The Net of R&D

Paul Kedrosky points to an interesting slide in Microsoft CTO Craig Mundie's recent analyst day presentation. The subject is R&D, the point Craig is making is that Microsoft is way outspending Google and others. Paul points out: Compelling, right? Microsoft's spending heavily on the Next Big Thing, while its…

Paul Kedrosky points to an interesting slide in Microsoft CTO Craig Mundie’s recent analyst day presentation. The subject is R&D, the point Craig is making is that Microsoft is way outspending Google and others.

200338426 9Af01Ae483

Paul points out:

Compelling, right? Microsoft’s spending heavily on the Next Big Thing, while its layabout competitors, you know, aren’t.

Well, I’m not so sure. Google has added over a $100-billion in market capitalization during the period, while Microsoft has shed around $30-billion in market cap. Similarly, Apple has added around 30-billion in market cap, while IBM has shed around $20-billion.

If you were of a grouchy frame of mind as a long-suffering shareholder, you could use this slide to argue that Microsoft overspends on R&D and investors would be better off off if it spent way, way less.

To which I’d add: I wonder how MSFT got these figures. Given that Google pretty much runs its engineering department as an R&D lab (ie, you can work on whatever you want at least 20% of the time), I’m guessing these figures are a bit off.

13 thoughts on “The Net of R&D”

  1. From what I hear that 20% number is quite exaggerated. But even if they had $5B in salaries (which they do not), 20% only adds $1B per year which still puts them well below Microsoft. If that Microsoft number is accurate, Microsoft spends more on R&D than Google spends. On everything. Sure Google added $100B to their market cap, but what have they done for me lately (last 6 months)? And the point of R&D: what are they going to do for me in the future?

  2. I always thought this 20% figure was just some PR trick.
    Engineering software is not a job with fixed 8am-5pm work hours and we always have much less time that what we actually need.
    So given the fact that hour working hours are “flexible” and that we’re always in short of time where does these 20% come in?

    I mean, I work at SAP and besides my regular tasks I can always find new ideas and proof-of-concepts to tryout and can spend the extra time to actually try them out. Is this extra effort count as 20%? and 20% from what?

    Software companies thrive on innovation and good engineers always have new ideas to look into. If a company actually needs such a “policy” then maybe it needs new people…

  3. I think they have just pulled the figures out of the air. Unless of course they are classing MS’s acquisition spending as R&D.

    As already said if they are correct then MS is overspending for what they are getting in return. Perhaps they need to have a closer look at where all this money is going.

  4. A couple issues ago Strategy & Business had a cover story on the results of heavy R&D spending:

    The jist:

    “Not so fast. The results of our recent study of the Booz Allen Hamilton Global Innovation 1000 — the 1,000 publicly held companies from around the world that spent the most on research and development in 2004 — may provoke a crisis of faith. The study, which we believe is the most comprehensive effort to date to assess the influence of R&D on corporate performance, suggests that nonmonetary factors may be the most important drivers of a company’s return on innovation investment (ROI2). The major findings:

    “Money doesn’t buy results. There is no relationship between R&D spending and the primary measures of economic or corporate success, such as growth, enterprise profitability, and shareholder return.

    “Size matters. Scale leads to advantage. Larger organizations can spend a smaller proportion of revenue on R&D than can smaller organizations, and take no discernible performance hit.

    “You can be too rich or too thin. Spending more does not necessarily help, but spending too little will hurt.”

  5. To echo Eran’s comment above, a Google engineer once told me that his 80% time often takes 120% of the time. So the only way you get to actually use your 20% time is if you stay late and/or use your own time.

    But no matter.. let’s suppose that everyone really did get 8 full hours per week to work on any idea they wanted to. In that case, there is still a huge difference between what Google is doing and what MS is doing. Google treats research as an engineering task. And thus really only comes up with engineering solutions. They see some problem that’s slightly broken, so they engineer a slightly better solution.

    With MS on the other hand, they’ve allowed funding for more pie-in-the-sky, long term projects, such as those that used to happen at PARC and Bell Labs. This longer-term research is not as connected to the parent corporation’s current business plan.

    I’ve heard Google criticize the long-term research approach, citing this “disconnectedness”. In treating research as an engineering problem, I’ve heard Google say, they are able to do work that has more of an effect on both the world and their business.

