Sunday, November 06, 2005
Missing the forest for trees and Google vs Evil
I don't know if you have been following the dialogue between Umair and others (including yours truly) on the Google Print thread, but it contains a pretty cool insight. People smarter than me may have got it rightaway, but here's my notes on what I learnt.
Briefly, here's the discussion: Umair asserted that Google is creating network FX by indexing infromation across different domains (websites, books, videos...etc). A reader and I didn't agree with it. I compared GOOG with a phone network, a classic example in network econ, and said that a phone network creates network FX for its users by connecting their consumption and thus increasing the marginal utility for each new member. Put it simply, the value of network increases for me if my friend joins it too because I can then use the network to talk to them. In case of GOOG, this is not readily apparent: if I can find a book on Google Print, it does nothing for my friend who still needs to use Google search to find the book she wants.
Umair's insight is that the network FX in this case are indirect, the real consumsers here being the advertisers. Essentially, Google attracts common users like you and me. As the number of eyeballs increases, the value of the network to the advertisers increases. This, of course, makes the network more valuable for an advertiser's competitors because they are trying to attract the same eyeballs. Thus, there are three actors here: Google, users, and advertisers. Therefore, the network FX are indirect because the utility of the network increases to advertisers through the increase in the number of common users. The network's value to common users (searching) is very different from that to the advertisers (buying attention).
It is very easy to see that the network FX created here lead to a natural monopoly dynamics in the advertisement market. However, there is one caveat: notice how Google depends upon capturing the eyeballs to create network FX. What if those eyeballs were lost? Clearly, this will destroy the network FX and make GOOG's business model useless. After all, we know that Google is still not a search monopoly: web search is pretty well an oligopoly among GOOG, YHOO and MSN. If I were a strategist for Google, what would I do to prevent the defection of the eyeballs I currently own? Where would I be speanding my dollars to attract more and more common users to my network? Thinking about this leads us to understanding GOOG's strategy.
The key to defending a strategic position is erecting a barrier to entry. Umair mentions in the thread that GOOG is indexing ever larger data cross domains which is more and more capital intensive, thus creating a finanical barrier to entry. However, this is not the whole stroy because AMZN, EBAY, MSFT and YHOO also have substantial economic resources as well as large user bases. For example, MSN+YHOO+AOL IM networks can create a huge user network. Search, by itself, is not a "sticky" service, people will quickly switch to another website if it can search stuff better. Let's not forget that GOOG itself was a fairly late entrant into the search market but it successfully competed against YHOO, Altavista and Infoseek to become a market leader.
So as he further elaborates , there is more to it than Google's searching prowess as well as the resources it takes to develop it. GOOG's indexing plus search algo's work on creating the barrier to entry on supply side. How'bout demand side? The answer to that is, as always, increasing the buyer's switching costs to reduce their power. So how does Google take its competence - indexing + search algos - and use it to increase the user's switching costs? By applying it horizontally across different vertical domains and creating linkages among them that create unique value for its users. My hypothesis is that it is this integration across domains that will be the key to Google's competitive advantage. This explains all of their recent efforts ranging from deep integration of GTalk with GMail to leverage common user ID to Google Print and Google Desktop Search.
However, the beauty of GOOG's strategy, as I understand, is that while it locks in consumers by creating these switching costs, it is the advertisers over whom GOOG exercises market power, not the consumers themselves! This is really quite cool : the traditional aim of a firm's corporate strategy is to increase its market power over buyers so that the firm can extract more profit from them through pricing of its goods and services. GOOG, however, does not directly charge anything to its consumers, rather, GOOG exercises its market power to extract something even more valuable in this economy : attention. GOOG then monetizes this attention through the money it charges to its advertisers.
Think about this scenario a few years from now: I am roaming in downtown Vancouver one saturday afternoon and using my mobile to find the exact location of a shop carrying a rare copy of an old Sanskrit text that I had found by searching on an Indian university's library archived using Google Print. Who's the ONE company that can find that shop for me? Of course, GOOG. Why? Because it has integrated vertical domains: web, languages, mobile, mapping imagery, geographic locations, businesses and print archives. How? It has applied its competence in searching across domains. Why? To increase the switching cost for a user like me and buy my attention for zero marginal cost. Why is that important? To sell my attention to GOOG's advertisers for profit and protect such advertising revenues.
If this really happens, GOOG monopoly will be more powerful than any other in known history, including our resident evil empire de jour from Redmond. The sheer scope of GOOG's ambition is breathtaking. Think of how valuable the access to Google' s network will be for the aforementioned shop which not only has its location, but also its entire inventory digitized and stored in Google's databases. Then think of how I'd ever be able to wean myself away from that network. Talk about being qwned!! Oh yes: it WILL be evil. It just has to.
Yeah, I think it's safe to say you got exactly what I was talking about!!
Tat's the power of the natural monopoly dynamics in this industry - your example is spot on.
BTW, this is a very important insight worth thinking about wrt natural monopoly dynamics:
(You said:) "...buy my attention for zero marginal cost."
That's the key to this whole thing - a leverage point, if you like. To Google, the marginal cost of your attention is low/decreasing, but to losing players, it's high/increasing.
Perhaps a nicer, more elegant way to put this whole discussion :)
Yup, got it. However, what I am now wondering is how exactly can we quantify such values? For example, what exactly is GOOG's marginal cost of acquiring my attention vs. that of YHOO, EBAY or AMZN? To make more sense of Google Print: exactly how much marginal value is created by addition of a content publisher? How do we compare that with the Adsense rev share model today? For example, how much marginal value does your blog create for Google and how much of that value is captured through AdSense? it'd be interesting to build these quant models. I know you have some of these in your ppt's, but it'd be fun to extend them some more and make comparisons among different companies.
