The Seekers

When natural search isn't cutting it, just get a little help

Imagine search that didn't need to be found. Well, it's not exactly search, then, is it? More like mystical psychic predictions (unless you're the one coding), recommendations-based algorithms are along the lines of saying to searchers, "Don't call us, we'll call you." One of the problems with this sort of thing might best be displayed in the title of Irvine Welsh's short-story collection If You Liked School, You'll Love Work. It doesn't always pan out, does it?

In The Graduate, young Benjamin Braddock gets a good piece of advice: plastics. (Of course, he doesn't take it, instead launching into a disastrous affair with the predatory wife of the hapless career counselor.) If only Mr. Robinson had made his recommendation online, Benjamin's life might have been so different. Today's netizen is all too happy to take advice from complete strangers, especially about shopping.

In this era of extreme cynicism about marketing, people are much more likely to trust online information about a product that comes from a third party, according to JupiterResearch. "Sometimes 'friends' are barely even acquaintances," says senior analyst John Lovett. The personal touch even beats the godlike algorithms of the search engines that promise to fine-tune organic search results. Consumers have been less than enthusiastic about opting in for Google's Web History and Recommendations, both released more than a year ago, or Yahoo's Mindset, out since 2005. "The expectation is that, when a visitor uses a search engine, they're going to get relevant results," Lovett points out.

Come and Find Me

Google describes its recommendations as "searching without a query" - somewhat alarming for search marketers. After all, if this takes off, it could mean that someone who relied heavily on Google's recommendations might never see your carefully managed search ad. And eliminating the query sort of eliminates the entire basis of the search-marketing industry. But that could be for the best: If the relevance engines are working well, that consumer wasn't interested in the first place.

This feature hasn't exactly made waves, but there could be a subtle effect on the results seen by heavy users, says JupiterResearch analyst Evan Andrews. If a searcher leans toward certain kinds of sites, those will show up on SERPS more and more - and marketers can no longer know where they rank in an individual's results.

"It's complicated on any search because you want to track and measure results. This is yet another complexity on top of the system - but it could yield great results," says Ellen Siminoff, chairman of search-marketing technology and services company Efficient Frontier.

That means marketers should pay even more attention to optimization tactics like tagging and title or meta tags, Andrews says. "The first impression is more important than ever, because once a user thinks content is relevant, they will always get that bias."

Microsoft has given its own spin to shopping recommendations with Live Search Cashback, a hybrid between recommendations and paid listings. They call the rewards portion of the program, where consumers accrue cash bonuses when they buy from participating merchants, "paid engagement," and the company believes it will lead to higher conversions for retailers, while the CPA model will make their ROI calculations easier.

Brad Goldberg, general manager of Live Search, says Cashback combines elements of search and discovery. When people look for a product, they may find and be drawn to merchants participating in the program. They can also start by searching for rebates. "The Cashback Offers page is a more browsing type experience," Goldberg says. "It allows people to discover at a category level."

Live Search is also blending recommendations into shopping searches by delivering summaries of user reviews. In addition, Goldberg says, Farecast, offered through Live Search, provides predictive pricing technology that could be applied to other kinds of search. When someone searches for an airfare via Farecast, it uses data mining to predict whether the fare might go up or down in the next few days. That kind of on-the-fly predictive modeling is a feature of the technology offered by vendors who say they can provide individually targeted recommendations for content and products.

Find and Seek

There's one crucial problem with search: You tend to find things you already know about. (If you didn't, you wouldn't be able to search for them, would you?) Recommendations, on the other hand, are all about discovery, suggesting things you might have looked for if you had known about them in the first place. Recommendations have been around since the earliest days of the Web, when Amazon began suggesting books. But relying solely on an individual's past behavior doesn't always lead to the best recommendations.

"Amazon continued to rely on historic information, so every time a visitor went to a site, he got repeated offers for similar products. I bought a watch, and now every time I go, I get offers for a watch. That model is broken," says David Selinger, CEO of richrelevance - and part of the R&D team that developed Amazon recommendations. Richrelevance is a new vendor that provides merchandising and personalized recommendations on e-commerce sites.

The latest technologies push recommendations onto the page at the moment someone is browsing. The newest trend, says Jupiter's Lovett, is to combine historical cookie information and audience segmentation with an analysis of immediate click-stream behavior to target recommendations on the fly. "Site operators must respect the context of the visit. I want them to tune into what my intent is at that moment," Lovett says.

Suggesting to customers on your site - customers you've already acquired - items they're highly likely to be interested in is a way to boost the bottom line without increasing customer acquisition costs. These suggestions can be in the form of personalized on-site search results, so that someone who searched for "sofa" first sees the five sofas most likely to appeal to him, or in the form of other options to consider - the "Want fries with that?" technique taken to a new level.

The basic methodology of next-generation recommendations engines is to use the visitor's cookie information to narrow down the options, then perform ongoing analysis of the click-stream and compare this behavior to that of all other users. When it finds the best matches, it looks to see what those people did or bought and offers the same to the current user.

Recommendations won't necessarily take the place of search, but they can be a way for marketers to reduce their reliance on search, Selinger argues. The cost of search advertising continues to trend higher as competitors bid up keywords. Meanwhile, most consumers start their shopping trips with a visit to a search engine, not to a retailer's site.

"Loyalty is being sucked away from retailers," Selinger says. "So, how do we invest in and leverage the customers we already have? Instead of trying to push shoppers down one channel or another, we present the consumer with options."

Content sites suffer just as much from the lack of consumer loyalty. David Marks, cofounder and CTO of Loomia, points out that when searchers follow a link from search to a specific piece of content, "they're there, they watch one video - and then they're gone." Publishers battle this hit-and-run effect with a variety of tactics, including delivering the searched-for content on the site's front page and showing lists of "most popular" or "most e-mailed" stories. But personal recommendations can boost clicks, as Loomia, which offers a hosted service for content and product recommendations, discovered firsthand in its earlier incarnation as a portal.

"Users don't trust machines to blindly recommend things as well as they would trust a friend," Marks says. So Loomia added social context by producing SeenThis, a Facebook app that lets people use their social networks to filter content on Loomia's CTRS went up as much as 500 percent, and a new business model was born: supplying SeenThis tech to sites including CNET and The Wall Street Journal online.

Right now, only people who have enabled Loomia on their Facebook accounts see recommendations from their friends or networks; Marks says the company plans to support other platforms as well, and its service can also be used to power recommendations for e-commerce sites. Efficient Frontier's Siminoff says marketers should test against recommendations engines just as they work to optimize their placement in organic search results. While the search landscape is more complex than ever, ultimately, she says, "the end user and the market will benefit."
Next story loading loading..