Stipple Builds Out System to Help Publishers Profit from Tagged Web Images

Between 30 and 40 percent of all files on the Web are images, and even pages dominated by text usually have a few photos. Yet most of these photos aren’t linked to anything interesting, the way text can be. San Francisco startup Stipple sees this as a waste—and an opportunity. Last fall the company rolled out a system that allowed publishers to add a new layer of information to online photos: “dots” that reveal details about people, places, or things when you touch or mouse over them. The details can include links, and if those links lead to e-retailing sites, the traffic they generate can turn into affiliate commissions—-creating one more much-needed revenue stream for publishers.

But as it turned out, there were three problems with Stipple’s original system, CEO and co-founder Rey Flemings told me recently. First, the publishers who adopted it weren’t always willing to put in the time to tag new images with useful information. Second, the tags they did create weren’t always accurate or consistent; the same black chair might be identified as one model in one picture, and as another model in the same picture on a different site. Third, publishers didn’t always have the proper licenses to use the images they were tagging—and while publishers might be able to get away with posting unlicensed images, the copyright holders are likely to come calling once they start make money from those images.

So Stipple, which has venture funding from Kleiner Perkins Caufield & Byers, Floodgate Fund, Parkview Ventures, Quest Venture Partners, and a number of prominent angel investors including singer/actor Justin Timberlake, went back into the darkroom to develop some new features. Late this spring the company introduced a “2.0” version of its system, divided into four components to solve the problems publishers were encountering and to get a new faction directly involved: namely, the makers of the products shown in Web images. Now Stipple comes in the form of products aimed at different audiences; Flemings calls them Lens, Pipeline, Network, and Want.

“I believe this new suite of products finally creates an image ecosystem,” says Flemings. “We think there is a great business to be built in providing accurate information to users when and where they want to see it.” And so far, Web users seem to be eating up the information on Stipple-enabled sites. When there’s a “people dot” on an image, about 5 percent of all visitors touch it, Flemings says, and when there’s a “product dot,” 12.5 percent of visitors touch it. Those engagement rates far exceed the fraction of people who click on traditional banner or text ads, which is typically below 1 percent.

Flemings walked me through a basic description of the startup’s new products and how they fit together. (You can see the system in action at Stipple’s blog.) It starts with Lens, an ever-growing collection of licensed editorial images from many of the country’s leading photo agencies and ad agencies. “The purpose is to assemble all of the editorial images that matter—the ones with really large audiences of people concerned about who’s in them and what’s in them,” says Flemings. For each incoming image, Stipple creates a permanent digital fingerprint. That way, dots created for a given image—say, a dot describing the dress Katy Perry wore at the Grammy Awards—will stay associated with the picture wherever it ends up on the Web. The idea, over time, is that publishers who subscribe to Stipple’s service will get more and more of their images through Lens, solving the unlicensed-image problem.

The second product is Pipeline, where the dots actually get created. Basically, Stipple’s solution to the workload problem is to have product makers themselves handle image tagging. Explains Flemings, “At the end of the day, who has the greatest motivation to tag, and who knows what’s in the photo? Is it the publisher? The person who took the photos? No, it’s the company that made the dress.”

All of the fingerprinted photos in the Lens collection are made available to brands, who use Pipeline—a Web-based image tagging platform—to attach dots with product descriptions and links for consumers. As soon as an image gets tagged, the dots show up with every instance of that image, even if the tagging occurred after the image was published. (And in a neat bit of digital magic, the dots stay intact and show up in the right locations even if an image gets resized or cropped.)

Publishers are still free to add their own dots to images—but by handing over most of the tagging work to manufacturers, Stipple hopes to solve the problem of inaccurate, inconsistent product information in images. “It’s extra work for brands, but with a high-value payoff,” says Flemings. “We started the [Pipeline] effort a very short time ago and we have been literally inundated with inbound interest from brands, who indicate that it’s a capability they’ve wanted for a very long time.”

Network, the third piece of Stipple’s ecosystem, is basically a rebranded version of Stipple’s original technology; it’s the system that ties together photo agencies, brands, and publishers. It ensures that “stippled” images selected by publishers and bloggers show up with the freshest tags; it also records all user interactions with dots and tracks the commissions due to photo agencies and publishers when dots generate affiliate revenue. Publishers don’t have to do much to activate Network—they just supplement their website template with a bit of code that calls Stipple’s system whenever a user’s browser loads a page with stippled images.

Finally, there’s Want, the only part of Stipple’s system focused directly on consumers. It’s a bookmarking service that lets Web surfers build a wish list of information about products in stippled images. Most product dots now include both a “shop” link that goes directly to an e-retailing site and a “want” link that adds the product to the user’s wish list. “There are clearly examples when a user doesn’t have the time to buy something right now; they want to keep moving,” says Flemings. “When they press ‘Want,’ it’s like a universal shopping cart.” Importantly, affiliate revenues on purchases from the wish list get credited to the original image publisher, even if the purchase occurs months later.

It’s still early days, but Flemings uses adjectives like “massive” and “staggering” to describe the rates at which users click Stipple’s want and shop buttons. (He declined to share exact percentages.)

Other companies offer systems for attaching product tags to Web editorial images, but Flemings says Stipple’s advantage, reflected in the high engagement and conversion rates, is in the accuracy and relevance of the information added by brands. “Engagement and accuracy are heavily related,” he says. “Our competitors have chosen to provide a different kind of experience. They’ll show a picture of Kristin Bell in a gray knee-length dress and they’ll include a link to a gray knee-length dress, but it’s not the same dress.” The logic is that the average Web surfer probably can’t afford Kristin Bell’s gray dress, so the link should lead to a less expensive knock-off. But Stipple’s philosophy is that “you may not be able to afford that red Ferrari, but you are free to aspire to it,” says Flemings. “We believe that when you mouse over a photo, that is a ‘tactile search,’ and the user should have the same expectation of accuracy as when they search using a keyword.”

Flemings says Stipple is still lean, with just six employees including business development exec Paul Melcher, a veteran of the image licensing industry. But expect to see Stipple’s ecosystem of services keep growing—Flemings says a big announcement is coming late next month. Already, 20 to 25 percent of all new editorial images are captured by Stipple’s system, “and that number is growing every time we add new agencies,” Flemings says. “It’s our goal to close out the year with a very high penetration of the images that matter [in the product merchandising area]. And you will see our product diversify quite rapidly to other populations of images.”

Wade Roush is a freelance science and technology journalist and the producer and host of the podcast Soonish. Follow @soonishpodcast

Trending on Xconomy

By posting a comment, you agree to our terms and conditions.

Comments are closed.