Yahoo Rallies For Better Search Metrics


Traditional metrics to measure searches don't work well with tools like Google Instant and Yahoo Search Direct, or with emerging technologies like voice and gesture.

Until now, the industry has relied on data firms like comScore, but Shashi Seth, senior vice president of search products at Yahoo, believes the problem represents deeper issues and the responsibility should be with execs at the largest engines.

"People at Yahoo, Google and Microsoft need to come together and share opinions on new metrics," he said.

Advancements in search and marketing technologies point to inadequacies in long-established methods to accurately measure query volume. Yahoo has begun to test the ability to start a query on one device and continue it on another. For example, searching for a local restaurant on a desktop PC, finding a location and saving the search to continue it on a mobile device, such as iPad or smartphone.

Signing into a Yahoo ID account allows the engine to connect searches on smartphones to tablets, desktops and laptops. Searchers will not only benefit from being able to save queries and continue them on another device, but advertisers will have an option to target ads across devices. Seth said Yahoo could release the tool within the next few months.

Both Microsoft and Google have been working to personalize searches by integrating screens and ad targeting across platforms when signed in through a Google account or a Windows Live ID account.

"Searches have been sliced and measured based on overall clicks, but not by the individual users or history," Seth said. You can imagine how social signals and interest graphs have begun to improve pushing out information."

Seth said crossing screens creates a huge data problem, but search engines are good at analyzing tons of data in short periods of time. Data also becomes an issue for image search -- another challenge Yahoo continues to address. The company is working on a feature supported by recognition software that lets consumers look for similar items and drills down to specific criteria on queries.

For example, a search for a lookalike Manolo Blahnik shoe at a cost of less than $200. Image matching and recognition means the algorithms must remove the colors and the curves of the shoes, and reduce the photo to a mathematical equation. The technology must be able to search large amounts of data and return results in milliseconds, either a typed query of voice search.

New search features and tools that provide answers, rather than links, produce ample data. So Yahoo has begun to build a content index, instead of Web index, for every HTML page that exists. It connects information in a way that search engines have not done before. "It's been a large undertaking, but should solve pretty deep problems when finding exact answers to questions," Seth said. "We're doubling the size of the index yearly."

Next story loading loading..