MS&E 235, Internet Commerce. Winter 2007-08.
Instructor: Ashish Goel
TuTh 2:15-3:30; Gates B03. This class is on SCPD, channel E4.
There will be a few optional televised discussion sections during the quarter: Fr 2:15-3:05; Skilling 191.
If you are auditing the class, please subscribe to msande235-win0708-guests by going to http://mailman.stanford.edu.
Instructor: ashishg at stanford.edu, Cell phone: 650 814 1478, Office: Terman 311.
Course assistant: Christina Aperjis. Email: caperjis at stanford.edu, Office: Terman 391.
Unless you need to very specifically contact the CA or the instruction, send email to msande235-win0708-staff @ lists.stanford.edu
Office hours: Ashish: Th 11-12. Christina: Mon 5-6.
Text-book: None; we will make lecture notes available a few days after each lecture.
Class description: The technology, mathematics, and economics of Internet commerce. Topics include: models of Internet commerce; online advertising; product recommendation systems and personalized marketing; pricing and delivery of digital media; web tools; piracy, copyright, and peer-to-peer networks; rating and reviewing of online businesses; and co-evolution of Internet technology and commerce. Hands-on exercises; group project.
Grading: Project 35%, Homeworks 25%, Midterm 20%, Final 25%. If you score more than 100, you get an A+.
Project groups: Due Jan 17th.
Midterm: Feb 12th, in class (except for SCPD students, of course, who can take it at their work place).
Interim project report: Due any time during the quarter. The sooner the better if you want feedback.
Project presentations: Mar 11th and 13th. At least one member of every group must be able to come to campus.
Final project report: Due Mar 16th.
Final exam: Thu, March 20th, 7-9pm (from the registrarís exam schedule)
Topics (subject to change; some of these will be covered informally by guest speakers).
1) Advertising and marketing
a) Push vs pull models; cost-per-click vs cost-per-action vs cost-per-impression
b) Technical topic: auctions (advertising auctions, truthful auctions, eBay auctions)
c) Technical topic: decision making under uncertainty and the multi armed bandit problem
d) Technical topic: long tails
2) Reputation, prediction, and recommendation
a) Reputation systems in practice
b) Technical topic: Eignevalue methods, eg. PageRank and HITS
c) Technical topic: Prediction markets
d) Technical topic: The arbitrage primitive
e) Technical topic: Singular value decomposition and collaborative filtering
3) Social networks
a) A survey of social networks
b) Social robustness; privacy vs functionality; anonymized consensus
c) Technical topic: the social cost of cheap pseudonyms
d) Technical topic: spread of influence in social networks; function follows form; and viral marketing
e) Monetizing social networks
4) A brief description of the client-server model
a) Email and the web
b) DNS and Akamai
5) Peer-to-peer systems
a) Piracy vs copyright
b) Technical topic: the architectures of KaZaa, BitTorrent, and Skype
c) Technical topic: Pricing of digital goods
Lecture Notes (provided by Shrikrishna Shrin)
Problem Set 1 (due Jan 31)
Project descriptions: The projects must be done in groups of three or four. Each group turns in an interim and a final project report and presents the project in class. For the Google Online Marketing Challenge the lengths of the reports are determined by Google to 2 and 10 pages. For all other projects, the interim report should be around 5-7 pages and contain a brief description of your approach to the project, a summary of your progress, and a detailed outline of the final project report. The final project report has no page restrictions, but around 15-20 pages is probably sufficient. All project reports will be on the class web page and hence in the public domain. But you must maintain confidentiality of any external data that you are provided and omit that data from the public part of your report.
Suggested projects (you are free to devise your own):
1) Google Online Marketing Challenge (teams must be of size 4)
Student groups will receive US$200 of free online advertising and then work with local businesses to devise effective online marketing campaigns. They will outline a strategy, run their campaign for 21 consecutive days, assess their results and provide the business with recommendations to further develop their online marketing. Each student group will choose a business that (a) has a website, (b) is not already advertising using Google AdWords, and (c) agrees to have a campaign devised and run for the purposes of the Competition. The instructor has a couple of businesses in mind and you are of course encouraged to find your own. You must be ready to go on Feb 10th so that you can finish in time.
Projects will be evaluated by Google on three criteria, two short written reports and the most effective Google AdWords campaign for their business. We will use the same criteria. At least one week prior to the campaign, groups submit a two-page Pre-Campaign Strategy, containing a client overview and proposed AdWords campaign strategy on criteria such as keywords, time of day and location. Teams submit a ten-page Post-Campaign Summary no later than three weeks from the campaignís end.
