Japanese Software Industry:
Where's the Walkman?
ProfessorEdwardA.Feigenbaum Professor of Computer Science
Software StudyDirector, Stanford Computer Industry Project
Both economists and businesspersons are interested inthe evolution ofindustries-the economists to understand the process
and the business people tobetter understandmarkets, competition and otherfeatures ofthe business environment. For all ofthese reasons, faculty members and students at Stanford University, supported by the SloanFoundation, are engaged in a study ofthe worldwide computer industry. Oneofthemajor foci ofthestudy is
the software sub-industry ofthe computer industry, a segment that is paradoxically booming yet troubled and difficult to understand.
Nowhere is this more true than in Japan, in many other areas a"heaven" for things technological. Stanford University has an ideal base for studying Japanese societal phenomena, the Stanford Japan Center. Itwas mygood fortune tohave afaculty assignment there that allowed me to conduct a study of the paradoxical Japanese software industry. I was doubly fortunate to have as research assistants fifteen very smart students from Stanfordand Dartmouth whowere"students abroad"in theSpring of1993. This paper willpresent, hopefullyin not too "academic" a manner, someresultsfromthis fascinating study.
A
Quiz
In the tradition of professors,let megive a quiz.
Question
1: Measured byannualrevenue, how much bigger is the US software giant Microsoft than Nintendo? (Microsoft, of course, is an everything-you-need software supplier and Nintendo specializes in games only.) The surprising answer is that Nintendo is 25% bigger than Microsoft (approximately 5 billion dollars ayearrevenues versusMicrosoft's 4billion a year). Ofthe four largest packaged software companies in the world, ifyouinclude the games software companies then the United Stateshas only two. Nintendo is number 1. The listofthe top four also includes Sega, another
gamessoftware company. That isoneside ofthe Japanese software industry. Second question: What is the best selling Japanese (non-games) software package in the American market? Ifthat one is too hard,try this variation. What isyourfavorite Japanese software package? The answeris: nobody has an answer to this question because there's no suchthing as a Japanese presencein the American software market. The Americansoftware market is huge and the Japanese have no presence at all. Now ask an analogous question about consumer electronics. What's your favorite Walkmaninstrument? What'syourfavorite CD player? What's yourfavorite television set? What's yourfavorite laptop computer? In this light, it is interestingthat the Japanesehave no presencewhatever in the American software market. Thispaperwilltry to explainthis fact. Itwill also address a broader range ofissues concerning the Japanese software industry in an attempttounderstand it.
It's worthwhile understanding the Japanese software industry, particularlyifyou're trying to deal with it,orsell into it. Iwas on the board
ofdirectors ofa software company thatwas doingsubstantial business with the Japanese; Inowrealize that thecompanyknew essentially nothing about the industry that we were participating in and the markets we were selling into!
Size ofthe Industry
What is the total size ofthe Japanese software business? It's about $100 billion annually for all kinds of software—that is, programs which are packaged, plus programs which are custom prepared by a contractor, or softwarethatismade inacompany (doingtheir ownprograms). Ofthat $100 billion, one halfis spent on in-house programming. Approximately $35 billion is spent on packaged software or custom programming services. Computer makers spend about $15 billion on system software for their equipment. In the Japanese way of doing things, customers like to have software customized for their own interfaces and their own ways of doing business; andthey're willing topayforit. Approximately 90% ofthe software sales are custom software sales whereas only 10% is packaged software sales. 1
Size ofFirms
Nintendo is thelargest packaged software company in the worldat $5 billionannualrevenues. In custom softwarethe largest from Japanis NTT Data, a subsidiary ofNTT (three and a half to four billion dollars a year
annualrevenues). Therearetwo Hitachi subsidiaries thatare each over one billion dollars a year in sales. A subsidiary of Nomura Securities doing
1 For softwareuserather thanrevenues, the ratio may be less because so much use ofpackagedsoftware is from illegal unpaid-for copies.
software in the financial area is about abillion and a halfdollars a year in sales. The largestindependent softwarehouse in Japanisafirm called CSK, the parent of Sega. Not counting Sega's revenues, CSK itself is about a billion dollarsa year(software plus services).
