Thursday, April 28, 2011

Finding Yahoo's Way


Last year I had the pleasure of presenting a seminar at Yahoo Research. One of the things that interested me about that part of Yahoo was the fact that it had, in my field of economics, made some very significant hires from academia. In a time where some academically oriented labs had fallen aside, Yahoo had moved in the other direction. And this wasn't some independent and loosely tied research organization. The economists I knew were heavily embedded and yet still publishing freely.
My presentation there was actually about just that. While many tend to think of commercial pressures as anathema to scientific openness, Fiona Murray, Scott Stern and I developed a hypothesis that suggested the opposite. When commercial prospects are strong — in that firms can commercialize innovations securely with less threat of entry — they will also be more receptive to allowing their scientists publishing rights. Put simply, when entry prospects are low, you don't have to worry as much about academic publications helping entrants. And why should firms take any risk on publication? Because scientists get personal (and possibly career) value from it and so you can pay them less. The implication that everyone in the audience was sacrificing income for working at Yahoo Research and publishing was lost on no one.
But Yahoo Research was more than just a potential source of cheap scientific and engineering labor. At least for the economic part of the group, there was a clear focus — for the most part what they now focused on was advertising. The focus was not exclusive — they had freedom to work on what they wanted — but it was there. Preston McAfee, who leads the group, gave a keynote lecture at the International Industrial Organization Society in April after receiving its prestigiousDistinguished Fellow award. The talk focused on advertising exchanges and the direct impact the theory of auction design (of which McAfee was a prime contributor) on the actual design of advertising exchanges at Yahoo.
This focus was natural. Advertising was responsible for most of Yahoo's revenues and this is an area the company leads in. Its ad exchange displays billions of ads across a large array of websites. Its sales of display ads exceed those of Microsoft, AOL, and Google/Doubleclick. And each time an ad is displayed, in the quarter of a second between the time someone requests a page and the ad loads, an auction is run for that ad space. In fact, as McAfee pointed out, it is the largest auction platform (in terms of number of units sold) ever. And that means that there is some serious muscle at its heart.
When you click on a Yahoo-served page, the content loads first and the ad loads second. I'm sure you have experienced the opposite, when you go to a news site but it fails to load because the ad it is trying to display has held things up. Yahoo uses that time to their advantage to make sure the right ad is selected and loads.
The auction itself involves some serious economics alongside the computer science. There isn't enough time to really select the best ad for a particular page call. But Yahoo's economists have reduced the complexity of the process and matched it with a good deal of economic theory to ensure that the systems at Yahoo learn what to do ahead of time. It is no wonder that Yahoo CEO Carol Bartz wears t-shirts bearing economic formulae.
What is interesting here is that advertising exists as one side of a two-sided platform. The other side is content. Yahoo often describes itself these days as a media company. It has acquired and developed many content providers from all over the digital world. While that might aggregate consumer attention for the purpose of displaying ads, it does not necessarily resolve the difficult task of matching ads to content. When content selects for consumers of a certain demographic of interest to advertisers, at the same time it thins the market for advertisers. And if Yahoo's technical and economic competence is in managing thick advertising markets, does that mean its content strategy is acting in opposition to that competence?
This raises a number of strategic questions for Yahoo. First, does Yahoo's suite of products remain just that — a suite of distinct products — or can it coalesce into a platform? In a recent book,Michael Cusumano suggests that platform arises when external actors direct effort and investment towards enhancing the value of a company's products. But who is doing that for Yahoo? Where are its complementors? Contrast it with Wikipedia — arguably the largest-focused media entity on the Internet — which has an army of complementors voluntarily contributing content.
Second, if media is not at the core of Yahoo, is it a distraction? Should Yahoo be providing content at all or should it be more like Google and focus on advertising and tools to manage content? The pros of disintegration may be some corporate focus but the cons may be difficulties in designing ad markets that efficiently work when content is designed to self-select for the right kinds of consumers.
Finally, my earlier argument that commercial interests and academic motives for openness can be compatible rather than in conflict rests on an assumption that those commercial interests are somewhat protected. To be sure, Yahoo is a holder of important patents on advertising auctions. But it is equally the case that there are many economic ways to generate efficiently operating ad markets. Moreover, the underlying theory that has formed Yahoo's own operations is commonly known and published. Hence, competitors may be on their way and any distinctiveness Yahoo has could be limited.
The researchers at Yahoo may already be on to this issue and may be finding a way to bridge the advertising/content divide in the market. Otherwise, one wonders how long the model of an academically-founded lab embedded within a commercial organization will persist. Instead, it may slowly evolve to a more autonomous arrangement with a longer-term focus rather than immediate practical applications with its scientists having to choose between a commercial or academic way.
Joshua Gans is an economics professor at Melbourne Business School and a visiting researcher at Microsoft Research. All views expressed here are his own.

No comments: