Web Advertising 12/26/2013 CS190: Web Science and Technology, 2010
Today's Plan Logistics Understanding searchers (Commercial Perspective) Search Advertising Next project: Google advertising challenge CS190: Web Science and Technology, 2010
Logistics Project 1 feedback/review returned: by tonight Thursday: Quiz (web search, advertising basics) Today: Project 2 details (will continue on Thursday) 2/26/2013 CS190: Web Science and Technology, 2010 3
Segway into Search Analytics Google trends http://www.google.com/trends/explore Google Flu trends http://www.google.org/flutrends/us/#us 2/26/2013 CS190: Web Science and Technology, 2010 4
Understanding Website Audience Who are these people? 1. Analyze URLs 2. Analyze Keywords Slide credits: Panos Ipeirotis, NYU
Analyzing Referral URLs: Demographics
Analyzing Referral URLs: Affinities QuantCast for CNBC
Analyzing Keyword Queries: the "intent funnel" What is the intent of customers that type such queries? Hint: What they searched before or after? Search Funnels: http://adlab.msn.com/searchfunnel/ How can you catch customers earlier? What customers do when they leave?
Keyword Composition Classification of queries in search engines Navigational Transactional Informational Keywords with your brand are navigational In principle, there should be no competition Site should rank high Typically in the head of the distribution Aware/Loyal Customers Non-branded keywords are the real target Competitive/Searchers Customers
Get Keyword Demographics Keywords have demographic signatures Microsoft adcenter Demographics Prediction: http://adlab.msn.com/dpui/dpui.aspx Quantcast http://www.quantcast.com/people.com#!demo Example: [posters]
Get Keyword Demographics Example: [Jessica Alba]
What is Search Engine Marketing (SEM)? Search Engine Optimization (SEO) Improve organic result rankings on search engines Adjust elements important to search engine algorithms Pay Per Click Advertising (PPC) Also known as sponsored results, advertising that appears to the top, bottom, or side of search engine natural results Typically cost is generated on a pay-per-click model, where advertiser pays for each click a searcher makes on an ad, vs. the traditional online marketing method of cost per impression.
Search Marketing Facts 80.6% of searches use 2 or more words (current trend is 2-3 word searches) 87% of searchers do not search past the first page of results 33% of searchers believe that the top ranking is also the top brand of the industry. 56% of retailers say that 10-40% of their total orders come from PPC advertising 62% of users do not know the difference between paid vs. natural listings (source: Pew Search Engine Users Study)
Why Search Engine Marketing?. Highly effective and targeted form of advertising High effectiveness given the cost
Some slides adapted from DoubleClick, Yahoo! WEB ADVERTISING 2/26/2013 CS190: Web Science and Technology, 2010 15
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 16
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 17
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 18
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 19
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 20
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 21
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 22
The Million Dollar Webpage http://www.milliondollarhomepage.com/ http://en.wikipedia.org/wiki/the_million_dollar_homepage 2/26/2013 CS190: Web Science and Technology, 2010 23
Web advertising Banner ads Direct advertising (search ads) Focus: direct textual ads 24
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 25
Ads as information I do not regard advertising as entertainment or an art form, but as a medium of information. [David Ogilvy, 1985] Advertising as Information *Nelson, 1974+ Irrelevant ads are annoying; relevant ads are interesting Vogue, Skiing, etc. are mostly ads and advertorials 26
Ads as information supply User profile & context Activity context: Browsing a certain content Avail info supply: Ads inventory Ad Selection Platform Matching Ads Economics Feedback User action: Click-thru Action 27
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 28
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 29
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 30
Content Ads 31
How it works Advertiser I want to bid $5 on canon camera I want to bid $2 on cannon camera Ad Index Sponsored search engine Engine decides when/where to show this ad. Landing page Engine decides how much to charge advertiser on a click.