Inpeople employed, the Japanese software industry firmsrange from the huge NTT Data downto what we would call garage-shop operations-threepeople, five people, eight people.
Problems ofthe Industry: Asummary
In January 1993, theJapanese government declaredofficially that the software industryisa "distressed" industry. Thatstatus allows theindustry to get certain kinds ofsubsidies and special treatment from the Japanese government. Everything Ilearned in doingthis study indicatedto me exactly the same thing about the troubled state ofthe software industry. The reasons are summarizedfirst, then discussedat length in the remainder of thepaper.
The Japanese industrylacks theproperinterplay ofmarketforces. It's
very much a narrow,focused industry that is not competitive on the world
stage. There is no honing ofJapanese software products in world market competition. There is no opportunity for the high volume sales needed to makepackaged software profitable.
Second, the industry has many different kinds of problems: the industry; the structure ofthe industry; the way the Japaneseculture treats software; andproblems in the educational systemsupplyinghumanresources forthe softwareindustry.
The officials at MITI, the Ministry of International Trade and Industry, refer to the software industry troubles as an "urgent" problem.2
The question is: Whyis theproblem urgent? What ismakingMITI nervous? What's making MITI nervous is a perception that is quite correct—that hardware items are becoming commodityproducts whoseprofit margins are eroding; that the highvalue-added software part ofthe computerindustry is theplacewhere thehigh profit margins are moving.
This is true not only ofthe saleof packaged software, but ofthe sale of software embedded in other products. Right now we are beginning to say
that "hardware is the box that software comes in." Softwareiswhere the real value is being added. The Japanese are worried about that. They look to a future, say past the year 2000, when software is really the place where the economic actionisin the computerfield.
Who We Interviewed
My study
3
was largely an interview-based study. We interviewed the large computer makers,Fujitsu, Hitachi,NEC and Toshiba. Also interviewed were packaged software companies. One ofthese was JUST Systems, the largest packaged software companyinJapan~sloo
billion revenues per year (compared with Microsoft's $4billion peryear). Anotherwas JapanLotus, a subsidiary ofLotus Development Corporation. I wanted to see how an American company operating in the Japanese packaged software industry saw the situation. IBM has established a newsubsidiary calledEncyclosoft, as a combination of software publisher and distributor, to try to enter the packaged software market in Japan. We interviewed this new company.2in the title ofa draft white paper circulated in 1993
We also interviewed two system integrators. System integration is playing an ever greaterrole in the software industry. Asthe industry moves away from mainframes toward the client-server architectures, the situation is
muchmore "plug andplay" and theusers are not knowledgeable enough to do
the"plug andplay" totallyon theirown. So theyhiresystemintegrators. We interviewed a Japanese system integrator, Argotechnos 21 (800 people, a spin-off of Nippon Univac) and Andersen Consulting, an American firm operating in Japan. We interviewed several small, independent software houses of size ranging from 15 to 80. To capture an academic view, we interviewed the Dean and faculty of a new engineering school ofKeio University. Looking to the government, we interviewed the Director of Division 3 ofMITI, the ElectronicPolicy Division; and the President andtop officers of a quasi-governmental promotion organization for the computer industry called JIPDEC—Japan Information Processing DEvelopment Corporation. In all, 20differentinterviews were done.4
Featuresofthe Industry
In contrast withtheAmerican situation, the Japanese large computer makers have established large numbers of subsidiaries in the software business. For example, Fujitsu has about 100 wholly owned software subsidiary companies. Thenumberof employees in asubsidiaryrange from about 80at the minimum to about 500 atthe maximum, with200considered by the Japanese to be an optimal size softwarefirm. Independent software houses serve these maker-subsidiary groups. The independent software
4How did I know that I didenough? I had a strongfeelingas we cametothe
end ofthese interviews that we werehearingthe same story over and over
againandatthatpointit looked like we hadreachedthepointofdiminishing
returns. We weren't learning anythingnew from new interviews. That made me feel very comfortable that we had essentially surrounded the issue.
house isthe lastfirmto getwork, andusuallygetsthe low-value-added work. The best work, the high value-added work, is donein the parent company itself. The Japanese have matched their liking for hierarchies with the hierarchies ofstructuredprogramming and mated the two together, saying "somepeople will dodesignandsomepeople will dotop-level codingandsome people will do lower-level coding." This is dysfunctionalinthesense thatthe people doing the lower level work never learn anything. They don't learn about how to do tasks at the higherlevel because they're never giventhose tasks. And the feedback loops are very noisy. It can take a long timefor information to get passed upfrom the coders (that some piece ofthe design won't work or some logical alternative wasn'texplored).
Much ofthe "custom" software produced bythe Japanesemayinfact be builtout ofhidden packages. Thebig system integratorfirms, including the big computer makers like Fujitsu may have libraries of large packages that they use when producing the next custom software system. Large custom software jobs may indeed have many "packages" inside. In this
countrywe would call them packages and sell them separatelybut in Japan they aresoldas integrated custom systems.
Japanese software distribution networks are not mature byAmerican standards. There arethree distributorsfor all PC shrink-wrapped software. Ifyou are not covered or handled by one ofthose distributors, you simply havenoway of sellingyour software. In the workstationarea I tried hardto find a similar network for workstation software and couldn'tfind it. I then went tothecompaniesthemselves andsaid, "How doyou sellpackages?" The answer was, "Well, number one—thepeople who sell ourhardware also carry oursoftware so ifyoubought aSun from C. Itoh (ahardware distributor),you
distribution system, ofa sort. It'sa phone-in service. You buy a piece of software by phoning up a number, telling them what you want; ifyou're in Tokyoyou get itdelivered the sameday, ifyou're anywhereelse in Japanyou getit deliveredtomorrow.
Iwent looking for Japanese databasesoftware. IBM sells a lotof DBII worldwide;Oracle sells about 1.8billion dollarsayear; otherlargeAmerican vendors areInformix, Sybase andso on. Whatarethe Japanese equivalents? The answer is~there is not a single Japanese equivalent. A buyer can get
database systems built by the system integrators or the large computer
makers but there is no separate market for Japanese database software. How about CAD software? With minor exceptions, the same is true of Japanese CAD software. The best selling CAD program in Japan is an American packageAutoCAD, and other Americanpackages dowell in Japan.
Problems ofthe Industry
Knowledgeable Japanesewho have studied the software industry told me(unanimously) that the Japanese industry would notbe able toprogress
unless itcompeted worldwide. Thefirms producing software had tohone up products. Unless there was some force to cause the introduction of new products,theindustrywould notbecompetitive even in Japan.
Theywould nothaveabig enoughmarket sizetocompete. In fact, the
market sizeis a very big problem forthe Japanesein the PC area. In this
countrywe havemanytens ofmillionsof PCs as a potential marketplace for anyparticular piece of software. In Japanthat numberis at least tentimes smaller! Compoundingtheproblem ofsmall market size, the competition in the JapanesePC industrytook a different form thaninthe United States. It
the marketfurther andcauses software companies to produce seven different versions oftheirprograms which adds expense.
That problemis going awaybecause ofMicrosoft. Microsoft jumpedin and said, "We will guarantee that Windows Version 3.1 in its Japanese version will be a commonplatform. That is, ifyou, the developer, will write for Windows Version 3.1, wewill guarantee itwill operate on all those seven
(otherwise incompatible systems)." Microsoft has created a sort ofacommon standard among the Japanese and are selling Windows like hotcakes in Japanbecause ofthat.
A major problem of the Japanese software industry is "bundling". Strange term. In the United States in 1968 the Justice Department forced IBM to sign a consent decree that saidthatit wouldnot include software in its hardware prices; softwarewould be priced separately. Thereason forthat wasobvious: you couldn't haveasoftware industry whenIBMwas producing all the software and giving itaway. A competitive software industry could not come into existence. Since 1968 the USA has had 25 years to build up a highly competitive software industry. It is almostimpossible to survive in the American software industry, it's so competitive. Microsoft and a few others are anomalies. Most companieshave avery difficulttimebecause the industry isso competitive. The Japanese havenosuch situation. Until afew years ago bundling was the norm. The large companies would bundle their software in with their computers and it was verydifficult for anyone else to survive, to compete in that industry. Then the government issued an "advisory", which normally is the equivalent ofthe governmenttelling firms whatto do. In thiscaseit didn't work. The advisory was "Don't bundle your software. Japan is supposed to goto unbundling." In 1993 the government
unbundle, we're going to take you before the Japanese Fair Trade Commission." Until the Japaneseunbundle software, there won't be any room for competition to arise to help invigorate the Japanese software industry.
The Demand Side
The independent software "body shops" supply workers for large software projects. The need forthe large number ofprogrammers from the body shops has decreased significantly because (1) there's arecession; ifa firm is going to save money, it'sgoing to savemoneyby not redoing asystem
that itmight otherwise have redone and(2) firms know now that they really don'twant to add new applications to mainframesystems. But they're not secure enough yet with the client-server idea to commit money to the reprogramming oftheirsystems for client-server applications. So they're just waiting. These factors have maderedundant at least 250,000 Japanese programmers from thosebody shops.
s
Finally on the demand side, the Japaneseare beginning to discover
(two or three years later than their American counterparts) distributed computing, client-server networks, open systems (the move away from
mainframes). The customers arebeginning to demand that. Theindustryis nervous about supplying it. They're unsure of their own level of skills. They'reunsure oftheirknowledge, and the softwarenecessaryto do the "plug and play" that you need for open systems. But customers are beginning to demand thistype of computing.
5 a bitter collection ofpeople, incidentally,drivingtaxis anddoingmenial
Illegal Copying
In my interview at JUST Systems, the largest Japanese packaged softwarecompany, the CEO mentioned the mainfactor inhibiting thegrowth ofhis company. Itwas notthe deep Japaneseeconomicrecession; norwas it heavy-handed governmentintervention. Itwas: the extent ofillegal copying of JUST's software! He said that JUSThas done measurements ofthis. A piece of software sold to a business is on average copied four to five times. Sold to the government, on average it's copied ten times. In spite ofwhat MITI says aboutbundling andunbundling, the Ministry ofFinance doesnot budget for government agencies to buy software. They only budget for computer hardware. This forces government agencies to pressure manufacturers to bundle their software. Otherwise, they scrape together some miscellaneousbudget tobuy a single piece of softwareandthenmakea largenumberofcopies.
One government official told me, "Japan is a copier's paradise." A contraview camefrom Japan Lotus. Theirviewwas: "We justlowered the price ofLotus 1,2,3. We had a price which was about $1000 taking into account the fact that so many illegal copieswere made. We justdecided to cut theprice in half. Wefiguredthat at alowerprice the Japanese willwant the manual andthey'll want ourservice, so they'll stop copyingoursoftware."
I interviewed another CEOwho told astorythat Ifound verypoignant. The man was the president of a small software boutique. He was the first personin Japan to produce a good word processor. He expanded that into a desktop publishing operation; now his firm does software for text-to-typesetting. When he introduced his word processing software many years ago, it became quitewell know. Everyone wanted it because itwasthe only good one. Heheardfrom afriendthat aretailer inthe electronics districtof
Akihabarain Tokyo wasgiving away his company's software on the street as a gift ifone bought an NEC computer. Sohewent to court to suethe people to stop that. As soon as he did so, the word gotout and his sales driedup.
