Auditing Paid Listings and Click Fraud Issues

It’s time for this afternoon’s Auditing Paid Listings and Click Fraud Issues panel moderated by Jeffrey Rohrs (Optiem, LLC), Shuman Ghosemajumder (Google), Tom Cuthbert (Click Forensics), Reggie Davis (Yahoo) and John Marshall (ClickTracks.com).

Quite a bit of star power on this panel. I’m curious to see whether it gets a little heated or not. We’ll have to see. Let’s hope the chairs are bolted down.

Clearly, click fraud is a major problem facing advertisers. How do you know if the clicks you were charged for last month were legitimate or if something else was going on? Are you reviewing your clicks? The hope is that this panel will show us how to spot abnormalities and follow-up with search engines if you suspect a competitor or someone else.

Up first is John Marshall to talk about distinguishing click fraud from poorly performing ads.

It turns out distinguishing a badly designed ad from click fraud is actually quite difficult because they look very similar. John gives some tips on how to distinguish them and some tips for spotting weak ads.

Case Study: John noticed he got lots of traffic on one particular ad without a corresponding increase in sales. It looked suspicious but is it click fraud? Maybe it was just that an ad appeared on a new affiliate site or an affiliate generated low-quality clicks.

He started going through a thought experiment, creating a hypothetical situation in his head. What if on deep analysis he saw that the clicks weren’t covering to sales and that lots of clicks are coming from India. Then it’s definitely click fraud, right? Not necessarily.

It could be that his ad got picked up by a publication that was particularly targeted toward readership based in India (riiiight.). Maybe when searchers clicked through the ad they found it was too expensive to ship the product to in India they went away immediately.

So how do you know what the true story was?

When he looked even more carefully he saw the clicks came from different IP addresses, most of the clicks were from the US, they had different user agents, they executed JavaScript and took cookies and loaded images. On the other side, he also found that a very high proportion were single-page visits and that the referring site looked very suspicious. Eventually, he concluded it was probably click fraud. He submitted the data and got his refund.

The moral is that distinguishing between click fraud and badly performing ad is tough. The data looks very similar and requires human judgment to examine what’s going on. It also requires knowledge of your specific Web site and visitor demographics.

Use statistical analysis to look at multiple data points for your clicks. Don’t rely on ROI. The computer shows campaigns which are worse than average in anyway. They could have a low conversion rate, have a lot of single-page visits, have a lot of clicks from China, etc.

If you go through this process you’ll also find poorly performing ads, which you can then fix.

Next up is Shuman Ghosemajumder from Google to talk about auditing paid listings and click fraud issues.

Click fraud comes from two main places – attacking advertisers and inflating affiliates – and it can happen in a variety of ways. It could be manual clicking, click farms, pay-to-click sites, click bots or botnets.

How about some definitions? Shuman gives some insight on commonly heard terms:

Invalid clicks: Clicks Google decides not to charge for.
Click fraud: Something which is very difficult to define because it’s a question of intent. You’re trying to read people’s mind to see why they clicked the ad.
Click fraud attempts: Clicks on AdWords campaigns made with malicious or fraudulent intent.

Shuman details the three step process for invalid click detection. The two initial stages are proactive detection and the third stage is reactive detection. Google wants all clicks to be caught in the first two stages so users are never charged. Also, they want most of the clicks to be caught in the first stage.

He says you want to make sure that your filters are sensitive enough that you’re really catching malicious activity.

Click fraud estimates vary slightly. In 2006 it was said to be 12% of clicks and in 2007, it is estimated click fraud cost advertisers $666 million. (See, click fraud is evil!)

The reality is that there is a significant number of clicks that are filtered (<10 percent). Nearly all invalid clicks are detected pro-actively. Reactively detected invalid clicks are a negligible proportion. Google wants to see more over-reporting than under-reporting. The level of reporting today is at a record high.

Where do fictitious clicks come from? Take a look at the scenario below:

1. A user searches on "flowers"
2. That user clicks on an ad and is taken to an advertiser’s landing page
3. Visits to landing page are tracking in the advertiser’s log
4. Advertisers log shows new hit
5. User hits the back button
6. Advertiser’s log records a reload of the page

Right there an advertiser sees an invalid click, but that’s not the case.

The only way to resolve this is to use redirects or to use AdWords auto tagging which appends a unique ID to the user befriending the back button.

Google has several unique features like AdWords Auto-Tagging, invalid clicks reporting, and enhanced reporting (coming Q2 2007)

Shuman says Google is trying to become more transparent but they don’t want to aid fraudsters from in attacking users. He blames CSI for fraudsters becoming more savvy. Weirdest. Analogy. Ever.

Tom Cuthbert is up next to talk about the progress on the click fraud front.

Tom says he’s hearing from advertisers that things are improving. He’s here to give some good news. Yay for good news!

Google revealed plans for an IP exclusion list and announced enhanced reporting format. Yahoo named Reggie David as VP level exec to combat click fraud. They’ve also announced domain level blocking. None of this, however, eliminates the need for third parties. (Or the need for you to study your own clicks.)

Other signs of industry progress include that industry progress, awareness of click fraud at an all time high, IAB Click Measurement Working Group, Click Quality Council meeting monthly, and the Enhancing Click Fraud Network launched.

Tom says his company will soon be releasing the click fraud numbers for quarter one.

What’s next? Working to take advantage of site exclusion opportunities by pooling community data and using it to make better decision.

Up last is Reggie Davis.

The mission of Yahoo’s click fraud team is to create the world’s highest quality search and display advertising network. It’s about additional quality initiatives, new features and functionality. It’s about greater visibility and control. To have more dialogue with advertisers and publishers.

Reggie says in the past an average of 12-15 percent of clicks were not billed because they were thought to be fraudulent. Yahoo’s committed to providing good quality traffic.

Reggie speeds through explaining Yahoo’s clickthrough protection system. I think he did it on purpose so that I couldn’t blog al of it. The first three steps are as follows:

1. Thousands of filters assess all attributes of each and every click.
2. results include black-list/white-lists
3. Behavioral pattern matching

To provide a better experience for users, Yahoo is aiming for an improved publisher assessment, deeper advertiser case studies, increased advertiser’s adoption of conversion tracking tools and improvements to matching technologies

As a result of this, Yahoo has seen a significant decrease in the number of claims being sent in.

Reggie briefly discusses Panama and says for 2007 Yahoo is working on quality based pricing, domain blocking, Advertisers Marketplace Quality Council, additional investigation detail, and to continue IAB efforts.

[Whose idea was it to space session 15 minutes apart? Why are people trying to kill me? (Mine and yes. I also told them to turn on the AC.–Susan)]

Lisa Barone is a writer, content marketer & VP of strategy at Overit Media. She's also a very active Twitterer, much to the dismay of the rest of the world.

See Lisa's author page for links to connect on social media.