Search Ads 33
Anatomy of a Search Ad 34
Landing Page Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 35
Basic Revenue Models CPM= cost per thousand impressions Typically used for graphical/banner ads (brand advertising) CPC = cost per click Typically used for textual ads CPT/CPA = cost per transaction/action a.k.a. referral fees or affiliate fees Typically used for shopping ( buy from our sponsors ), travel, etc. but now also used for textual ads (risk mitigation) 36
Keyword matching: issues For advertisers What keywords to buy? How much to pay? Spamming is an economic activity For search engine owners How to price the keywords? Let the market decide: bidding! For both What are good performing keywords? What extra keywords to buy/sell? 37
Digging deeper: beyond keyword matching Relatively simple on bidded keywords What about queries on which there is no bid? Advertiser can bid on broad queries and/or concept queries Suppose your ad is: Good prices on Seattle hotels Can bid on any query that contains the word Seattle Problems What about query Alaska cruises start point? What about Seattle's Best Coffee Chicago Ideally Bid on any query related to Seattle as a travel destination Should these broad matches be priced the same? 38
Finding the right ad (cont d) Ads Database = Keywords + Title + Description + URL Ad Query = Search Keywords + Context Search problem similar to Web Search, but Ads have different structure Ad database is (somewhat) smaller Ad database entries are small pages *+ URL+ Ranking could depend also on bids 39
Similarity to Web search Web search trend over the past decade: From syntactic matching (common strings between query and document) to semantic matching (understand the query intent) Sponsored search From ads driven directly by the bid phrase (syntactic) to ads driven by the search context (semantic) 40
Three sub-problems in ad selection 1. Match ads to query/context 2. Order the ads 3. Pricing on a click-through Econ IR 41
2. Order the ads For each ad we now have A query-dependent score and A $ bid from the advertiser that can be used to compute an Econ score Ordering of retrieved ads Most generally, composite query+econ score, e.g., expected revenue Original GoTo/Overture scheme: Order by bid 2. Order the ads 42
How to price a click? For the keyword emory, assume advertiser has a value of $10 per click. How much should she bid? How much should she be charged? The value of a slot for an advertiser, what he bids and what he is charged, may all be different. 43
PPC: Pay Per Click PPC Terminology CTR: Click Through Rate Percentage of clicks to impressions CPC: Cost Per Click Price an advertiser pays for each click on his/her ad Daily Budget Amount advertiser is willing to spend, each day, on PPC ads
The PPC Auction Model Basic model Highest bidder gets highest slot Winner pays winning bid Not ideal! Why? Vickrey-Clarke-Groves (VCG) auction Second price auction Optimal for single slot Reveals true willingness to pay, no need to be strategic Winner for slot i pays maximum bid of bidder that get i+1 slot Still not the best for Google, Yahoo, MSN etc. Why?
Ad Quality Score + Landing Page Quality Search engine needs repeat customers Needs to improve user experience for long term Bid transformation: (Bid $ amount) x (Ad Quality Score) ~ (Bid $ amount) x (CTR rate + relevance) In other words, rank by expected revenue for Google High quality ads Lower bid amount (incentive compatibility) Main difference between Yahoo and Google until 2007 Details for quality score: https://adwords.google.com/support/bin/answer.py?answer=10215
Example Google AdWords Report 47
Case Study: Cost per Click and Holidays CPC increases before the holidays Should advertisers refrain from bidding? How to approach the problem? CPC increases but conversion rate (CR) increases more! Cost per Acquisition (=CPC/CR) drops! Holiday season offers better bargain than other days
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 49
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 50
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 51
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 52
Spring 2009 Eugene Agichtein CS 190: The Web: Concepts and Technology, Emory University 53
Improving PPC ROI The Long Tail Concept The Long Tail by Chris Anderson An aggregate of less popular products can sell more in the long tail than the most popular products Same applies to PPC An aggregate of specific, less searched terms can provide better ROI than highly searched terms
The Long Tail Again
Does It Really Work? YES. Major athletic retailer case study 80% of PPC sales come from long tail keywords Specific, product-name keywords Only 20% of sales come from broad terms, like BRAND shoes or BRAND jacket Most long tail keywords are inexpensive [web hosting] $8.30 [freebsd web hosting] $0.10 Targeted at customers with specific needs, closer to purchase
Project 2: Google Ad Campaign NOW: Pick your teams: 2 or 3people each. Preferably, not same as Project 1 Pick a small business (or non-profit or student organization) with a website that has not used google AdWords in previous 6 months Sign up for Google advertising challenge Today/tonight: Setup Google AdWords account; send your CID to me Monday 3/4, send professor your campaign strategy (1 page, including website of business you want to promote). I will give you feedback in class on Tue. Thursday (3/6) Your Advertising campaign starts and continues at least through spring break. Thursday 3/21 each group does a short presentation about campaign results. Wednesday, 3/27, submit a short (1-2 page) report about your ad campaign. Note: If you want to be considered in the Google competition for fame, fortune, and extra credit, there are additional requirements, discuss with the instructor. 57
Step 1: selecting a business Try to work with businesses relevant to the types of search queries that Google users conduct. A good example would be a traditional retail business, such as a home wares store, a vintage fashion store or a niche beauty store, or a nonprofit/student organization. You should be aware that Google has content guidelines and will not run AdWords for sites promoting inappropriate items such as academic aids, alcohol, bulk marketing, counterfeit designer goods or cigarettes. Please note that the business or organization should not currently use AdWords in any capacity and not have had an active account within the last 6 months. 58
Example: Emory CS PhD Program https://adwords.google.com/select/snapshot 2/26/2013 CS190: Web Science and Technology, 2010 59
Start planning your campaign Pick a business preferably, non-profit/student org, but real small business OK Follow the multimedia lessons in Google Adwords: http://www.google.com/adwords/learningcenter/index.html Especially: 1(a, c) 2 (a, b, c, d) 60
Further Reading: Student Guide to the Challenge http://www.google.com/onlinechallenge/student_guide.pdf Note that it contains additional requirement only needed to participate in Google s competition. Not required for the class project. 61