Stanford University September-December, 2009
 
Introduction to Computational Advertising
September 25-December 04, 2009 - Stanford University, California


Contents

Course Information

Overview
Computational advertising is a new scientific discipline, at the intersection of information retrieval, machine learning, optimization, and microeconomics. Its central challenge is to find the best ad to present to a user engaged in a given context, such as querying a search engine ("sponsored search"), reading a web page ("content match"), watching a movie, and IM-ing. As such, Computational Advertising provides the foundations for building ad matching platforms that provide the technical infrastructure for the $20 billion industry of online advertising.

In this course we aim to provide an overview of the technology used in building online advertising platforms for various internet advertising formats.

The course is aimed to help industry professionals engage in building systems for online advertising. We also aim to describe the research fronteers of this young discipline to allow graduate students to tackle problems in the area of Computational Advertising. The course will start with an overview of the core Informational Retrieval and Machine Learning techniques that provide the underpinning for the Computational Advertising methods. Then we will describe the marketplace designs for selling, buying and charging for online ads. The lectures on texutal online advertisments will describe both Sponsored Search and Contextual Advertising. Next we will focus on Display Advertising and Behavioral Targeting for banner ads. We intend to have a class project and a panel at the end of the course (content and schedule subject to changes).

Teaching Staff
The best way to reach us is via email at: msande239-aut0910-staff@mailman.stanford.edu
Instructors TA
Meeting Time/Location
Fri 10 am-12:50 pm, 380-380Y (Math Corner)

Grading
There will be 4 -5 homeworks and one major term-project involving real-world web advertising. The homeworks will account for 50% of the grade and the project will account for the remaining 50% of the grade. The term-project should be done in teams of up to 3 people. However, homeworks must be done individually. It is an honor code violation to collaborate in any form on homeworks.
Course Schedule (tentative)

  • 09/25 Overview of Computational Advertising
  • 10/02 Marketplace & Economics Considerations
  • 10/09 Textual Advertising & Sponsored Search
  • 10/16 Sponsored Search (part 2)
  • 10/23 Search Methods for Textual Ad Selection
  • 10/30 Content Match
  • 11/06 Collaborative Filtering
  • 11/13 Graphical Advertising: Guaranteed Delivery and Behavioral Targeting
  • 11/20 Other Contexts: Mobile, Social Networks, and Rich Media
  • 12/04 Project Presentations
Lecture Handouts

  • Lecture 1:
  • Lecture 2:
  • Lecture 3:
  • Lecture 4:
  • Lecture 5:
  • Lecture 6:
  • Lecture 7: (Guest Lecture by Prof. Ashish Goel, scribed by Rio Goodman)
  • Lecture 8:
Readings

Assignments

Policy
  • Assignments must be done individually. It is an honor code violation to collaborate in any form on assignments.
  • Recognizing that students may face unusual circumstances and require some flexibility in the course of the quarter, each student will have a total of three free late (calendar) days to use as s/he sees fit. Once these late days are exhausted, any homework turned in late will be penalized 50% per late day.
Assignments
  • Assignment 1 is out. It is due back Monday, November 2nd by 5:00 pm in Terman 393.
  • Assignment 2 is out. It is due back Friday, November 13th in class.
Project