    However, I worry that this engineering approach is shallow and short-sighted. If you’re always only working toward a solution to a problem that users are having today, you never have any time to dream up the new, creative, big ideas that change the future. To use a machine learning analogy, Google is really good at iterative hill-climbing. I.e. they are masters of the beta/feedback/rapid release cycle. But the problem is that hill-climbing is only going to get you to a local maximum, rather than a global maximum. You will only get to the top of your current hill, and won’t ever realize that maybe there is a better, larger hill that you should have been climbing instead. Finding (and climbing) those larger hills is the advantage of MS-style longer-term research.

    A valid point in favor of Google is that, too often, the parent company that invents the new, groundbreaking technology is not in a position to take advantage of that technology commercially. On the other hand, I think the last eight years have shown that Google really is struggling with the local maximum problem. What have they done that is really earth-shatteringly new, and not an incremental copycat of some already-existing service?

    Don’t get me wrong.. I think the incremental improvements are very nice. Gmail and maps on ajax are beautiful. Having a spelling-checker while searching is quite helpful. And obviously took know-how, creativity, and skill.

    But they represent the tops of very low hills, relatively speaking. And I am hard pressed to see how giving 1000 employees their 20% time is going to lead to a radical jump to some bigger, new hill. With 1000 people @ eight hours a week, you won’t get big, incremental leaps. You’ll get 1000 incremental coding hacks.

    No, frankly, I’d rather have 200 people dedicate 100% of their time to research, rather than 1000 people dedicate 20% of their time. And I think this is the approach MS takes. (MSR Asia + Redmond + Cambridge have hundreds of researchers.) They actually give people time to think up new and interesting, “creative leap forward” problems.

    My point is that I actually can believe that MS is spending much more on research than Google, given Google’s attitude toward research. Google really thinks of it as engineering, not research.

  6. John, those figures are easy to understand.

    MS produces a zero-cost marginal “product”, therefore their R&D is not only research, but also _development_ of the final product, but also “production”. That’s the figures they show: their research, development and “production” costs.

    The two followers, IBM and Sony are doing also real manufacturing (i.e. high marginal costs’ products), so their real investment in “research and development” is proportionally much higher than Microsoft’s.

    MS just like to ignore these facts, or they stupid to confuse “software development” with the D of R&D, or they think they can cheat all the others, all the times.

  7. So, I work at Google, and I see an amazing amount of research going on. The entire company is staffed with people with academic backgrounds in disciplines like computer science, computer engineering, mathematics, and so on. To imagine that we’re not doing research constantly seems bizarre to me. The question of “yes, but how much basic research” you’re doing also seems weird to me. When running your company involves solving fundamental problems in computer science and mathematics, that’s what you do as your bread and butter.

    If Microsoft is really going to throw up charts and graphs, it’d be interesting to see their cumulative spending on R&D in their sixth year of existence as compared to Google’s. As for the particulars of this graph, it appears to imply that Microsoft has spent nearly 40 billion dollars on R&D (cumulative) to produce a business that has about 40 billion a year in revenue. Meanwhile, Google has spent about 2 billion for a business that has about 10 billion a year in revenue. Apple has about 13 billion a year in revenue.

    What those numbers mean is I guess left to the investors pouring over the tea leaves, but I do think it says at least a little something about the efficiency of traditional corporate R&D in terms of return on investment.

  8. Random Googler writes: “The question of “yes, but how much basic research” you’re doing also seems weird to me. When running your company involves solving fundamental problems in computer science and mathematics, that’s what you do as your bread and butter.

    I respect the fact that Google is working on some very difficult technical problems. Kudos for all the effort that went into things like MapReduce and Google File System. But with MapReduce, for example.. how much of that was “basic” research, and how much of that was taking research theories from 15-20+ years ago, and being the first company to apply those ideas on such a large scale? My understanding is that it was more the latter. Oh, it is an extremely worthy accomplishment, don’t get me wrong. But it is an engineering and development problem, not a research problem. The functional programming ideas were already there. So where is the research to replentish your stock of scientific ideas? I realize that these technical problems are your bread and butter. But bread and butter is for the eating. Where are your seeds, for the replanting?

    About half a year ago I heard Google’s director of research, Peter Norvig, on an IT Conversations podcast talk about a recent PhD hire’s experience there. The new hire wanted to do research, to publish papers, to grow the realm of interesting ideas. To create more seeds. Norvig’s response (I hope I paraphrase this correctly — it is memory and not malice if I don’t) was that it was much more interesting to put incremental ideas into the hands of the general public. Some time later, the new hire concurred, and not only was in the process of launching some new product that week, but also said he never wanted to go back to that “old style” of research.

