Reputation Management: How the system identifies relevant listings

How does the system identify relevant listings?

We use Best Match – an algorithm that determines if a listing belongs to a business’ profile. Best Match compares the anchor data (Name, Address, and Phone Number plus Website) to the information found on the listing, determines if a listing is worth considering, and then ranks the listing on how closely it matches the anchor data.

A listing’s page is determined as a “potential listing” if the best match score reaches a threshold. By default, this is 50% (140), but on some sources, we expect a better score or don’t expect as high a score. For example, Twitter pages don’t allow users to enter much information, so its threshold is quite low. Whereas Yahoo! Local has a lot of information and is usually quite accurate, thus its threshold is rather high. If none of the listings returned in the search reach the threshold, there is no Best Match, otherwise, the highest-scored listing (or listings in the case of a tie) are the Best Match listing(s).

The best match considers phone, name, city, zip/postal, and address independently and scores the match between 0% and 100% for each. Each type of data has a different weighting on the Best Match result:

• Phone: 1.00
• Name: 0.55
• City: 0.10
• Postal code/ Zip: 0.55
• Address: 0.70

A phone number has the most influence and the city has the least. A perfect score is 290, but if there are multiple listings above the threshold, all listings with scores greater than 90% of the listing with highest score will have a tie since a profile could legitimately have multiple listings on a site and any listings with such similar high scores are likely real listings.