fashion nova los angeles
he holds academic appointments at the universityof cal, berkeley--in several departments. has his bachelor's degree from mit in 1969--sameyear i got my bachelor's degree from michigan state, hal.our football team would beat your football team--[laughter]--back then.but that's about all, probably, would beat mit, i must say. [laughter]ma in mathematics, phd in economics from uc, berkeley in 1973.he's taught at mit, stanford, oxford, michigan and other universities around the world.he's a guggenheim fellow, member of the econometrics society, american academy of arts and sciences,a co-editor of the american economic review from 1987 to 1990.holds several honorary degrees.
he's published numerous papers in economictheory, industrial organization, financial economics, econometrics, information economics.author of two major economics textbooks which have been translated into 22 languages.he's the co-author of the best selling book on business strategy called information rules:a strategic guide to the network economy. he wrote a monthly column for the new yorktimes from 2000 to 2007. it gives me great pleasure to introduce dr.hal varian. [clapping from audience] >>varian: thank you. [pause]
thank you very much for that kind introduction.i'm very happy to be here, particularly since it's only about 10 miles from where i live.i didn't have some of the commuting issues that you may have had.i want to start this morning by talking--talking a little story about some duck hunters whowent hunting up in northern michigan. and when they got there, they decided theyneeded a hunting dog, so they went to the dog rental outlet, and they were all out ofdogs. he says, "well, i've got one here that justcame in. it's a young dog--almost a puppy--we call him assistant professor. [laughter from audience]now, he only rents for ten dollars a day,
so he's a pretty good deal."so they rented assistant professor, went out, and he was a very enthusiastic dog.he ran through the forest and jumped into the water and fetched those ducks.didn't quite know what he was doing in some cases, but he performed enthusiastically.so the next year when they came back, they said, "let's get that assistant professordog again." and they said, "well, i'm sorry.we named him associate professor. now, he's twenty dollars a day." so they rented associate professor, and hewas great. he was--not only was he enthusiastic, he was really
pretty skillful by this time.he was able to fetch all the ducks, he was able to swim in the cold streams, they threwhim a few scraps of food at dinner, and he was happy.so the third year they come back--and i think you could see this coming--and said, "well, i'm sorry. we renamed him professor, and now he's thirtydollars a day." so professor--well, he was little less reluctant to jump into that ice cold water.and he demanded more than a few scraps at dinner time.so the next year they came back, they mulled it over, and finally decided let's do that--forold time sake--let's get that dog professor.
and they said, "well, i'm sorry.you can't have him. not at any price."why is that? "well, we renamed him dean, and now all hedoes--" [laughter from audience] "all he does is he sits and begs." [laughterand clapping from audience] so-- well, as you can tell from that story, i wasonce a dean myself. and after seven years of labor, i got timeoff for good behavior, and i thought it might be fun to go try something in the real world.but then the real world seemed kind of scary, so i decided to go to google instead.so that was a great decision, because when
i went to google, there were about 400 people.that was in 2002, and i was there to watch it develop from 400 to 20,000 people thatwe have today. so what i want to talk about is a little bitabout the way we do things at google, and that's for two reasons.one, i think it's interesting--there are interesting models there that can be exported to othersituations, including the situation you find yourself in.and even for those aspects of the company that can't really be exported, i think they'reinteresting from the scientific point of view to understand how organizations can work inthe modern age. now google's corporate mission is to managethe world's information to make it universally
accessible and useful.so our employees are, quite literally, information workers.they don't work with steel and glass. they work with information.or maybe to use a term with a slightly longer pedigree, they're knowledge workers, in thesense that peter drucker described knowledge workers way back in 1959.these are the computer engineers and the project managers that build the products and toolsthat allow google to manage the world's information and, by extension, knowledge.now, of course, you as deans, you also deal with knowledge workers because the facultyat your universities organize information, create and impart knowledge, and helps usto better understand the art and science of
management.and as peter drucker observed those 60 years ago, managing knowledge workers requires aparticular approach--that organizations that can do that successfully have a huge competitiveadvantage. now, i like to think about this from the historicalperspective. in the first decade of the twentieth century,industrialists like henry ford and frederick taylor developed techniques of scientificmanagement, which in those days, of course, was primarily focused on manufacturing.the factory system that was pioneered by ford and by taylor has been criticized for itsemphasis on efficiency at the expense of workers, but the fact remains that it was these innovationsthat resulted in this tremendous productivity
boom that we saw in the twentieth century.economic historian, paul david, has a very nice description of how and why we got thisproductivity boom. as he indicated, back in the beginnings ofthe nineteenth century, factories were mostly run on water--there were water wheels.and you had a big shaft down the center of the factory.and then all the tools ran on a belt that was connected to the shaft.you had the drills and the lathes and the saws and everything scattered throughout thefactory, but primarily you tend to put all the drills together and all the lathes togetherand all the saws together, and then the work was carried back and forth from station tostation to run the assembly.