The potential customersknew that they too makeillegalcopies; and thought actionmight be takenagainstthem. Furthermore, using lawyers is notthe Japaneseway of doing things. Sales driedup. He then backed off of legal action, and his sales picked up again. The man was deeply angry. He told me that Microsoft has now sent out 5,000 letters to the 5,000 largest purchasers of Microsoft software in Japan informing themrather bluntly, that Microsoft will prosecute to the fullest extent ofthe law anyone caught copying. My CEO friend said, "That's great—only the Americans can tell them that and we're very happythat Microsofthas done that."
The laws regulating copying are satisfactory. The government enforcement ofthe law is essentially nil; and anyway people don't enforce it for reasons I mentioned— dislike of going through the legal system. The
societal ethos apparently doesnotcontaina "don'tcopy"restriction.
HumanResources for theIndustry
In all of myinterviews in Japan I didnot find a singlepersonwho said that he was satisfiedwith the quality ofthe outputfrom Japan's university computer science or information science departments. Not one. Recently I have donemany similar interviews in the UnitedStates. I haveyettofind a single American software industry executive who said that he or she was unhappy with the output ofour American computer science departments. Thus, there is a major difference between Japan and the US in perceived quality ofthe university contribution. Furthermore, in Japan one sees far fewer students graduatingincomputer majorsbecause ofthelimitations that
the Ministry of Education puts on the flow of students through each department. A typical department of computer science has about 40 undergraduate students; a few Master's students; and maybe one or two Ph.D. students. Some departments have had dispensation to graduate 80, but that israre. The Japanese are trying tofix the problem ofinadequate numbers. They set up,for example, two bigcomputer sciencedepartments at new institutions, one nearKyoto and one atKanazawa. In each place, they are trying to hire 64 professors. Each department will have hundreds of students graduatingper year inthefuture. Untiltheycan trainmorePh.D.'s
(which theyhave notbeen doing) tobecomeprofessors in those departments, it'sgoing totakealongtime to achieve the goals: perhaps tenyears.
The Japanesesystemgraduates many people called programmersfrom trade schools and high schools; these are the people who are becoming increasinglyredundant, the real need now is for highly trained University
graduates.
CulturalFactors
Underlying all ofthe malaise in the software industry there is a problem that became evident after many interviews. The issue is deeply cultural—does the culture respect software? Is software "real"? Years ago, softwarewriters were called programmers; programmerswere paid less, and were treated at a lower level, than "real" engineers. The "real" engineers were electrical and electronics engineers, the hardware engineers, etc. As
softwarewriters cametobe ingreat demand, the Japanese firms changedthe name "programmer" to "software engineer" and changed the salary levels. But, alas, the status level didnot change. In the Japanesemind, softwareis not something real. Software is considered something like a service—
somethingthat comes alongwith the hardware. You expect it; it'ssupposed to be there; it makes the hardware run; but it's not distinct. It is not a product initsown right.
It is a kind of "vapor". You can'tfeel it. You can't lift it. Can you really sell it? Is one comfortablewith selling somethingthat is not tangible,
isnothard?
Is software worthy ofrespect as a field? Ifit's not worthyofrespect, is itworthy of investingyour moneyin it? Ifit'svapor, it's nothing. Wouldyou
investyour moneyinnothing? Wouldyouinvestyourcareerin nothing? Today it is hard to motivate young people to go into the software business. The Japanesedo not have afolkhero who made six billion dollars in thesoftware industry.
GovernmentActions
What is the government doing about the problems of Japanese software? Theyhave issued that directive on unbundling. Theyhave issued subsidies to the packaged software industry. Good ideas for packaged software can winlarge government grants. This is not venture capital. The
government doesn't take equity. MITI is pushing the Ministry ofFinance to introduce software line itembudgets intothe agencies, toreduce the amount of copying. The government is taking astrong standon copying, instructing agencies not to copyillegally. The government isjawboningbusinesses about not copying software illegally. The Ministry ofEducation, as I mentioned, is starting more computer science departments. M.O.E. is also trying to introducemorePCs intothe schools.
