Tuesday, December 22, 2009

Google and Open Systems

Google’s SVP-Product Management Jonathan Rosenberg has published a manifesto explaining Google’s commitment to open systems. The post — otherwise a paean to the virtues of openness — contains a crucial caveat about the proprietary platforms that provide 99% of Google’s revenue. Rosenberg says:
While we are committed to opening the code for our developer tools, not all Google products are open source. Our goal is to keep the Internet open, which promotes choice and competition and keeps users and developers from getting locked in. In many cases, most notably our search and ads products, opening up the code would not contribute to these goals and would actually hurt users. The search and advertising markets are already highly competitive with very low switching costs, so users and advertisers already have plenty of choice and are not locked in.
So, search and related ad markets are highly competitive. Really? To echo Rosenberg’s point about the definition of open, competitive is “a Rashomon-like term: highly subjective and vitally important.” Many antitrust economists would balk at describing a market with just two players as highly competitive. That aside, Rosenberg’s analysis brings to mind the browser business in 2000, when Internet Explorer and Netscape had 80% and 20% market shares, respectively. That market was also “highly competitive with very low switching costs.” By Rosenberg’s logic, there was no need for an open source browser.
Has Rosenberg given us reliable rules for when it makes sense to open a platform? Chris Dixon doesn’t think so. He sees Google’s “closed core” strategy as a classic case of a platform owner seeking to commoditize complement providers and thereby extract more rent from its ecosystem.

I have two additional observations about the post:
  • Rosenberg’s claim that MBAs don’t learn about the merits of open systems is wrong.
  • His call for a clear and precise definition of open is on target.
Open Systems and MBAs. Rosenberg asserts that open systems are counter-intuitive for traditionally-trained MBAs, who are “taught to generate a sustainable competitive advantage by creating a closed system, making it popular then milking it … [seeking to] lock in customers to lock out competitors.” He promises that Google’s strategies for open systems will “rewrite the MBA curriculum for the next several decades.”

I hope so. I make my living teaching MBAs about platforms, and in the spirit of openness, I welcome new ideas from Google. However, I take exception to Rosenberg’s assertion that MBAs don’t learn about the merits of open systems. This topic wasn’t covered when Rosenberg got his MBA in 1985, but most business schools now have courses that examine open platforms and open innovation (for example, at Harvard Business School: Managing Networked Businesses, Strategy & Technology, Competing with Social Networks, and Managing Innovation). Scholars have learned a lot about the merits and drawbacks of open platforms. My working paper with Geoff Parker and Marshall Van Alstyne, "Opening Platforms: How, When and Why?" summarizes some of this thinking.

Defining “Open.” Rosenberg says “in our industry there is no clear definition of what open really means.” I agree and would add that in classifying a platform as open or closed, it’s vital to specify which platform layer is being referenced. Conversations about open systems often get confusing because open source and open APIs refer to fundamentally different platform layers. The following excerpt from my working paper with Parker and Van Alstyne expands upon these points.
A platform is “open” to the extent that: 1) no restrictions are placed on participation in its development, commercialization or use; or 2) any restrictions—for example, requirements to conform with technical standards or pay licensing fees—are reasonable and non-discriminatory, that is, they are applied uniformly to all potential platform participants.
Platform-mediated networks encompass several distinct roles, including: 1) demand-side platform users, commonly called “end users”; 2) supply-side platform users, who offer complements employed by demand-side users in tandem with the core platform; 3) platform providers, who serve as users’ primary point of contact with the platform; and 4) platform sponsors, who exercise property rights and are responsible for determining who may participate in a platform-mediated network and for developing its technology. For a given platform, each of these roles may be open or closed.
The Linux platform, for example, is open with respect to all four roles. Any organization or individual can use Linux (demand-side user role). Likewise, any party can offer a Linux-compatible software application (supply-side user role). Any party can bundle the Linux operating system (OS) with server or personal computer hardware (platform provider role). Finally, any party can contribute improvements to the Linux OS, subject to the rules of the open source community that maintains the OS kernel (platform sponsor role).
By contrast, Apple’s iPhone is closed with respect to all four roles. In the U.S., only AT&T Wireless subscribers can use an iPhone. To buy one, other mobile carriers’ customers must switch to AT&T, incurring inconveniences and contract termination fees (demand-side user role). Software applications for the iPhone are only available through Apple’s iTunes Store. Apple reserves the right to reject third-party applications due to quality or strategic concerns, and often does so (supply-side user role). Finally, only Apple manufactures and distributes the iPhone (platform provider role) and Apple is solely responsible for the iPhone’s technology (platform sponsor role) 
Between these extremes, we find platforms that mix open and closed roles in different patterns (see figure above). For instance, Microsoft’s Windows platform is closed at the sponsor level but open with respect to other roles. Apple’s Macintosh platform is closed at the sponsor and provider levels but open with respect to both user roles. Since all of the platforms in the figure are successful, it should be clear that without careful definitions, we cannot make general statements about the attractiveness of open versus closed platform strategies—notwithstanding enthusiasm about the profusion of open source software and content created in collaborative communities like Wikipedia’s.


