The “Web 2.0” moniker is pretty flexible in its connotations, but one thing it does seem to represent is an emphasis on collaboration and consumer participation in a site or service. If that’s true, then Microsoft has gone all Web 2.0 with the new set of tools it’s developing for Web advertisers at the adCenter Labs site (http://adlab.msn.com).
AdCenter Labs is the R&D arm of Microsoft’s online advertising platform, where 120 or so researchers and engineers work on tools marketers will be able to deploy to get more mileage out of not only their pay-per-click search campaigns but from advertising on emerging media (think mobile, video and Xbox Live) and on Microsoft’s soon to be launched contextual network.
“AdCenter is our platform for Microsoft Digital Advertising Solutions, and as we progress, adCenter will control more and more of those solutions,” says James Colborn, product manager for search strategy at Microsoft. The adCenter Labs division grew up about 18 months ago to incubate tools and platforms for these ad products.
And the adCenter Labs Web site grew up as a place to let potential users get their hands on those tools while they’re still in beta form. “Tools can be showcased so that folks can go and work with them, and so we can gather feedback as to whether these tools are market-ready or how they’re handling in relation to what the market is actually looking for,” Colborn says.
The Web site launched last June as a kind of look-but-don’t touch demo showcase for tools in development. The marketer community appreciated the inside look, Colborn says, but many visitors also told Microsoft they wanted more: They wanted to be able to see how the tools actually worked in analyzing and recommending keywords. So adCenter added data from 5 million MSN keywords to the test bed, allowing marketers not only to evaluate the tool products but to get some useful insight into their search campaigns in general.
And the data integrated into the tools available at the adCenter labs site is not only accumulated from adCenter customers but also incorporates predictive algorithms, so that visitors can in many cases get a look into what Microsoft engineers forecast for keywords over the next month to six weeks, based on past performance.
“Now when people come in and use these tools, not only can they get an idea of where Microsoft is going, but they can use them and put them into their everyday operations,” Colborn says. “With adCenter, it’s what we know. With adCenter Labs, it’s where things could be, based upon analysis of data trends.”
The adCenter tools and that community input received top billing at last month’s third annual Demo Fest, where Microsoft invites its biggest online advertisers to Redmond WA to preview future developments. There Tarek Najm, distinguished engineer and general manager of Microsoft adCenter, credited user suggestions with speeding up the development of many of the tool items shown.
“At last year’s Demo Fest we said that this technology would take us three to five years to get to market,” he said. “But due to overwhelming community feedback we’ve been able to release it in just one year.”
Visitors to the adCenter Labs Web site will find demos sorted into four primary areas: paid search, contextual advertising, intelligence (audience and demographic understanding) and emerging media. Right now paid search gets the most traffic because that’s how most users engage with Microsoft advertising; but the other three categories will grow in importance as the company rolls out new ad products in those channels, Colborn says. (In fact, Microsoft already has selected advertisers entered into a pilot of a new contextual ad program on the fledgling MSN Content Network.)
Among the paid search tools on display, Colborn says a new tool for analyzing the search funnel has gotten a lot of attention. Advertisers can enter a term and get a look at where users who searched on that term in MSN Search searched next. In the case of a brand name such as BMW, the search funnel tool reveals that about 5% of users went to a subsequent search on “mercedes” or “lexus”, while slightly smaller proportions went to “audi”, “honda” or “mercedes benz”. The tool can also be reversed to show the incoming funnel of searches done before a “bmw” search.
Colborn says the funnel tool can provide marketers with search insights that help tailor their creative to fit their specific search environment. For example, the outgoing funnel for the term general “cameras” turns up two specific retailer brands in the top five results, eBay and Best Buy. Entering “cameras” turns up three, with the addition of Wal-Mart, and a specific brand, “Canon”. That can be useful knowledge if you’re an independent retailer looking to compete with those big marketers.
Another tool under the paid search heading is keyword forecast, which provides trend data on searches for multiple terms over the last 12 months and algorithmic forecast data for the coming quarter. Colborn types in “Xbox;wii;PS3” to show the amount of search activity on those keywords and finds that while Xbox searches spiked in March 2006, with the introduction of the 360 console, searches for Nintendo’s Wii and the PlayStation 3 hit peaks in November. PS3 searches on MSN Search then declined through the holidays and will continue at a low level this quarter; but Wii searches have continued to rise and should do so through March.
The keyword forecast tool also includes distribution breakouts by age and gender. But the data contained in these displays can be dubious right now, Colborn admits, reflecting their prototype status. For example, the tool tells users that almost 100% of those searching on the term “Honda” were male—news that the car maker would probably be able to dispute pretty easily.
“AdCenter Labs is an iterative process, and we learn more as we go forward,” Colborn says. “When something like that [the anomalous Honda gender reading] comes up, it’s nice to see it in the Labs so we can work on it before the tool goes into production. This is a very popular tool, and users are helping us identify areas we need to work on.” By the time this tool makes it into the adCenter toolkit, he says, that demographic predictor will have been refined and validated against the market.
Another adCenter tool currently under development, and available for testing, examines the likelihood that someone searching on a particular term or URL is not simply doing research on a topic but has a “commercial intent”. For example, someone searching on “patriots” is probably searching without any intent to buy New England Patriots merchandise; the adCenter online commercial intent tool gives that search less than a 2% chance of resulting in a purchase. But a search for “patriots jersey” has a 97% chance of having a buying goal.
Those examples are obvious for illustration purposes, but the tool can prove useful in less clear-cut cases. “Imagine you’ve got a big keyword list and you’re not quite sure of the variables that go into it,” Colborn says. “Understanding the different intents for the keywords users type in may affect your approach to bidding on those words or to writing the creative around those terms. It may even help you decide whether you want to bid on those keyword terms at all. Advertisers with small budgets might want to focus on keywords with high commercial intent, while others might want to bid on both research and buying terms to get a broader awareness across the spectrum of searchers.”
Here too, user feedback is helping to set the development path for the tools Microsoft will eventually bring to market. Colborn says users of the online commercial intent tool have said that while they find it useful, they’d like to see how that intent varies seasonally. For example, it’s pretty reliably certain that someone searching on “ski vacation” in November or December is ready to book a trip. But how likely is someone doing the same search in May or June to buy, compared to the number of research-only queries?
“This is one instance where the [adCenter labs] Web site has allowed us to put the tool out for user comment and garner some valuable feedback in return,” Colborn says. “That’s why adCenter Labs and this Web site are in existence. My job is to heighten awareness of the labs and the site and get more people using these demo tools to increase that flow of user input.”