• by Phil Leggiere , September 5, 2007

The Web, as David Weinberger has written, is about reminding us we are "connected creatures in a connected world." For targeted advertising to function constructively in that space, it must itself become a vital connector between people and their shared passions and relationships, as Sharon Peyer, director of business development at media sharing site Pixamo, explains below.

Behavioral Insider: Media sharing sites are becoming very popular but monetizing them has remained  a huge challenge. How is Pixamo looking at the potential for targeting media sharing behavior?

Sharon Peyer: For us, there are two significant ways to encourage tagging of content. One is to give people a lot of different ways to tag, while providing tools that speed up the labeling process and promote consistent tagging of similar content across different users' collections. For example, we offer four tag categories that users can apply in describing their content: people, places, events/things, and dates. The dates are pulled in automatically based on a user's camera settings.

Also, our tags, once created, are saved for future use and sharing.  So, when someone begins entering the first few letters of a tag, like ‘Amy Smith,' for example, Pixamo will present a list of potential matches. The list consists of tags in the Pixamo database that are presented in the following order to promote relevance and consistency: first, tags that the user in question has previously used. Then, if there is no match, Pixamo reaches out to the users' friends, then to his friends' friends, and finally to any other Pixamo user/the Pixamo database. Once a match is found, the user adopts the suggestion without having to manually enter text.

BI: How do you do that?

Peyer: Through tools that demonstrate the benefits of tagging extensively. We do this by presenting browsing users with content that is relevant to them, such as photos others have of them, or videos with other content that we think might interest them.  We can illustrate this using the Amy Smith example from above. After uploading his photos of her, Pixamo would generate an invitation for John to send Amy. When Amy get John's invitation, she reads: ‘Hi Amy! Check out my photos and videos of you on Pixamo!' Upon activating the invitation, she sees John's photos of her, and while she's browsing his photos, Pixamo tells her that Alfred also has photos of her that she can view. As she's browsing Alfred's photos, Pixamo tells her about other content (this might be content that is frequently associated with the tag Amy Smith' or related tags in John's, Alfred's, and their other common friends' collections), providing her an opportunity to explore that content as well. What we see is that people like Amy will like the tag-enabled, relevant browsing and sharing so much, that she'll start contributing her own richly tagged content to share with Alfred, John, and others.

BI: How do users react to this?

Peyer: It's completely up to the person if they want to pursue any of the suggestions or recommendations. But what we've learned is that a very high percentage do. It's because there's a benefit in personalized content delivery.

We encourage people to upload their own photos and videos, using as many descriptive tags as possible. 

BI: How does your technology platform track tagging?

Peyer: Behaviorally our technology is designed to derive associations among tags used, tags viewed or searched for, and wider interests. Say someone looks at a number of pictures of islands, oceans, beaches and the Caribbean. One thing we can derive from that is that they may be interested in travel and vacation spots. This kind of association can also extend to, and be combined with, other types of tags. For example, if someone in a group affiliated with a graduate school, say, has lots of friends and we find a wide range of them sharing photos with a tag for an event at a bar or club near the campus, we can use that data to target ads for that club or similar locales. We can also do things like combine information about people appearing in photos or videos with the browsing behavior those same people display on our site, for very personalized targeting.

BI: What kinds of metrics are you paying most attention to?

Peyer: For us the most important metric is user engagement. We can show that the more tags a given content has, the more page visits are generated, and the more time is spent on the site. One application with enormous potential is, we can cluster together tags that seem to be characteristic of a given group or community of interest, be it people within the same school or workplace. Say we discern that there's a pattern where people in a school group have frequent tags related to athletic shoes and also tags related to the color red. That behavior would be of great interest to a Nike.

BI: How is behavior and behavioral targeting diverging in media sharing from what's going on with, say, social networking?

Peyer: We've noticed that engagement and page view metrics increase dramatically if you go after and emphasize real-world connections. What I mean is, social networks will cluster together by subject interest primarily, and that's useful, of course. But we find that the really vital nexus of user behavior (especially as it relates to rich content generation) is much more intimate than that. It's the tight-knit group rather than the broader network. The group implies a much stronger bond and, therefore, more valued interaction. It could be a shared school, an alumni group, a corporation, a military base, a sports team, a town or neighborhood. With those real-world bonds as a base, sharing photos and videos becomes a way of cementing ties to an offline community.