The Japaneserealize that value is moving from hardware tosoftware; but they feel more comfortable selling software in a substantial package. They have invented a term that describes their new attitude towards software. The term is pure, wonderful Japlish—the "new hard". The "new hard" is thenew way toview hardware: "hardware is the box thatsoftware comes in." Think ofa function: imaginethe software capability that's going torealize that function; then invent a box that actually does the job. You embed software in the box; then sell the box, not the software. This is becomingan increasingly common practice intheUS. Atarecent meeting at Stanford, Joel Birnbaum, the Vice President for Research of Hewlett Packard, told abouta newHP stethoscope. Itwill havea microcomputer init
thatwill dosignal processing ofthe sounds that the doctor ishearingandwill be able to give intelligent readouts. That's an example ofthe way that an intellectually andeconomicallyvaluableproduct gets embodied in something that is actually hardware. What the customer buys is an intelligent stethoscope, not a piece of software that is read into a stethoscope. The Japanese plan tomake amajorbusiness outofthisparadigm.
The Japanese seemultimediaas a majorfuture trend. There is great uncertaintyabout "when" and "how much" butnot about "if. In aninterview I asked one ofthe Executive Vice Presidents of Fujitsu: "Given that IBM's mainframe business is declining and Fujitsu's business is modeled after
IBM's, what is Fujitsu's future? Theanswer was "multimedia."
The traditional large Japanesecomputer makersareworried about the upcoming confluence of consumer electronics, communications and computing. Theyworrythat thefirms thatwillmaster this newworldwill be thefirms who have mastered theconsumer electronicsindustry in the past-Sony and Matsushita, specifically. Sony, indeed, is a company considerably
skilled in computer science and technology. Nintendo is also a potential winner. A Nintendo game computeris the computer that's present in more Japanese and American homes than any other computer. If Nintendo's alliances withSilicon Graphics orsome other firm bear fruit, Nintendomight become amajorplayer inthenew eraof computing.
Conclusion
I wishtoconclude withahypothesis about how the Japanese software industry is going to evolve. This is anhypothesis ofseveralfaculty members ofthe Stanford Computer Industry Project.
Webelieve that,on the onehand, thelarge computermakers in Japan
(like Fujitsu) will circle the wagons around their customer bases. These companies andtheir customers are not going tomove away frommainframes rapidly. At allcosts the companieswill protect their cherishedrelationship with the customers in their customer bases. Hence, they will vigorously oppose open systemstrends which allow customers to mixand match andbuy hardware andsoftware from many differentvendors. Here is a good example of protecting the customer base: Japanese customers seem to want to buy Sunworkstations. Fujitsu offersitsversion ofthe Sun workstation, licensed through Sun. But the workstation says "Fujitsu", not "Sun". Ifyou're a Fujitsu customer, youbuy theFujitsu box, not the Sunbox because Fujitsu doesn't want Sun relating to its customers. It will be very difficult for American firms to break into the domestic marketplace because the relationships the Japanesevendors have with their customers are of
long-standing duration and are difficult to disrupt.
The secondpart ofourhypothesis is that there will be start-upfirms, smaller firms that believe they can prosper outside Japan, that will
energetically pursue markets in the United States and elsewhere on a Nintendo-like model. Somehow Nintendo made it. Somehow Nintendo introduced aproprietarystandard;set upa distributornetwork; madeforeign alliances, andso on. There probablywill be some nicheproducts, defined by Japanese companies, that make dents in the American market. These companies may doverywell, asNintendo did. Except for Nintendo there is absolutely no evidenceforthisportion ofthehypothesis. Timewill tell.
My colleagues and I feel that American firms will do very well competing with the Japanese in the packaged software industry where the US currently has a dominantposition on theworld scene. The US packaged software industry has a 75% market share worldwide. But Americanfirms should not set their expectations too high for penetrating the Japanese marketfor bigbusiness software. The alliances andrelationships which the Japanesehave developed over a long period oftime are very durable, and