Details are at http://www.google.com/onlinechallenge/. The following links are useful:
The Student Guide - http://www.google.com/onlinechallenge/student_guide.pdf
The Guide to Running Your AdWords Account - http://www.google.com/onlinechallenge/adwords_guide.pdf
The Guide to Marketing and Advertising Using Google - http://www.google.com/events/business_educators/files/MarketingAndAdvertisingUsingGoogle.pdf
2) Build an Internet-based business (same as last year: here is an article about this project from the Stanford Daily)
Each group must invest no more than $225 (suggested but not mandatory breakdown: $200 in advertising, $25 in operational costs) to build an Internet-based business. The projects will be evaluated on (a) Strategy and execution [20 pts] (b) Whether the group generated revenues  (c) How much revenue the group generated  (d) Whether the group turned a profit , and (e) How much profit . We would suggest investing $100 each towards advertising on Yahoo and Google (or MSN or any other combination of two venues that you prefer) to attract customers into an Amazon associates site that you create for selling products of your choosing, but feel free to use any other legal, socially responsible, and ethical business model. You can reinvest your revenues into advertising/operations, but if your net out of pocket expenses hit $225, you MUST stop. Submit an interim progress reports and a detailed final report including accounting and receipts. Donít panic if you go bankrupt Ė as you can see from the relative weights, it is more important to have a convincing strategy and good execution than to be too narrowly focused on turning a profit, which is hard in so short a time. Try to invest gradually so that if one business idea is not going anywhere, you can try another to at least generate some revenue (for example, a blog that attracts ad-sense revenue). Start early. Discuss your project with us early and often. And remember, content rules! You should have no direct inventory. You have to submit an expense summary and an honor code statement by all group members that you did not exceed the prescribed maximum.
3) Write a Facebook plugin to implement a new social primitive
The instructor has some ideas, which he will be happy to share. But you must be very comfortable with web 2.0 coding.
4) Advertising banner optimization (Via adchemy.com, an advertising start-up in the valley)
Adchemy provides the students with data about the performance of some of their advertising banners for a given advertising campaign. The students would be expected to analyze the data and provide recommendations as to which banners they would like to show and in what proportion during an evaluation period. During the 1-2 week evaluation period, Adchemy will implement the studentsí recommendations on a subset of their impressions. It may be possible to make this an iterative process so that there are two or three evaluation periods (and hence opportunities for the students to adjust their strategy). The studentsí performance could be compared against that of one of our automatic banner allocators.
5) Reputation System for the GlobalGiving foundation
GlobalGiving is a marketplace that connects donors to projects via their website, as well as through custom giving services--helping corporations, foundations, affinity groups, and other institutions with their specific giving needs. You can find out more about the foundation at http://www.globalgiving.com/.
The goal of this project is to design a reputation system for project leaders by incorporating data such as page hits, Google rankings, whether a given report from a project leader actually led to repeat donations, exits from the project page, etc. The President of the global giving foundation will be available to consult on this project.
The group should consist of at least two people comfortable with coding and at least two people with basic algorithms, economics or optimization knowledge.
6) Inspect-Utilize Model
Run experiments to report the accuracy of the Inspect-Utilize model, which is described in http://www.stanford.edu/~rajatb/robustranking.pdf . This is an open-ended project, which may have a high impact. There is a possibility of getting data from Microsoft.
The group should consist of at least one person comfortable with coding.
7) Robust ranking
Run an experiment to simulate the incentive based reputation system in the paper http://www.stanford.edu/~rajatb/robustranking.pdf . This can be in the form of a web-site with a game that you devise for the class to play.
Check out blinds.com, which is an example of a business which has succeeded in monetizing the long tail.
Watch this clip (http://www.thebusinessmakers.com/2006/09/02/episode-65-jay-steinfeld)
and read this article (http://www.internetretailer.com/dailyNews.asp?id=16816) for more details.
An article by Chris Anderson explaining long tails:
A recent Internet advertising prediction:
An old Internet advertising prediction:
Internet advertising and the generalized second price auction: Selling billions of dollars worth of keywords. B. Edelman, M. Ostrovsky, and M. Schwarz.
Position auctions. H. Varian.
Truthful auctions for pricing search keywords. G. Aggarwal, A. Goel, and R. Motwani.
Google press release: http://www.google.com/intl/en/press/annc/online_marketing_challenge.html
Algorithms and incentives for robust ranking. . R. Bhattacharjee and A. Goel.
The HITS paper:
Authoritative sources in a hyperlinked environment. John Kleinberg
Auren Hoffman (CEO of Rapleaf)
Title: Reputation, Portability and Privacy
When: 2:15PM, Tuesday February 5, 2008
Where: Gates B03
The Future of Online Advertising
When: 2:15PM, Thursday February 7, 2008
Where: Gates B03
Rob Goldman, formerly General Manager, shopping.com
Vanja Josifovski, Principal Research Scientist at Yahoo! Research.
Anil Kamath, CTO, Efficient Frontier
Rob Luenberger, Chief Scientist, advertising.com
Murthy Nukala, CEO, Adchemy Inc