    You mention the hordes of academics who have joined Google. I know, they’re there. But if they’re all busy launching products, who is creating the seeds for the next generation? Is someone working on that, in their 20% time? Really, from the outside, I see academics go in, but never come back out. It is like a research black hole. Even Yahoo Research Labs publishes orders of magnitude more work than Google.. and relative to Microsoft, YRL is basically the same age as Google.

    Maybe I’m just not understanding well enough what Google does. Maybe there are new fundamental seeds you are planting and I’m just not quite seeing it. Well, you need to do a better job of marketing because it is hard to see the big leap forward with Google Answers, Google Base, Google Bookmarks, Google Calculator, Google Calendar, Google Catalogs, Google Chat, Google Click-to-Call, Google Co-op…[and]…Google Video, Google Web Accelerator, Google Zeitgeist, etc.

    Random Googler, the last line of your response just about sums it up: “I do think it says at least a little something about the efficiency of traditional corporate R&D in terms of return on investment.”

    With all this focus on ROI, on what is profitable now, you realize, don’t you, that the sort of research you are talking about is naught but us eating our seed corn. A recent Ars Technica article says it much better than I ever can: “There’s no doubt that the information economy continues to create a lot of wealth, but I think it’s fair to ask if it’s also creating enough science to replenish the stock of scientific capital that it’s still burning through. I think it’s clear that chaotic, market-driven change is a good way to bring ideas quickly and efficiently from concept to profitable product. However, such a rapid churning of the institutional and cultural landscape ultimately may not be conducive to the kind of steady, expensive, long-term investment in fundamental research that produces the really big ideas that somewhere, at some completely unforeseeable point in the future, change the world.

    It is just my fear that, because of the overwhelming short-term business success of models like Google’s, the rest of the world will adopt this approach to research (academics in, no publications out — and the research that is done is only done with an eye toward ROI), and we will collectively end up eating all of our seed corn. The next “two guys in a garage” will not be able to come up with the next PageRank, because there will not have been a “Eugene Garfield” (the creator of citation analysis) publishing work upon which they can build.

  9. This is an interesting take on a classic argument. As someone who spent years in a corporate R&D laboratory, and then went on to work in the start up arena, I must say I really question whether standard corporate R&D is valuable to a large coporation in a truly competitive environment. A lot of the classic R&D labs held up as models arose from a need for a monopoly to create an illusion that the company itself was on the technical cutting edge. The reality was far different.

    I once spent 4 hours sweet talking an executive outside our lab about ostrich farming (the closest thing to computing he could relate to), just so we could get his signature. With another executive, after many hours of presentation on a DejaNews style usenet search and browse engine (this was 1995), I distinctly recall scratching my head while analyzing our log files trying to figure out why this executive was not clicking on anything. It turned out, after days of explaining to us how great our work was, this 7 figure compensated executive had not realized that hyperlinks were to be clicked.

    My point is, coporate R&D, often and perhaps fundamentally, goes no where. I would argue the US would be better off with this money being either taxed and pumped back into the NSF, NIH, etc., or, routed through VCs for proper technology transfer. It was depressing to see so many accomplished and gifted researchers slowly corrode as their lives became more and more comfortable in an almost utopian suburban setting.

  10. Great….. spend spend spend. It isn’t like MS has a proven history of executing on innovative products.

    If anything they usually try to do something innovative, fail, wait for it to be implemented by another company 3-4 years later, freak out, clone them and try to rip them off and end up building a mediocre product.

  11. Note that this is cumulative R&D spending for the past six years. That very much overrepresents Microsoft’s spending as it penalizes young companies like Google, which may have had little R&R in 2000 and 2001 or even later. I’m willing to bet that Google’s current per year R&D is probably close to Microsoft. Certainly closer than the graph.

  12. Look at the graph which provides the relavent data points.
    Microsoft 2003-2005: 22b -> 37b ($15b in research)
    Google 2003-2005: 0b -> 3b ($3b in research)
    Thats still 5x the amount of spending in research in the time period you consider more relevant. Certainly less than the 12x displayed on that graph but still a suprising amount none the less.

Leave a Reply

Your email address will not be published. Required fields are marked *