now, when electric power came along, the firstthing that happened is they replaced the water wheel with a big electric engine.and so the electric engine drove the shaft, and that gave you some productivity gains,because when the water spring dried up, you could run the electricity motor.and when the creek froze or the water froze, you could still run the motor, and so youwere able to produce more days of the year. but it didn't really change the way productionwas organized very much. and then as time went on, the electric motorbecame miniaturized, and so you'd put an electric motor on each tool.and this gave you the flexibility to move the tools around.but in general, that flexibility wasn't exercised.
why?because everybody knew you always put the drills together and the lathes together andthe saws together. and it really took henry ford to get the brightidea--the assembly line-- that you could re-arrange the tools in a way that would accelerate theproduction process.instead of clustering them together, you put them in a line, and you move the product downthe assembly line, and we're able to engage in manufacture at a much, much more efficientway. so during this period in the early part ofthe twentieth century, ford and the entire management team were down on the factory floorevery day, tinkering with the assembly line--lowering
this, raising that, speeding this up, slowingthis down, building a new jig, etc. and, of course, that continuous optimizationended up dramatically improving productivity of manufacturing.and today, of course, manufacturing is hugely more efficient than it was a hundred yearsago. now, just as the miniaturization of the electricmotor allowed you to improve the efficiency and the organization of assembly work andthe flow of production across the factory floor, the miniaturization of the computerhas improved the organization of knowledge work, because it's made information deliverableto the worker, to the employee that needs the information at the time and place thatthey need it.
so just as the flow of product was improvedby the assembly line--the miniaturization of the electric motor--the flow of informationthrough the organization is improved by the availability of the pc.in the 1990s we saw the significant productivity improvements from computers and local areanetworks, which allowed the knowledge work to be distributed efficiently in the organization,and now we're seeing the further improvements that are available from the widespread availabilityof the world wide web and the global internet. so if you go back and look at the tenets ofmodern manufacturing that really emerged from this assembly line, mechanical work model,they all have their analog or extension in the knowledge work side.i want to talk a little bit about that.
so the six principles that we see in modernmanufacturing are routinization, modularization, standardization, continuous production, continuousimprovement, and miniaturization. so each of these principles translates toknowledge work in--i think--a nice way. so we already talked about the miniaturization,the standardization and the modularization--those are integral parts of information technologyin general. and nowadays, in many organizations, you seethis kind of continuous production because production is world wide.and in fact, one very important benefit of this kind of miniaturization and the commoditization,i guess, of information technology is the rise of what i call the "micromultinationals."so there are thousands of companies with dozens
of--small companies with only a dozen or soemployees that are scattered around the world. i've run into friends here in the bay area.i say, "what are you doing?" they say, "oh, i have a company."i say, "how big is your company?" "well, we've got two people in italy, threepeople in india, four people in boston, and then two people here in san francisco."and you can do this kind of micromultinational because you have access to technologies thatonly the largest multinational corporations had access to ten or fifteen years ago.so these micromultinationals can use e-mail, skype, wikis, chat, google apps--all thesevery inexpensive forms of communication that are primarily available at no cost, and they'reable to coordinate the production of work--the
knowledge work that they engage in--in realtime around the world. so communications technology that was onlyavailable to the mega corporations fifteen years ago is now available to the smallestgarage startup. now, small companies represent 99.7 percentof all firms. they create more than half of private, non-farm,gross domestic product, and they create 60 to 80 percent of net new jobs.so the fact that information technology can be made available in this very efficient wayto these tiny firms, i think, has huge implications for the productivity of the economy goingforward. now, routinization--they talked about a minuteago--that's also a very critical force.