// Mahashunyam // 2:13 AM
"Therefore, the network FX are indirect because the utility of the network increases to advertisers through the increase in the number of common users."
This strikes me as very wrong. There's no network FX for advertisers here.
Sure, the value to advertisers scales as the audience scales. But this differs from oldskool broadcast media how, exactly?
Advertisers don't benefit in the slightest from *other* advertisers using the ad-platform. Contrariwise, it just drives up the cost-per-eyeball - exactly the way it does in traditional media.
Yes, the connections between different Google "components" (as Scoble calls them) are valuable. But this has nothing to do with network effects.
// Composing // 4:46 PM
As Rajan has pointed out, 1.0 businesses, like broadcast nets and newspapers *do* realize network FX.
These network FX are not between advertisers, they are between the pool of consumers and advertisers, mediated indirectly by whatever price mechanisms the firm is using.
Why don't they create a huge amount of value?
Well, they did. But their growth was artificially limited by regulation.
They also didn't microchunk like Google does = explodes the size of the network.
How do we know this? What you call "driving up the cost per eyeball" is exactly an indirect network effect.
Think about it this way: why does the cost per eyeball increase in network size? If there was no network effect, it should stay flat, even as audience size grows.
But, as you've pointed out, it doesn't. That is, my private value remains the same, but my social value increases when you join the network - advertisers are willing to pay a higher price for me.
To add to Umair's point, we can also see this from the persepctive of how eyeball shopping is done differently done differently in traditional media vs. Google. On TV or in print, you buy fixed chunks such as column inches or 30-second spots. The price is essentially for the real estate categorized in different slots. A 30-second ad on ESPN broadcasting Superbowl costs a lot more than a similar spot on C-SPAN. Column inches on front page of NYT cost a lot more than the 18th page of the same newspaper. However, notice that there is very coarse correlation between the real estate you buy and the attention it grabs. Your ad on Superbowl can be viewed by anywhere from 10 to 100 Million people : who knows? So, to you as an advertiser, the marginal utility of acquiring another viewer has very coarse correlation with the number of viewers that the network has attracted. Contrast that with Google where every clickthough is tracked and the higher number of consumers on the network directly translates into more click-thourgh's, which are a proxy for the amount of attention you were able to buy from the network. That is why Google has a variable pricing model : you are not buying a fixed real estate on Google page, you are buying specific keywords which show up on search results page only if somebody is searching for them. Once an advertiser has grabbed my attention with the key words, the value of my attention increases even more to its competitor because now they know that I am interested in their stuff.
This is also why indirect network FX are extremely powerful for Google: Google can reuse and re-sell the webpage real estate on a user's computer differently for literally every consumer who's accessing the network while still getting a piece of monetized attention! Remeber that Google click-through rates have remained constant as a proportion of the size of the network. Compare this with the broadcast network : the best it can do is show 200 different ads on the 200 channels broadcasting the same 200 programs in a region - essentially selling a fixed 30-second time slice. They cannot resell the same timeslot on differnet users' television sets to different advertisers. This then leads you to a great illustration of the difference between how their revenues scale vs. those of Google. Broadcast media hit limit on their revenues because time is finite - there're only 24 hours in a day and only 200 channels on my TV. Going from 200 to 500 channels is actually going to get even less attention per channel from me! Google's revenues, on the other hand, can scale virtually ad infinitum. This is because Google realizes network FX, but 1.0 broadcasters realize economies of scale.
// Mahashunyam // 9:33 PM
Hmmm. Maybe I misread the original post. But now I'm more confused than ever.
As I understand, we're talking about Google Print. And the question is whether it has network FX.
What I thought Mahashunyam said was that he originally thought not. But then that you convinced him it has : in that there was something like a "secondary" effect. Advertisers found Google more valuable the more viewers Google has.
I don't disagree that Google is building something close to a monopoly, but I'm still worried by the term "network effect" used in this situation. I think it might bring more confusion than understanding, especially as most people think of the "telephone example" for their intuitive idea of a network effect.
Let me present you with an analogy. Suppose, by careful dealmaking, and smart exploration, I managed to get hold of most of the oil-wells in the world.
I could start to increase the price of each barrel of oil. As my collection of oil-wells got larger, the remaining supply outside my monopoly would decrease, and its price would rise too.
Yet I don't see that "network effects" would be the best way to explain this price rise, nor that there's reason to think that were new supplies to become available, my customers couldn't shift to them.
Now replace me by Google Print, oil by attention and oil-customers by advertisers, and it seems we have the same situation. Google has managed to get hold of most of the attention in the world, but it didn't get that through network effects.
If Google and Microsoft's booksearch offer identical service, then neither the number of current users nor number of advertisers affects my decision as to which one I should use. But if the network effects are not keeping me loyal to Google, then nothing is keeping the advertisers loyal to them. The advertisers will follow the readers.
There *are* barriers to rivals : there are synergies between Google components, there's Google's accumulated expertise. But these are *not* network effects.
Also : I buy everything you say about Microchunking. It does release more value (did you read about the guy who sells magazines by the page? http://sethgodin.typepad.com/seths_blog/2005/10/understanding_t.html)
And there's an opportunity for more Google lock-in there, if Google can, for example, persuade people to use it's addressing scheme for the microchunks (ie. blogs link to particular book-pages via a Google URL) then it might be able to derive some network fx from people using the URLs to communicate. (Much as Amazon already has for addressing individual books.)
// Composing // 10:20 PM
Post a Comment