir hypotbesis and various al-ternatives. To test a hypothesis,try to disprove it.

HARVARD BUSINESS REVIEW November-December 1996 173

mined its credihility just when itneeded it most. Then there was thecase-hy-case return policy. AlthoughIntel wasn't denying applications forreplacement chips, the policy creat-ed the perception that the company,not the customer, was deciding whowould and wouldn't get a new chip.None of this looked too good comingfrom the people who make the hrainsof the computers we all rely on.

The Pentium episode reminds usof an academic seminar at which thespeaker began hy laying out all thesbortcomings of his model. Evident-ly, he wasn't worried that admittingto some flaws would negate thevalue of the model. This was an actof confidence. The audience saw themodel's overall value and had onlynice things to say. After all, all thenegative points had already beenmade - hy the speaker himself!There was no reason for the audi-ence to repeat them. We thinkthere's a general lesson here. Whenyou criticize yourself, you preemptcriticism by otbers. If someone elsedoes repeat the criticism, you canpoint out that you agree-you'veeven said the same thing yourself. Inthe case of the Pentium, when Inteldiscovered the flaw, it could have an-nounced, "We've identified a prob-

B O O K S i N R E V I E W

lem; we're working to fix it. In themeantime, here's what we know."Chances are there would then havebeen no crisis at all.

The Next ChapterCan Grove finally relax? Afraid

not. Today Intel is facing anotherSIP. What's making Grove a littleparanoid today is the processingpower that people already have ontheir desktops.

Over the years, Intel's engineershave done a brilliant job of develop-ing increasingly powerful chips. Andthat's the problem. Intel's currentehips are so good that most peopledon't feel any pressure to upgrade.They already have more processingpower than they need to run theirfavorite applications.

Grove recognizes that, right now,people don't have a strong enoughincentive to huy Intel's next-genera-tion chips- such as the newly re-leased Pentium Pro. He knows thathe must actively engineer demandfor the new chips. If he doesn't, themarket will hecome saturated andIntel's competitors-AMD, Cyrix,and others-will catch up. That iswhy Intel is currently at a SIP.

Grove has a plan. He's teaming upwith several of Intel's complemen-

tors to develop products tbat willpush the limits of processing power.Thus, Intel is working with MCI toprovide more bandwidth for net-works. After all, without more hand-width, people can't access the largequantities of data that require tbenumher-crunching power of a Pen-tium Pro. For the same reason, Intelis working with others to develop"hybrid applications" such as inter-active games on the Web. It is evenventuring outside its core husinessto ensure that essential complemen-tary products get developed, Thecompany has invested more than$100 million in ProShare, a video-phone product. If desktop videocon-ferencing takes off, so, too, will de-mand for the Pentium Pro.

To see what Intel is doing to devel-op the market for its future-genera-tion chips, go to its Web page. Thereyou can read not only about Intel'smicroprocessors but also ahout allthe complementary products thatthe company is promoting. Amongthem are interactive games, Inter-cast, and an Internet phone that's aprecursor to Internet videophones.This is where the next chapter of theIntel story is unjolding^ §Reprint 96604To order reprints, see tbe last page of this issue.

"I'm doing my geography homework - what are the principal exports of your country}'

CARTOON BY GEORGE DOLE 175