• by Susan Kuchinskas , July 22, 2008

They will find you: A new generation of services reveal so much about consumers, it's scary

Sitting around having beers with the people on your softball team, you find out some crazy stuff. Paul is an ex-Jesuit who makes a wicked gazpacho. Despite her wild shoe collection, Jennifer is a politics wonk and a green consumer. And big, burly Raj, when he's not working on his car, spends hours rating videos on blip.tv. Oh, and he has a thing for fine Italian shoes.

But ad networks can be blind to such nuances, and targeting software pushes consumers into buckets. You can identify Raj as a shoe shopper or as an auto nut, but you may have no idea what he's doing on social media sites.

A new generation of services promises to connect the dots, not only building a complete picture of an individual's interests and behavior, but also making highly informed predictions about what he's likely to want next. These services combine behavioral targeting with prediction algorithms that let you define your target and then find him and people like him. The secret ingredient is the ability to marry a wider variety of behaviors to in-the-moment click-stream analysis, and then to make a nearly instantaneous comparison of this anonymous user to every other user in a vast pool of data. While behavioral targeting looks at what you're doing or have done, this is predictive behavioral targeting: If you look a lot like someone who's bought a car, you're probably going to buy a car, too.

The result is almost a searchable index of the entire online audience. Instead of using audience analysis services to take your best guess as to where your target market goes, you can use these indices to select people who fit your profile and target them whenever they show up - even on niche Web properties that don't show up in comScore or Nielsen NetRatings.

When a behemoth like Google gets into the game, you know it's serious. In June, the dominant search advertising provider announced Google Ad Planner, a free tool that lets a media planner enter the demographics and sites associated with a target audience. The tool delivers a list of sites both on and off the Google content network with similar audiences - in effect, Google audience search.

"These new entrants argue very effectively that you can use them to go after the long-tail sites," says Andy Fisher, director of analytics for Avenue A | Razorfish. "Panel collection methodology is extremely useful, and we use them all the time to do planning. However, for the panel collection methodology to work, a site has to have a certain volume."


Quantifiable

Newcomer Quantcast claims it can uncover those long-tail sites that will work as well for an advertiser as for more mainstream - and expensive - properties. Quantcast is an open Internet ratings service that gives advertisers audience profiles for more than 20 million Web sites and services that have become "Quantified" by adding tracking codes.

Avenue A's Fisher says, "Quantcast vastly expands the number of sites you can consider for your media buy. One of the ways we use Quantcast is ... to look at things that would never show up on comScore or Nielsen."
There are two sides to the service. On the publisher side, those that enable Quantcast on their sites contribute to a blind data pool of user behavior that powers the audience analytics. Quantifying their sites can give publishers valuable information about their audiences that they can use to increase the value of their inventory.

Gawker Media uses Quantcast data to package custom audiences and offer them to advertisers. In the past, Gawker placed advertisers that wanted to reach a female audience on sites like Jezebel and Gawker, while letting them target males on Jalopnik and Gizmodo. Quantcast lets the network of blogs identify and segment a unique audience across multiple titles. Today, the publisher offers two audience packages, Gawker Chic and Gawker Geek.
"The custom breakout in Quantcast for those two clusters gives us unduplicated reach and frequency in those subsets," says Chris Batty, vice president of sales for Gawker Media. "It eliminates the overlap on the readership."

The flip side of the coin is an audience search engine called Quantcast Media Planner. It lets media agencies find hidden gems among the shifting sands of niche sites. Instead of - or in addition to - buying the sites they know their audience visits, agencies can search for more sites like them. The Web-based interface looks and works a lot like traditional search: Media buyers can define the audience they seek via a query with criteria that include audience demographics, category or similarity to another site, called affinity. Publishers who don't want to let agencies search for their audiences can control how their data is aggregated and how it shows up in the audience search tool.

An auto manufacturer could index its audience of everyone who had gone through the configuration process within the last three months, then find all the other places on the Web where those people and people like them go.
While retargeting your own site's visitors is a familiar tactic, Adam Gerber, CMO of Quantcast, says, "The bigger opportunity is to help marketers understand who people are who tend to come to the site or have taken action, and find more like them."

Mediasmith was a beta tester of Media Planner, and founder David Smith says it makes it easy to evaluate all the sites in a network that wouldn't normally show up in comScore or NetRatings. In an early test, Smith says, the agency gave Quantcast a list of eight long-tail sites it had identified as likely to do well for a particular client. "They came back with a list of literally hundreds of sites that were appropriate - that we didn't know existed. It helps us assure the advertiser we have our arms wrapped around the audience we're buying, and it helps the planner and the buyer at the agency to feel more confident in a site."

Smith's comment illuminates a growing concern over the ability of traditional Internet audience measurement services to keep up with the expanding universe of consumer-generated and niche content. Says James Kiernan, the MediaVest vice president and group client director who runs the agency's P&G business, "Some of the tools we've been using over the last decade still have a difficult time in getting down to these niche sites. That's partly because some of these don't have the clout to work with the comScores and Nielsens or to be registered on their platforms."

MediaVest often works with ad networks to follow people into those niche environments, but, Kiernan says, "we need to dive deeper." Using Quantcast Media Planner for the baby category, for example, Kiernan was able to quickly identify lesser-known sites that indexed well against babycenter.com, one of the market leaders. In just a few seconds, he says, "I can drill down into justmommies.com and see how potentially this site could offer me a competitive advantage."


Look-alikes

Quantcast aims to be the matchmaker between advertisers and publishers; aCerno uses similar modeling strategies as the basis for a buyer-side ad network designed to increase ROI for advertisers.

The aCerno network has its own blind data pool, derived from anonymous shopping and purchase behavior across 450 participating multichannel retailer and product manufacturer sites. It watches what more than 140 million people - or rather, their unique cookies - look at, what they check out, what they abandon in their shopping carts.

The company's technology makes inferences from recent behavior to understand who customers are and to predict what customers will buy next. Say one bought a dress and shopped for shoes, while another shopped for luggage and researched airfares. "People doing these things are likely to either need or do certain other things," says Bennett Zucker, vice president of marketing. "Every time we see the cookies of people who we know are in the market for what an advertiser is selling, we'll serve its ad to that person, whether they're checking their e-mail on Yahoo or reading the news." It will also show those ads to people who look like those people.

Zucker says that because aCerno models so many variables, it can figure out, for example, whether someone who's been to a retailer's site five times but never bought is just about to convert or is never ever going to drop a dime. It can also quickly figure out that this guy may have bought a dress for someone else, but it's not going to be a regular thing for him.

"You can get CPA advertising targeted to people all over the Web who are likely to do what you want them to do," Zucker says.

This company, too, says it can uncover hidden audiences. For example, 3M wanted to target female paper-pushers with fashion sense for a new line of Post-it notes. Instead of display ads on content that correlated with those interests, aCerno used its shopping data to deliver the ads to women who bought fashion apparel, worked in home offices and had purchased designer materials and objects.

Predictive behavioral targeting can turn up some counterintuitive things. For example, Mindset Media, an ad network that lets brand advertisers find people who fit their psychographics, discovered a correlation between buying sneakers and leadership traits. They've also found that people who buy organic are highly creative.

"Marketers have known for a long time that psychographics do correlate with buying behavior," says Mindset cofounder and coo Sarah Welch. "The innovation we offer, in addition to creating profiles, is giving you a way to reach them online."

Mindset conducts surveys, in partnership with Nielsen Online, to type a target audience on 20 personality elements, each refined by five levels of intensity. The resulting MindsetProfile lets brand advertisers identify the psychographics that drive its brand or its competitors. Once the MindsetProfile is defined, the advertiser can produce creative that appeals to those characteristics, and then reach millions of people who fit the profile across the Mindset Media ad network.

"Behavioral targeting is fantastic at harvesting smaller groups of interested prospects at or near the point of purchase," Welch says. "A Mindset Media buy would be used much earlier on in the media plan to fill the upper funnel with flocks of people who have the psychographic makeup to end up at the bottom of the purchase funnel."


Friendly Wager

But what about those small groups who gather online around specific interests? In social media, especially user-generated or user-posted content, demographics and context shift under the marketer's feet. Lotame lets clients build a customized audience among 52 million users across 20 social media sites and then target relevant advertising against it. It identifies the most influential members by their actions: how much they post, comment, rate, etc. For example, a media planner could ask for all the people who uploaded, rated and shared golf and finance content in the previous 72 hours.

The analytics also help agencies discover new audiences. When Lotame, which specializes in social media, ran a campaign for a major sports league, it found that influencers in the arts and leisure audience performed eight to 10 times better than those in the vertical sports audience.

Lotame's Crowd Control does the same thing for publishers, letting them package their audiences into custom segments - and sometimes uncover audience segments in the process. Take social network Meez, which lets you build animated 3D avatars and use them in profiles, games and videos. Michael Lehman, vice president of marketing, says Lotame allows Meez to collect and analyze information about what people do on every page. Meez uses that information to decide what assets and services to create, as well as to segment the audience. The analysis showed that among the preponderance of teen and college users was a small but passionate group of mothers. "On the surface of things, you wouldn't come to Meez for moms," Lehman says. "Now we can validate that for an advertiser." In fact, for such packages, Meez charges more than double the CPM it would get using a generic remnant ad network.

Taylor Valentine, associate director of strategic services for Horizon Interactive, has used Lotame in six campaigns so far. He appreciates that the system is based on true behaviors and interests. "Social media has been so hard to capture," Valentine says. "You won't see the same kinds of response rates within social media, but you can identify the people who are genuinely interested in what you are talking about."

"Think about the reams of data these social media sites are sitting on," says Kiernan of MediaVest. "Not just what their favorite shows were three months ago when they took a survey, but right now, because they updated their profile in the last hour. Using that data to identify new trends ahead of the market is a huge advantage for agencies. Lotame is one of the first to roll that out and allow advertisers to target and advertise against those affinities."

It's an exciting new frontier of targeting, but it still needs to be held accountable to measurement and ROI standards, Kiernan says. "A lot of panels give us the opportunity to reach the consumer, but we're going to need to get down to the expectation that we definitely reached that consumer."