Shopping Search in the Market for Feed Standards

Alan Rimm-Kaufman is an idea guy. That works out well, because as the head of The Rimm-Kaufman Group, he’s called on to offer search marketing and e-commerce help to retailers of all sizes, from start-ups to Fortune 500 companies.

So when he realized that many of the largest e-tailers, marketing agencies and search engines would be assembled in Atlanta earlier this month for the annual Shop.org First Look conference, an idea occurred to him: Why not use that opportunity to start the ball rolling on creating industry-wide standards for data feeds to shopping search engines? Just possibly, with some early agreement and a lot of luck, those standards could be in place by the 2006 holiday season.

Shopping search engines are gaining traction with both users and advertisers. That was made clear last year by the fact that three of the biggest deals in direct marketing involved the purchase of online engines devoted to finding products and/or comparing prices: of Shopping.com by eBay in June, Shopzilla by E.W. Scripps days later, and PriceGrabber by Experian in June. In addition, both Google and Yahoo! have taken strong steps to beef up their shopping search offerings. In further addition, a crop of smaller shopping engines have grown up to appeal to niche markets and offer specialized tools. And last week, from Ask Jeeves parent IAC, came notice that it too will field a comparison shopping engine named Pronto.

The problem is that a marketer looking to be included in these engines still needs to handcraft a data feed of product information, model number, SKU and price for each one. That calls for more custom-tailoring than most small online retailers can afford and may lead them to pass up the smaller engines that seem to promise smaller returns. For the big guys with plenty of IT resources, this balkanization of shopping standards creates tremendous inefficiency.

And the problem isn’t only one of standardized input: Shopping engines vary widely in the amount and usability of the data they report back to marketers on clicks and costs. That variation makes it even more difficult for marketers to plan efficient shopping-engine campaigns.

Rimm-Kaufman likens the shopping-search problem to the crazy quilt of track gauges that prevailed in the early railroad industry. “In the 1820s, the tracks were all different sizes, and you couldn’t move a train from one system to another without great difficulty,” he says. “In the short term, there may have been some benefit to that walled-garden approach. But in the long term, it was just friction and not good for the state, the railroads, the travelers or probably even the track makers.”

Retailers can outsource the data-feed heavy lifting to firms like RKG, but that just transfers the cost inefficiencies outward, Rimm-Kaufman says, and puts full shopping-engine participation further out of reach of the small players.

“We run an interactive marketing agency, but it struck me how much time we were putting into things that didn’t seem to be marketing work,” he says. “It was just dumb IT work.”

So Rimm-Kaufman spoke to the organizers at Shop.org and asked for some time and space to get retailers, engines and agencies together in a room for a preliminary discussion on where standards were needed and how they might be developed. He invited about representatives from all three categories and got about 40 confirmations. Bad weather delays meant that about 30 of those made the meeting, but they were from some of the most important brands in their fields: Yahoo!, MSN, AOL, Become and Shop.com among the engines; Red Envelope, CompUSA, Cutter & Buck, Coldwater Canyon, REI, Sierra Trading Post and Bare Necessities from the retail side; and from the agencies, Performics, Mercent, Mars and RKG. The National Retail Federation’s Association for Retail Technology Standards (ARTS) was also represented.

For two hours the group shared their frustrations with the current system, or lack of same. Many of these pain points broke out by category, Rimm-Kaufman says. Retailers were most stressed by problems involved in listing or displaying their products on the engines. For example, many retailers customize and even trademark their color descriptions. Fine for them, but the search engine taking the data feed needs to know whether “heather” is blue, green or purple in order to serve users who are looking for a purple sweater. Translating “Lilac Dawn” into a Pantone color number could help all parties involved.

Other merchant pain points a standardized shopping-engine data feed might address:

* Volume discounts: A greeting-card retailer might have one per-card price for quantities of 250 or less, another for 251 to 500, and a third for orders of 501 cards or up. That marketer may ignore shopping engines that can’t accept and display those variable discounts.

* Some online retailers offer outdated merchandise; in fact, it’s a popular way for marketers to reduce old inventory. But shopping engines sometimes force the inclusion of a manufacturer’s product number on listings. If the product number has been purged, the item may not appear; if not, it may appear as the current product, guaranteeing disappointed buyers.

* Search taxonomy- that is, what a category is called. Rimm-Kaufman said one retailer is adamant that “they sell ‘trousers’, not ‘pants’”, a distinction featured in all their ad copy and one they don’t want to rewrite. But far more people search on “pants”; this retailer wants its listings to show under “pants” but always to refer to “trousers”. In the world of search, that’s not as easy as it sounds.

* Merchant control over keyword searches and ad syndication: Retailers may want to specify that their listings are served only when the correct search is done. For example, a lingerie seller may want to appear when undergarment shoppers look for “thongs” but not when 13-year-old boys do. And some marketers may have agreements with manufacturers that their ads will not appear on certain Web sites—something retailers have little control over now.

Rimm-Kaufman admitted he was a bit uncertain of the response from the search companies to any proposed standardization but found to his surprise that all the engine attendees (as well as Google, which discussed the issue separately) backed the notion of a unified shopping feed because they’re “bombarded” with incorrect feeds that require translation.

One unnamed engine told the group that of the millions of data rows they receive, each representing a product URL, 10% to 15% on average are broken so badly that they can’t be loaded and another 20% to 35% don’t fit the engine’s taxonomy—so that even though they load, users can’t find the products they define. That means as many as half the feed content that engine receives is useless to the shopper, and therefore earns the merchant no conversions and the engine no ad revenue.

Another engine said that in the “brand” column of their data feed, marketers were about equally likely to supply the manufacturer’s brand (which the engine wants) and the brand name of the retailer (which they don’t). Occasionally, marketers even inserted the name of the search engine itself.

Rimm-Kaufman was less surprised that the attending ad agencies were also in favor of standardizing data feeds. “We get paid for our time, and we spend so much time just getting the feeds not to be broken that we don’t have time to optimize or tune them,” he says of RKG’s workload. Being able to tell easily which listings work for which products and which should be turned off are obvious no-brainer benefits.

When it came to talking formats, Rimm-Kaufman says, the group broke along lines of size. Small marketers tended to favor simple text files because they were easy to compile. But the sense of the group was that both merchants and engines should have the option to offer and use feed features like keyword limitation and volume discounts, which won’t apply to everyone. Linear text files separated by commas won’t provide that optionality; you need a more robust format like XML. If the shopping search industry can settle on single data-feed standard, software makers should quickly begin offering simple “export to feed” applications for small marketers without IT savvy.

Finally, the discussion turned to how to generate these standards. In the past, suggestions have been made that the shopping-search industry simply adopt the systems used by one or another large product aggregator—Amazon, for example. But that’s not the way to generate a universal standard. “Large retailers such as Wal-Mart and Best Buy will object that Amazon is not the whole Web,” Rimm-Kaufman says.

That’s when ARTS stepped up and pointed out that it has successfully developed similar wide standards for other portions of the retail technology industry. “They have the process, and they have the experience and the credibility,” Rimm-Kaufman says. “This would be done in bright sunlight by a neutral organization, and the results would be completely voluntary.” Past ARTS standards projects have also used some component concepts—promotions and item categorizations, for example—that might be re-engineered for a unified data feed.

To top it off, ARTS had already scheduled a general association meeting for Feb. 27 in Menlo Park, CA—right in the search engines’ backyard. Now that meeting will include a few developmental sessions on the specs that belong in a standardized way to present product information on the Web.

And Rimm-Kaufman is working to get the word out to the online retail community that shopping-engine standards are coming, and that they should take this early opportunity to make their needs and wishes heard.