edison reportedly said that invention is onepercent inspiration and 99 percent perspiration. well, at google, we believe in cyber sweat.so the idea is the people should have the inspiration, and it's the computers that shoulddo the perspiration. if a task can be routinized at all, a computershould be doing it. continuous improvement--i talked about a minuteago--is not only due to globalization, but there are also other forces at work.so traditionally, continuous production applied to the assembly line model i discussed earlier.back in the 1910s, henry ford and others were tuning the assembly line.they aimed for continuous improvement in the manufacturing process.every day they wanted to make it better.
and if you look at the japanese method ofmanufacturing that was touted in the u.s. in the 1980s, they were talking about kaizen--kaizen,which means continuous improvement. and with web-based production--producing thingsusing information technology--you can do the same thing, only it's much, much easier becauseyou're working with bits, not with steel and glass.so go look at google. at any one time google is running literallyhundreds of experiments. so there are hundreds of experiments goingon, where you're tinkering with the search algorithm or the add algorithm or the userinterface. in some cases, in ways that are so subtlethey barely register with the conscious mind.
you might see that if this interface is alittle bit different, but you don't quite know how it's different or why it's different.so each of these experiments is a carefully designed treatment control model, so you comparethe new proposed treatment--the new change in the algorithm--the new change in the searchalgorithm to what you had before--and after a few days, you discover if you've reallymade an improvement. if you made an improvement, then you rollit into the product, and the product is improving, literally, on a week-by-week--and in somecases, a day-by-day basis.in fact our big problem now--our worry is that someday it might become conscious, butwe do have an agreement.
the governor of california, arnold schwarzenegger,said that he'll step in if this happens, so we hope he keeps his word on that.now, as for standardization, it goes without saying that information technology is builton standards. so when we wrote information rules, we talkedabout the huge profits accrued at a company that controlled a proprietary standard, dueto the network effects and the eclectic switching costs that these standards conferred.well, somebody must have been listening because today it's very, very hard to build a businessaround a proprietary standard simply because people recognize those threats, and they wantto defend against the--ceding control to a company that built as their business modelaround proprietary standards.
and google, of course, is built on open standards.we use open source software almost exclusively and support its development both financiallyand by contributing code. now, why is that?well, the easy answer is it's free, so if you have to scale to arbitrarily large operations,it's nice to have a free operating system or free code.but the more important consideration is that you don't want to allow a core piece of yourbusiness by--controlled--by someone else. and so by using open- source software, we'reable to run our business the way we would like to do it, rather than being beholdento an outside provider. okay, well i've talked a bit about the technologyside of knowledge work, and now i want to
talk a little bit about the management.so as peter drucker explained many years ago, there's some special cares and special issuesinvolved in managing knowledge work. and i want to talk about some of the principleswe use at google. and again, i'm not necessarily saying thatthese could be exported to every business, but i think they're worth looking at.so the first one is hire by committee. so every person who interviews at google talksto--at least half a dozen interviews--usually more--drawn from both management and potentialcolleagues. and everyone's opinion counts in this interviewprocess, making the process more fair in pushing the standards higher.now, it's true that it takes a lot longer.
that's the big gripe about it.but if you hire great people and you involve them intensively in the hiring process, youget more great people. so it's setting up that positive feedbackloop way back when the company was founded, and that's ended up having a huge payoff.one kind of subtle side benefit is because people have gone through this extensive interviewprocess, by the time they start to work, they're pretty much trusted to be able to performat a high level, and so there isn't the kind of testing or trial period that you see insome organizations. the second thing is we cater to their everyneed. so peter drucker also said this.he said the goal is to strip away everything
that gets in their way.so we provide a standard package of fringe benefits, but on top of that, we have first-classdining facilities, gyms, laundry rooms, massage rooms, hair cuts, car washes, dry cleaning,commuting buses with internet connectivity on all the commuting buses--just about anythingthat a hard-working engineer might want. let's face it, programmers want to program.they don't really want to do their laundry, so we make it easy to do the laundry so theycan do what they want to do, which is the programming.third thing is we pack them in. so almost every project at google is a teamproject. the way things get done is with small groups--fiveto ten people.
and the teams have to communicate, and thebest way to make communication easy is to put the team members within a few feet ofeach other. so google has a cubicle culture, like mostof silicon valley. we have cubicles with four or five people.you don't have your private cubicle as they do in some organizations.your cubicle is shared with other workers who are working the same project.so when you want to know what does this variable do or why was the code structured this wayor where do i fetch this piece of data, you turn around to the guy behind you and youask him. so it's something where the communicationflows, you don't have telephone tag, you don't
even have e-mail delays.you're able to get the answers to your questions--immediately. so when a programmer needs to confer witha colleague, the colleague is right there. now, of course, there are also many conferencerooms where people go to have private conversations or to talk, because you don't want to necessarilydisturb your co-workers if it's something that doesn't involve them, but it gives youthe best of both worlds--immediate communication and, when necessary, you can have this extensiveverbal communication. in fact, i will tell you that even the ceoat google shared an office for the first several months he was there.in fact, what happened is he was assigned a private office, but one of the engineersnext door felt that his office was too crowded,
so he picked up his computer and moved intothe ceo's office. and when the ceo came back, he saw this guysitting in his office--didn't know quite what to do--he introduced himself.and he said, "i'm eric schmidt, the ceo, and the other person gave his name, and he said,"i'm the chief lumberjack." the chief lumberjack was actually the personthat managed the logs, right? okay.so--what was great about it is that was--he was employee number seven, he'd been therefor a long time, and so every time eric had a question about how something worked or whatsomething was done, his office mate was able to explain how that particular piece of technologyworked.
so, in fact, the ceo sharing an office--atleast for this initial period--that really helped him get up to speed very quickly andget a deep knowledge of what was going on in the company from pretty much a direct lineto people who managed it. so it was an incredibly effective educationalexperience, and i think both parties agree to that today.all right. make coordination easy.so because all members of the team are within a few feet of each other, it's very easy tocoordinate projects. and in addition to this physical proximity,each googler e-mails a snippet once a week to his work group, describing what it is thathe accomplished in the last week.
so those snippets are up there on the web,and you can look at what everybody's doing, see what their--happening.it's easier to monitor a progress towards completing projects and synchronize the workflow, because you can tell what it is that's going on.in addition to the snippets, we also have okrs, which are objectives and key resultsthat are put together once a quarter, where you list your objectives and key results forthe coming quarter. and then at the end of the quarter, you gradethose--see how well you did in accomplishing those objectives and key results--and setup the okrs for the next quarter.so again, all of this is online.
all of it can be looked at by your colleagues,your co-workers, your managers. even senior management in the company looksin on these okrs periodically--to just get a feel for how the work is flowing in theorganization. next principle--eat your own dog food.i guess i'm getting back to the northern michigan theme here.google workers use the company tools intensively. now, the most obvious tool, of course, isthe web, and we've got an internal web page for virtually every project and every task.they're all indexed, they're available to the project participants on an as-needed basis.and we've developed many, many other organizational tools which are rolled out as products.so you look at gmail--
i don't know how many of you use gmail, butit's a fantastic e-mail system. and the reason i think it's so good is itwas tested extensively internally. so it had to meet the standards of this very,very demanding set of customers, mainly the workers--the knowledge workers at google.and google docs, for example, has been fully integrated in the work flow at google, soyou have multi-author documents. again, in your environment, i know many documentshave several authors, and we've got this rather arcane system of sending files back and forth.and you have version control problems, you have tracking problems, you have people wholoose them or don't get them submitted on time.well, of course, a much better model is to
have a single copy of the document sittingup there in the cloud, and then people can modify that document in real time, you cantrack the changes, you can undo anything you don't like, you can add notes to it.so you've got a much more efficient way to produce multi-authored context.that's the same thing, whether it's a word processing document, a spreadsheet, or forthat matter, a piece of computer code. encourage creativity.so google engineers can spend up to 20 percent of their time on a project of their choice.now, there's an approval process and a bit of oversight, but basically, we want to allowcreative people to be creative. one of our not-so-secret weapons is the ideasmailing list.
so there's basically a suggestion box or anideas mailing list, company-wide, where people post ideas that range from parking proceduresto the next killer application. so the software allows everyone to commenton and rate ideas and provides a way for the best ideas to percolate to the top.now, people are always a little surprised about this 20 percent time.they say, "well, what can one person working alone accomplish?"and the answer is one person working alone can accomplish a huge amount, and oftentimes,they'll co-opt a few other members to come into their project, and then they can accomplishtruly amazing things. so some of you may have used google scholar,which is the--what we call vertical search
engine--it's searching scholarly material.that was developed by one engineer in his twenty percent time over a period of aboutsix months. and when it came down to the final month beforerelease, he managed to get an intern to come in and help him out--do the final processesto get the release out. so i asked the person, "how is it that youwere able to develop this incredible product virtually alone?"and he said, "because i have this extremely powerful infrastructure to draw on."because the infrastructure was available to each engineer--to each person who wants touse it, without getting prior approval or committee sign-off or anything like that--theycould use that infrastructure to develop projects.
