Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Online advertising


Published on

The business of advertising, Pro & Cons, cost and benefits. Creatives effectiveness

Published in: Technology, Business
  • Be the first to comment

Online advertising

  1. 1. Data Management for the Web Giacomo Giorgianni
  2. 2. Outline:  Briefing  Monetization  Auction Mechanism  Types of online ads  Approaches to online ads  Case Study: Modelization of the effects of different Creatives and Impressions history  New Trends (?)  Considerations
  3. 3. “QUITE” A BIG BUSINESS More than 36 BILLIONS $ in US during in 2012!! United States Only Objectives: • Increase visibility of a brand (Brand ads) • • Stimulate users to immediately buy some product Collection of user’s data and behavior to deliver him more appropriate ads Main Actors: Advertiser, Publisher, User Benefits for users: • • Free usage of applications (Too) Quick responses to their needs Briefing
  4. 4. Briefing – Pro & Cons: • Ubiquity • Frauds • Speed • Fragmentation • Low Cost • Ad-blocking • Measurability • Banner Blindness • Creative • Privacy concerns  Difficulty to track users • Customization of target
  5. 5. Briefing – Some history May 1978: Gary Thuerk, emailed ARPANET's user, DEC computer. 1993: First clickable ad sold to a Law firm by Global Network Navigator. Have some time? CHECK THIS OUT!! HotWired made banner ads mainstream 18 January 1994: Large scale (RELIGIOUS) email born SPAM. 1998: the first search advertising keyword auction
  6. 6. Let’s talk about money: Pay per click (PPC) 32% : Payment based on number of click received from ad. Not good for brand awareness Pay per action (PPA) 2%: Ad is clicked and user performes desidered conversion (purchase, form fullfilling, … ) Pay per impression (PPI) 66%: Based on number of times ad is shown. Usually stock of 1000 (CPM) Widely used in Display Advertising
  7. 7. Auction Mechanism:  Bid for keywords or better position/ranking  First-Price Auction:     English auctions: public offers Sealed-bid auction: Single and secret offer Winner pays the amount he bid (the highest) Bid are lower than WTP of bidders  Want some profit  Second-Price Auction:  Highest offer wins, second offer is paid  REVENUE!!  GSP (Generalized second-price): Ranking in slots assignment based on bid+quality. Widely used among Search engines
  8. 8. Main Types of ads:  Email & Newsletter marketing: An ad copy inside email message Consent  Opt-in /Opt out • Search advertising: Pop Under: Pop – up & Advertisements on resultsover the main Small window pages Based on queries. browser Sponnsored Search • Display Ads: Multimedia content appears on Web pages. 4 main types of Display ads (Which WE know very well!) Email advertising
  9. 9. Interstitial Next slide will be available in 7seconds. 7 4 2 6 3 5 1
  10. 10. Frame ad FLOATING ADS.. QUACK!! AND SO ON…
  11. 11. Technological PoV: Approaches Filtered: Specification of general Constraints (time, age..) Untargeted: Fixed ads displayed for a scheduled time period Personalized: - Ads exposed based on user’s behavior (history, data…). - Machine Learning and Web Mining - CHALLENGING!!
  12. 12. Technological PoV: Challenges Objective: Exploit users’ navigation history to deliver better ads General Problems: Technical Problems: • Preferences vary over time • Cold start • Inaccuracy of information • Potential customer vs Information seeker • Appropriate learning technique • Privacy constraints • Boredom prevention
  13. 13. Study case:  Facts:  Individual who sees an ad occasionally treated as individual who sees it repeatedly  different goodwill wrt the ad.  Not all creatives have the same effect on individuals.  Act: Mathematical model that consider: • Importance of different ad creatives along the campaign • “Goodwill” advertising response model • Effect of individual’s ad impression history on future exposures
  14. 14. Study case – The boring part: Ad Stock: A= Ad stock i = individual t=time α=decay E = Effect of all creatives AD= Effect of the Whole Campaign Wearout Restoration C= Effect of the creative j R= Restoration Rate ρ= restoration param. τ= time from last exposure
  15. 15. Study case – The boring part: Data Likelihoods 3 related processes (zero-inflated): 1) mit: Impressions arrival; Poisson 2) vit: Visits; Poisson distribution 3) sit: Conversions; Binomial 3 parameters in the likelihoods: 1) λ: Impression rate parameter 2) μ: Visit rate parameter 3) p: probability of conversion after visit NB: (1-r): take account of 0-inflation. Modelization of visits and conversion parameters as functions of Ad stock. Xt: Vector of variables  time varying Fixed effects γ= vectors of coefficient Offline advertising effect 1 2 3 Effect advertising on behaviour
  16. 16. Study case – Model test: CONTEXT: On Automobile Brand 10 weeks in Summer 2009 5809 individuals randomly selected Data Observed (powered by Organic): • N_Impression per creative • N_session with at least one visit • N_session with conversion 15 different creatives Benchmark with 4 Models in the Observation Period: 1. No Ads Effect 2. Campaign Ad Effect 3. Creative-Specific Ads Effect 4. Full Model
  17. 17. Study case – Results: Indicator: MAPE (Mean Absolute Percentage Error) Low MAPE  Real behaviour with less error Ad Effect over time Advertising Impression Effect Model fit comparisons
  18. 18. Considerations: PROBLEMs:  Wear-in  “Cold Start”  No Example Reported  Theoretical model  Practical results SUGGESTIONs:  Scheduled ad-exposure  Interaction among website and ad creative
  20. 20. New trends: CORRELATION Video Advertising Mobile Advertising Social Media Marketing
  21. 21. Social media marketing: Scope: Create brand awareness trough social web Viral concept (good or not) eWoM Earned media rather Than paid media COBRAs (Ex. New Converse sneakers to Facebook) Special deals with Tweets or Repost Usage of social networks Interaction with smartphones (QR code) Direct interaction among Companies and users
  22. 22. Mobile advertising: In-App Advertising Sms Advertising Mms Advertising Form of advertising via mobile phones Ubiquity CPI (Cost per install) Smartphone Technologies Battery concerns Incent for Users Interaction with Classic Advertising (Bar code/ QR code) Video: most effective mobile advertising
  23. 23. Video advertising: • Video content in a MPU • Streaming Events • Cut TV Spot before Streaming Felix Baumgartners’ Jump: Big Adventure  Around 10M users watched streaming  Big Visibility for RedBull
  24. 24. Considerations: • Is “AdBlock” a good thing?.
  25. 25. References:  Statistical Techniques for Online Personalized Advertising: A Survey (Maad Shatnawi and Nader Mohamed)  Online Display Advertising: Modeling the Effects of Multiple Creatives and Individual Impression Histories (Michael Braun, Wendy W. Moe)  Video + Tablets: The Mobile Catalyst for E-Commerce (  IAB internet advertising revenue report  Web Information Retrieval (S. Ceri, A. Bozzon, M. Brambilla, E. Della Valle, P. Fraternali, S. Quarteroni)