  1. Great addition to the 'open' conversation!

    As a tangent, I would love to see if/how your working paper expands on the revenue models of the 'openess matrix'...what I mean is in your completely closed iPhone example, they clearly have great revenue streams...but as you point out, the other models are also successful (though to varying degress of financial success). So I would love to see some fincial data overlying some of the popular/successful models you mention.

  2. Kevin: We don't explore business models/financial performance in the working paper. There's been a fair amount of theoretical modeling work on the economics of the open/closed choice (including some great analysis by my coauthors Parker & Van Alstyne). There's been much less empirical work that actually measures financial results. However,it's clear that closing a layer creates more business model options. iPhone is a great example. It profits from device sales, from bounty paid by AT&T, and from toll taken on paid apps. Contrast that to a Linux OS vendor who profits solely from service contract.

  3. Great post! I was wondering, how would you classify the openness of Google's search/advertising business using these four roles? And who is Google's demand-side user?

    The openness of roles seems to depend a bit on who one views to be the demand-side user. When I view Google as a search business with searchers as the demand-side user, then I think the column would look like [O,O,C,C]?

    When I view Google as an advertising business with their search engine as just one means to generate ad inventory, I think the column would look like [O,O,O,C]?
    Demand-side user - advertisers (anyone can advertise)
    Supply-side user - web sites participating in adsense (anyone can put up an adsense bar)
    Platform providers - 3rd party ad management software ISVs (I believe anyone can use Google’s AdWords API to become a demand-side user’s primary point of contact)
    Platform sponsors - Google

  4. Thanks, Chris. Both of your interpretations make sense, but I favor #2. The ad management ISVs do indeed play the platform provider role when they serve as the primary point of contact for an advertiser. A parallel would be the Windows platform. With minimal restrictions, almost anyone can license Windows then build and sell a PC. MSFT is the platform sponsor (closed) but the platform provider role -- a consumer's primary point of contact with the operating systems -- is the PC vendor (open). A practical test: the platform provider is the entity you call first when something doesn't work. Probably Dell if your PC won't boot (but of course this is always confusing in the Windows world). Certainly the ad management ISV if you are unhappy with ad results, billing etc.

  5. I suspect that Google -- and other large search providers -- could turn their search technology into profitable open infrastructure that would allow other players to develop better search algorithms.

    The trick is to become a cloud computing provider that sells not just compute cycles, but compute cycles with local map-reduce access to a full web index.

    Essentially, this turns Google into a utility. The probably don't want that. But I bet that everyone else does, and I suspect that this is actually a big opportunity for disruption.

    For more, see http://jonathanstray.com/why-we-need-open-search

  6. Thanks, Jonathan: your post about the merits of opening GOOG's search platform is interesting and provocative. The post from Chris Dixon that I cite above also explores this question, and in case you missed it, Charles Ferguson wrote a pair of articles on this idea in Technology Review back in 2005 http://bit.ly/5gDkVT and http://bit.ly/4KqEPG