and then if the project's successful, theycan argue for deployment, and we see, as a result, these fantastic projects like googlescholar, which are developed by a very small group of people and, yet, have very, verypowerful results. strive to reach consensus.so the modern corporate mythology has a unique decision-maker as a hero--ceos kind of elevatedto this mythical position. but we adhere to the view that the many aresmarter than the few, and at google, the role of a manager is a person who aggregates viewpoints,not a person who dictates decisions. so building a consensus sometimes takes longer,but it always produces a more committed team and, generally, better decisions.next point.
data drives decisions.so at google almost every decision is based on quantitative analysis.we build these systems to manage information, not only for the internet at large, but alsothe internal information. and we have dozens--actually now--hundredsof analysts who plough through the data, that analyze performer metrics, they plot trends,they do experiments. we have roughly two hundred statisticians,econometricians, operations resource, technical analysts sorts of people, and we do this,often summarizing this information in online dashboards for every area we work in, so managementcan get a snapshot of where we are in pretty much every dimension of the corporation.so ultimately in business, there's two ways
to make decisions.you can use hppos, or you can use data. now what's a hppo?hppo is a highly paid person's opinion. [chuckling in audience]so the trouble is with highly paid people, you don't want them to sit around and debatewhether the best background for an ad is yellow or blue.you want them to say here's the experiment we should run.we'll let the data drive the decision, and now we can move on to the next topic of theagenda. and indeed, just a few months ago, we rana very extensive set of experiments about what the best color is for the ad backgrounds,and it turns out--in case you ever need to
know--the answer is yellow.so now, i want to expand a little bit on that data drives decisions, because that's reallythe area that i work in at google. several years ago i had a group of recentmba graduates, from top institutions like yours, that came to me and they said, "wewant some classes." and i said, "what kind of classes do you want?"and they said, "we want classes in databases and statistics."now, when i tell this to my colleagues in business schools they burst into laughter,because everybody knows those are the least popular classes in any mba curriculum.but in this case, it was absolutely clear to these recent graduates that this is whatthey needed to know, because if they went
in to an engineer and try to get the engineerto go to the database and pull their data, or they had to go find a statistician to tellthem how to conduct a t-test, they would never be able to deliver their project on time,because management demanded use of the data, the analysis of the data, and scientific proceduresto analyze that data. so i ended up teaching a little course.we called it statistics for people who've forgotten statistics, and we got some otherpeople to teach the database course, and now the mba students are quite happy, becausethey have access to the raw material that they need to help drive the decisions thatmake the organization move forward. so at google, if you want to succeed in yourjob, you have to be able to do these things
yourself.so i want to remind you that some of the things we do at google can't be exported to othermedia. i don't think your faculty would necessarilywant to share cubicles. although, i have to tell you we had some visiting--wehad a visiting computer scientist from berkeley--a friend of mine.when he went back to his department at berkeley, he rounded up some colleagues, and they agreedto tear out a corridor of their building and turn it into cubicleland, rather than usethe traditional private office approach. and so they have these cubicles working therewith their graduate students, because in general, their model is to work on team projects whereyou have four or five people involved in each
project, and he says it's been a tremendousboost in productivity in the research work that they're doing.and now, they're thinking about tearing another corridor.so it may well be that for some kinds of jobs that is a much better operating environment,particularly jobs like the computer systems development that i was describing at google.on the other hand, i think that there are many mbas that you're turning out that aregoing to be working for companies where data is widely available, and data can confer competitiveadvantage if you are able to analyze it successfully. so my view is that maybe putting a littlemore emphasis, either from a curriculum side or at least from the propaganda side, couldbe helpful in terms of turning out managers
who could be successful in this information-drivenculture. so even if the universities can't or won'temulate some of these aspects of google, i think that they're still very worthy of study,because these are the kinds of environments that many of your graduates will be workingin. so let me stop there.i'm going to open the floor up for questions, and thank you very much for your attention.[clapping from audience]
0 komentar :
Posting Komentar