Click Forensics Traffic Quality Management Tom Cuthbert President

Click Forensics Traffic Quality Management Tom Cuthbert, President & Founder Tom Charvet, Vice President & Founder IEEE March 19, 2009

Agenda • • • About Company Background The Problem of Click Fraud Our Approach Technical Overview

Tom Cuthbert – EDUCATION:  · Attended M. I. T.  · BSEE with Highest Honors, Georgia Tech · MSEE, Southern Methodist University · Ph. D. in Electrical Engineering, S. M. U. – PUBLICATIONS:  · Circuit Design Using Personal Computers · Optimization Using Personal Computers Applications for Electrical Networks  · Broadband Direct-Coupled and Matching RF Networks · Four articles, RF Design, 1994, 1997, 2000 · One article, IEEE Trans. Microwave Thy. Tech. Newsletter · One article, Applied Microwave and Wireless· Paper and presentation, IEEE International Symposium on Circuits and Systems – EXPERTISE · Design &Synthesis of Electrical Filters &Impedance-Matching RF Networks · Amplifier Design by Geometric Model and Mapping · Electronic Circuits in Frequency Ranges from VLF through K Bands · Optimization (Nonlinear Programming) · Numerical Methods and Analysis – JOINED IEEE IN 1958, LIFE MEMBER

About the “Tom’s” • Tom Charvet • Tom Cuthbert • The company • How we got here…

It’s Google’s World…

Click Fraud is like spam… • Spam was up 25% in 2008 • When Mc. Colo went offline in November, the rate dropped 70%, it’s up 156% since • An unprotected user would have received 45, 000 spam emails in 2008 • Virus volume increased 6 X in 2 nd half ’ 08 • Spammers have reinvested and are constantly evolving their efforts

…and spam is getting worse

How Click Fraud Works • Methods – Individual – Click farms – Bot and botnet • Motivations – Competitive – Malicious – Financial

Types of Traffic Quality High Likelihood of Conversion Valid Traffic Low Other Invalid Traffic Click Fraud Invalid Traffic

Invalid Click Detection Click Type Characteristics Click Bots Valid Traffic Programmed for low frequency repeat clicking Click Farms Humans coached to use proxies and clean cookies Competitors Business & Keyword competitors Internal Invalid Traffic Own & Partner clicks Search Bots Valid Bots that don’t properly identify themselves

Q 4 2008 Quarterly Stats Click Fraud Index™

Overall Click Fraud Rate by Quarter © 2009 Click Forensics, Inc.

Content Network Click Fraud Rate % © 2009 Click Forensics, Inc.

Botnet Click Fraud Rate by Quarter © 2009 Click Forensics, Inc.

How We Do What We Do

Who We Help • Buy Side - Advertisers • Sell Side – Search Engines, Publishers, and Ad Networks

The Buy Side

Invalid Traffic Is More Than Just Click Fraud Types of Low Traffic Quality Reason • Low conversion rates Low Quality • CPA too high • Broad match conflict • Outside target service area Invalid Traffic Click Fraud CONFIDENTIAL 18 • Competitor Clicks • Unacceptable type of site (MFA, Adult, etc. ) • Malicious intent • Click Farms • Bots/Botnets

Five Major Threats To ROAS In CPC Campaigns 1. Like spam, bots and botnets are becoming more sophisticated and difficult to catch 2. Over 70% of all sites in Google Adsense and the Yahoo Publisher Network are made for ad sites or parked domains 3. Over 22% of all invalid activity comes from outside the targeted geo area of advertisers 4. Competitor clicks deplete ad spend 5. Brand infringement violations use established brands to send customers to competitor’s sites

Competitors Click On Ads CONFIDENTIAL 20

Clicks Are Coming From China… CONFIDENTIAL 21

Your brand is sending traffic to your competitors CONFIDENTIAL 22

Fully Automated Click Tracking Reconciliation FACTr ™ • FACTr completely automates the submission process of invalid campaign activity evidence • The evidence report is a verifiable third-party audit containing a click-for-click breakdown of invalid activity • Google, Yahoo, Looksmart and Miva are partners in this program with more to be announced soon CONFIDENTIAL 23

Case Study: Click Forensics 6 -Month ROI FACTr q 45% reduction in invalid clicks § Over $200 k monthly savings § Eliminated very expensive, low-quality clicks § FACTr process produced credits for exceptions (>$100 k/mo) q 7% increase in conversion rates

The Sell Side

Traffic Segmentation Best 1000 • Route Traffic Most Effectively • Maximize Earnings Potential Marginal Click Fraud 0 • Traffic to Discard

Reading the Scoreboard § Click. Score - A three digit score that indicates whether a click is valid or invalid and serves as a predictor of conversion. § Visitor. Score - A three digit score produced by each individual visitor. Represents the quality of visitor. § Site. Score™ - A three digit score that represents the traffic quality of an individual publisher or network. Click Forensics 2008

Primary goal is to supply high quality traffic to advertisers from publishers to maximize revenue Ad Network Click Forensics 2008

Click Forensics measures the quality of traffic from publishers and report the results Ad Network Click Forensics 2008

Click Forensics augments existing blocking mechanisms that eliminates the poor quality visitors from viewing ads Ad Network Click Forensics 2008

The result is higher quality impressions and more valuable clicks delivered upstream Premium Standard Discard Ad Network Click Forensics 2008

Reporting Tools • View the Site. Scores™ of the publishers in your network • Monitor the Visitor. Scores™ of your worst offenders • Gain insight into the geographic areas delivering high and low quality traffic • Analyze trends within your traffic • Export data into other applications for added analysis • Setup near real time email alerts to monitor spikes in unwanted traffic • Drill down for publisher and advertiser dispute resolution Click Forensics 2008

Data Flow Diagram Raw click data – transmitted via scp, sftp, or ftps. This can be scheduled near real time. Each click is rated and assigned a Click. Score™ Each visitor is rated and assigned a Visitor. Score™ Ad Network Each publisher is rated and assigned a Site. Score™ The Click Forensics API returns: • Valid and Invalid Clicks and their Click. Scores™ • High threat visitors and their Visitor. Scores™ • Publisher quality index and their Site. Scores™ • Known Offender Lists Click Forensics 2008

Putting It All To Work • Blocking - Invalid traffic can be blocked before it ever reaches an advertiser • Routing - Ad networks can route traffic appropriately to maximize earnings • Intelligent Pricing – Traffic segmentation allows for quality-based pricing models • Publisher/Advertiser Review - Ad Assurance automates the process for detecting malware in ad creative and sites. • Reconciliation - Evidence of invalid traffic is recorded so that ad networks can reconcile payment with both publishers and advertisers

Thanks! Q&A