Web Advertisement Rong Jin Basic Forms of Advertising

Web Advertisement Rong Jin

Basic Forms of Advertising o Brand advertising n o Creates a distinct favorable image Direct-marketing n Advertising that involves a "direct response”: buy, subscribe, vote, donate, etc, now or soon

Examples

Web Advertising There a lot of ad. on the web …

Search Advertising

Display Advertising

Revenue of Internet Advertising

Different Forms of Internet Ad. Internet advertising revenues by format in percents - 1998 -2004.

Search Advertising o o Advertising that is related to the web search Becomes the driving force sustaining monetization of Web services n n o 63% of the advertisers want to target users according to their interests 49% want to target users on their geographical locations Two types n n Keyword-targeted advertising Content-targeted advertising

Keyword-targeted Advertising o o Introduced by Overtune in 1998 Ads are related to the user’s queries Paid list

Keyword-targeted Advertising o Advertiser n n Associate ad with keywords Bid on the keywords o o o Presentation of ads n n n o The higher the bid, the greater the chances that the ad will be shown the paid list Only pay when users click on the ad A title, a short description, and a URL Creative = title + short description Buzzing word: “free”, “click here”, “enter now to win …” Conversion: the user jump to the landing page and starts transaction

Content-targeted Advertising o o Ads are related to the content of triggering pages Automated selection of ads n o Assign each page with keywords and key categories The dominant contextual approach in Web marketing. Paid list

Search Advertising Business o All large players (Google, Yahoo, MSN) now own: n n Web search engine Advertising platform + network o o Google n n o Own engine for keyword ads (Adwords) Bought Applied Semantics (context ads, Adsense) Yahoo n n n o They are both a search engine and an ad agency Bought Inktomi, Alta. Vista, Allthe. Web, Overture Search: Inktomi Ads: Overture (keyword, context) MSN n n Used search from Inktomi + own corpus. Has now its own (Bing) Used ads from Overture, has now own platform (“Ad. Center” in 2006)

Search Advertising Network o Failure of search advertising in the early days n Poor matching between ads and keywords/content o n Questionable practice o o Irrelevant messages annoy users popping up ads in pages without the permission of their publishers, associating their images with improper companies Solution: search advertising network n Form a network so that all the participants are benefited

Search Advertising Network Search advertising network

Search Advertising Network: Broker o Responsible for the maintenance of the network. n n o Decide advertisers and publishers that participate of the network Decide the publishing policies Responsible for the auction system n n Interfaces, databases, controlled vocabularies Match ads to keywords/content

Search Advertising Network Search advertising network

Search Advertising Network: Advertiser o Goal: n n o o o Popularize their brands and commercialize their products and services. Their ads are referred to quality users Compete among themselves for keywords by bidding in an auction system. Pay to the broker according to the traffic Tune the parameters dynamically according to the performance report

Advantages of Web Ad. o Accountability n n o Measure the performance and provide feedback about marketing campaigns and strategies. Dynamically adjust their strategies Flexibility n Target users according to their past behavior, interests, demographic information, and local information

Search Advertising Network Search advertising network

Search Advertising Network: Publisher o Goal: n o o o Simply displaying ads will annoy most users Search-related ads are more targeted positive user attitude Provide brokers with the description of their pages n n o Monetize their pages through the loyalty of their audience Manually assigned keywords or categories Essential topics derived automatically Diverse publishers n n search engines and online directories Small publishers

Search Advertising Network: Users o o Goal: getting relevant information from the publishers. Describing their information needs by keywords or by surfing in Web pages n o Average four hundred million queries per day in 2003; 40% were of commercial nature Occasionally click on the ads exhibited, jump to the advertisers’ pages, and start commercial transactions.

Example (I) o o We advertised a tutorial of WWW 2006 Bid on keywords n n n www 2006 “www 2006” “www conference” Prabhakar Raghavan Andrei Broder $0. 20 per click $0. 07 per click $0. 10 per click $0. 07 per click

Example (II)

Example (III): Report

Distribution of Cost Per Click (CPC) Not a Zip’s law !

Challenging Questions o For advertisers n n n o What words to buy? How much to pay? Arbitrage among keywords/suppliers, try to cherry pick demographics How to fight with spam Dilemma between the overall cost and the number of clicks For publisher (search engines) n How to price the words? o n n Let the market decide: bidding! When to place ad? (a matching problem) How to price “extended” or “broad” match ? Long term costs (bad user experience) Dilemma between relevance and short term profit

Find the Right Ad o Relatively simple on bid phrases Similar to Web Search, but o Ranking depends on both bids and relevance n Each ad entry = a “small page” n Different metadata (keywords, title, URL) n Enormous historical data (billions of searches/ad/clicks records) What about queries on which there is no bid? o n

Find the Right Ad o Advertiser can bid on “broad queries” and/or “concept queries” n o Bid on any query that contains “sigir”or the “sigir”concept. Pitfall: partial matching n n Ad: A seller of car water pumps might bid on “pumps”. Query: “breast pumps” or “black pumps”

Auction and Pricing o o o Advertisers compete for each keyword Advertiser pays the search engine when a searcher clicked on a displayed ad Challenges n n n Which keywords to bid? How much to bid for each keyword? How to adjust the biding according to the performance?

Search Advertising System o o o D: a collection of documents (i. e. , ads) q: a user query M: q D {0, 1} n n o A matching function Decide if an ad is relevant to the query q R: q D [0, 1] n n A ranking function Decide the relative rank of an ad to query q

Search Advertising System Query Miele Ad Database ad 1 ad 2 ad 3 ad 4 ad 5 ad 6 Identify relevant ads by M Bid different keywords Click Advertisers Data ad 1 ad 3 ad 4 Rank ads by R Change bid

Relevance Matching o o The quality of relevance matching affects the user’s perception of the web content Exact match and approximate match n n Exact match: query = ad keywords Approximate match: partial match o o Similar to typical IR methods Query expansion to include synonyms and related terms

Conversion Rate vs. Query Size Why? o Intuition: the longer the query the higher the conversion rate

Conversion Rate vs. Query Size It is because most brand names are single word o Intuition: the longer the query the higher the conversion rate

Conversion Rate vs. Query Size Why? o Too specific queries

Ranking o Satisfy interests of different parties n n n Users: relevant information Advertisers: quality traffic at a minimum cost and with a minimum risk of negative user attitude Brokers and publishers: maximize their revenues at the minimum risk of negative user attitude

Click Through Rate vs Ranking A simple strategy: always put the most expensive ads on the top (what is wrong with this idea? )

Paid Placement Strategy Perceived Relevance o o Expected Profits Rank by willingness to pay (WTP ranking) Rank by willingness to pay times click-through rate (WTPC ranking), i. e. , the product between their bids and their expected click-through rate. n The reward for a click is larger if it is received in a lower rank.

Fraud Detection o o The more publishers have users clicking in the ads shown in their pages, the more advertisers will pay to them. Publisher can fake traffic to attract advertisers n o Comp. USA spent more than $10 millions in 2004 due to fake traffic. Characteristics of fake traffic n n n Distribution of clicks over time Distribution of clicks over users A standard classification

Measurements o o Advertisers need to get detailed feedback about their performance Cost per thousand impression: n n n o Traditional pricing model Payments were measured mainly based on the quantity of impressions of ads Uncertainty on the benefit of advertisers Performance based metrics n Direct estimate of advertiser’s return of investments

Measurements o o o User clicks may not convert Require reliable estimate of the return of their investments (ROI). Keyword-targeted advertising performs better than content-targeted advertising n Users are less likely to generate a conversion while surfing.