Advertising Project Description
The project for this class will involve hands-on experience with real-world web advertising. Students may (and are encouraged to) work in teams of up to 3. Each team will find a local business (e.g. restaurants, doctors, laywyers, etc.) and work with them to create an effective online advertising campaign for them on Yahoo!. Alternatively, if a team is passionate about a cause (e.g. Autism), they can create an advertising campaign for that cause. The broad steps involved in the project are as follows:
  • Each team finds a firm (or a cause) to create a campaign for. The firm does not have to be a large well-known public company. It can be a local restaurant, a friend's startup, a grocery store, etc. They should preferably be local so that it is easy for the team to interact with them on a regular basis.
  • The team creates a website or (if a website exists) gain access to the website to instrument it for metrics collection (e.g. "number of page views", "number of unique visitors", "number of online orders", etc.). Note that the effectiveness of your advertising campaign (and therefore your project grade) can only be demonstrated using metrics, so it is important to pick the right metrics. A few suggestions:
    • Find which metrics are most relevant for your firm: for a pizza shop the number of online orders per day may be a good metric, but that clearly does not work for a lawyer. You should be able to justify why you picked the metrics you did.
    • Find which metrics are helpful in demonstrating effectiveness. For example, if a store gets 5 online orders a month, and after your campaign they start getting 10 online orders a month, you cannot claim a 100% increase in orders because of your campaign. Keep in mind the notion of statistical significance when choosing your metrics.
    • Lastly, as with most things, the more metrics you collect the better your report will read: more metrics will give you more insight into what works and does not work, and (probably) why. It will help you craft more realistic, meaningful recommendations for the firm.
  • Setup an account on Yahoo. More information on this to follow. Basically, you will have access to a small amount of money (something like $200) to run your campaign. You can create and schedule campaigns in your Yahoo! account. It will also give you access to some metris that Yahoo! collects for advertisers (e.g. number of impressions, number of clicks, etc.).
  • Come up with an advertising strategy in conjunction with the firm. This involves answering the following kinds of questions (this not a comprehensive list, just a small representative set):
    • What demographics and geography to target?
    • What search terms and web properties to advertise on?
    • How to split the advertising dollars between search and display ads (and potentially other forms of advertising)?
    • What should the banner ads and creatives look like (messaging)?
  • Execute on the strategy agreed upon above; run the campaign. The campaign should run for at least 5 weeks. The longer it runs, the more data you collect. We encourage you to do a mid-way analysis and tweak your strategy and/or your bids, creatives, etc. if required. It is better than letting the campaign run unsupervised for 5 weeks and to find out at the end that it had no impact.
  • Collect metrics and write up a report. The report should focus on the following:
    • What business did you work with? What was their current market position and future growth trajectory?
    • What advertising strategy did you come up with and how that ties into the growth plans of the business?
    • What mathematical models did you use to arrive at your allocations, bids, etc.? Why? (e.g. Multi-armed bandits, greedy strategies, etc.).
    • What metrics did you measure? How? Why?
    • What recommendation do you have for the firm based on your understanding of their business, the advertising campaign you ran and the metrics you measured?
Timeline
The expected output and timeline for the project is as follows:
  • 10/09: Submit a team name, list of team members (via email to the staff mailing list).
  • 10/16: Submit the name of the firm you are working with (via email to the staff mailing list).
  • 10/30: Submit a 2-3 page write up on your advertising strategy with some justification for it. Due in class. (25%)
  • 12/04: Project presentation (15 mins per team). (25%)
  • 12/11: Final project report (max. 15 pages) due. Hardcopy due in Terman 393 by noon. (50%)
Resources
Here are a few links/docs/tutorials that you might find helpful: Algorithmic Project Description
The details of the Algorithmic project can be found here (updated on 11/14/2009).
Short Bios

    Andrei Broder is a Fellow and Vice President for Computational Advertising in Yahoo! Research. He also serves as Chief Scientist of Yahoo’s Advertising Technology Group. Previously he was an IBM Distinguished Engineer and the CTO of the Institute for Search and Text Analysis in IBM Research. From 1999 until 2002 he was Vice President for Research and Chief Scientist at the AltaVista Company. He graduated Summa cum Laude from the Technion, and obtained his M.Sc. and Ph.D. in Computer Science at Stanford University. His current research interests are centered on computational advertising, web search, context-driven information supply, and randomized algorithms.

    Broder is co-winner of the Best Paper award at WWW6 (for his work on duplicate elimination of web pages) and at WWW9 (for his work on mapping the web). He has authored more than ninety papers and was awarded twenty-eight patents. He is an ACM Fellow, an IEEE fellow, and past chair of the IEEE Technical Committee on Mathematical Foundations of Computing.
    Vanja Josifovski is Principal Research Scientist and the Lead of the Textual Advertising Research group at Yahoo! Research. He joined Yahoo! Research in late 2005 and has since spent most of his time designing and building Yahoo!'s next generation online advertising platforms. His research interest include ad selection for sponsored search, content match and graphical advertsing; search engines adaptation for ad selection; data mining and information retrieval techniques for improving ad quality; and click and query log data analysis. Previously, Vanja was a Research Staff Member at the IBM Almaden Research Center working on several projects in database runtime and optimization, federated databases, and enterprise search.

    Vanja has published over 50 peer reviewed publications and has authored over 40 patent applications. He has served on the program committees of WWW, SIGIR, ICDE, VLDB, SIGKDD and other major conferences in the database, information retrieval and search areas. He holds a MSc degree from University of Florida and a PhD degree from Linkopings University in Sweden.

Related courses