In the session “Search Engines on Auditing” the speakers discussed how search engines handle auditing concerns.

Richard Zwicky, Founder & CEO, Enquisite, discussed auditing issues with PPC and how they try to figure out what went wrong.
 
Analytics ecosystem:

Collect

Analyze

Report

Optimize

Monetize

Ways ad networks protect you:

Invalid/discounted clicks

Real time behavior analysis

Proactive filtering

Over-time analysis

Mathematical models

Honeynet projects- open proxy

Zwicky said ad networks don’t see what happened after the click left their network, which is why it is important for you to audit your campaign.

You can make ad networks better by using the parameters they have to tighten your campaign and actively control your results. Not everything that goes wrong is fraud.

“When you have an issue, file a claim – it’s the only way you will receive help,” Zwicky said.

Shuman Ghosemajumder, Business Product Manager for Trust & Safety, Google, said there are 2 main incentives for click fraud, attacking advertisers and inflating affiliates.

Numerous methods:

Manual clicking

Click farms

Pay-to-click sites

Click bots

Botnets

Ghosemajumder said Google uses a three-part system for invalid click detection:

1. Filters- automated algorithms filter out invalid clicks in real time, analyzes all clicks, and account for the vast majority of invalid click detection.
 

2. Offline analysis- automated algorithms and manual analysis
 

3. Investigations- rare

“When we mark clicks as invalid, we’re not charging the advertisers for those clicks,” said Ghosemajumder.

Nearly all invalid clicks are detected proactively. Reactively detected invalid clicks are a negligible proportion (less than 0.02%).

“The impact of invalid clicks at Google is minimal.”

Several features help and protect advertisers:
SmartPricing
 

AdWords Auto-Tagging
 

Site Targeting
 

Site exclusion
 

Invalid clicks reports
 

Placement reports
 

IP exclusion
 

AdSense click area change

• Takeaways:
  
• Validate data collection methods
• Define conversions carefully
• Use advanced analytics features to spot trend in data
• Conduct experiments to get more data