Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Method and System for Ranking Keyword Suggestions in Online Advertising

IP.com Disclosure Number: IPCOM000217456D
Publication Date: 2012-May-08
Document File: 5 page(s) / 145K

Publishing Venue

The IP.com Prior Art Database

Related People

Sebastien Lahaie: INVENTOR [+2]

Abstract

A method and system for ranking keyword suggestions in online advertising given an estimate of the value of the keyword suggestions to the advertiser is disclosed. The estimate of the value of the keyword suggestions is obtained by building a model based on keyword characteristics and observables such as bids of the advertiser and click-through rates.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 35% of the total text.

Method and System for Ranking Keyword Suggestions in Online Advertising

Abstract

A method and system for ranking keyword suggestions in online advertising given an estimate of the value of the keyword suggestions to the advertiser is disclosed.  The estimate of the value of the keyword suggestions is obtained by building a model based on keyword characteristics and observables such as bids of the advertiser and click-through rates.

Description

Sponsored search is a form of online advertising where advertisers bid for placement next to search engine results for specific keywords.  As search engines compete for the growing share of online ad spend, it becomes important for the search engines to understand what keywords advertisers’ value most, and what characteristics of keywords drive value. Ranking of an advertisement is determined by a combination of bid and quality score.  The quality score is used to capture relevance of the advertisement to the keyword.  The quality score represents probability of an advertisement being clicked in a search engine also known as click through rate (CTR).  Here, the advertiser only pays when the advertisement is clicked.  The price of a click is called cost per click (CPC).  Advertisers receive some expected value per click, which is private information, and together with the advertiser’s estimates of CTR and CPC, drives decision making of the advertisers.

Disclosed is a method and system for ranking keyword suggestions in online advertising given an estimate of the value of the keyword suggestions to the advertiser.  The estimate of the value of the keyword suggestions is obtained by building a model based on keyword characteristics and observables such as bids of the advertiser and click-through rates. The keyword characteristics are demographic and textual features of keywords.

A quasi-linear model of advertiser utility on each term is chosen.  Let   be the advertiser’s value per click on term , where an individual advertiser has terms in its account with used to index terms.  Let  be an extended real-value function that gives the expected cost per impression  of obtaining a click-through rate (CTR) on term .  The purpose of introducing into the range is to implicitly encode the domain of as .  For instance, negative CTRs are infeasible and would have a cost of .  It is assumed that is strictly convex and   differentiable.  Further, it is assumed that an advertiser’s utility for clicks on a term is quasi-linear in cost, and takes the following form:

Let be search volume for term , and let be the vector of cost functions. The aggregate utility to the advertiser of obtaining the vector of CTRs across the terms in account is . The CTR is chosen by the advertiser via its bids on the term and cost function is estimated.

The term subscript is suppressed on values, costs and CTRs for clarity.  Dual space is the vector space of all linear functions on .  Convex conjugate of cost func...