Yahoo admits no one can detect every fraudulent click
December 31, 2005
Google and Yahoo owe a lot of their growth to the pay per click phenomena of advertising. The majority of Google’s total income (estimates are around 98% of their total income) is from search based advertising.
Pay per click (PPC for short) is a simple system where advertisers bid on a keyword, and then when someone searches that keyword (often called a user query) on a search engine, the advertisers ad is displayed along with the usual search results. If a user clicks on the advertiser’s ad, then the advertiser is charged based on how much they are willing to spend per click.
Advertisers only pay when a searcher clicks on an ad, hence, pay per click.
Click fraud has been unofficially labeled as clicks on pay per click ads originating from illegitimate sources. This can occur for a large variety of reasons from advertisers competition to publishers wishing to line their pockets.
Click fraud has been widely discussed over the past year. The most recent mainstream press was from a Wired article that received some excellent discussion in a WMW thread.
It’s pretty common for search engines to be tight lipped about click fraud, detecting clicks, and what users can do about it.
While click fraud is an issue, so are traffic spikes that appear to be fraud – but aren’t. A syndicated ad onto the front page of CNN or ESPN can receive a huge traffic spike, unfortunately, often these additional clicks don’t always turn into additional sales.
News can also trigger these spikes. In fact, on the same page that Yahoo admits the potential for click fraud, they also acknowledge traffic spikes:
Some keyword markets have a lot of volatility in their quantity of clicks, sometimes leading to traffic spikes that might be interpreted as fraudulent activity. Imagine how many added clicks bidders for the term “Janet Jackson” received after the infamous Super Bowl incident. If you see an unexplained surge in traffic, it can often be related to a news story. Our Click-Through Protection team is trained to identify these click spikes and their possible legitimate sources.
This is a very accurate statement by Yahoo. On several occasions there have been large traffic spikes for niche keywords that appear to be click fraud based on historical search history. However, investigation into the traffic often shows why the spike happened and that they were legitimate visitors.
This is also one reason that it’s very important to stay on top of the news and your keywords. If the advertisers for ‘Janet Jackson’ had anticipated the additional traffic and paused those keywords, they wouldn’t have received clicks from those searching for additional info on the incident.
However, this isn’t to say one should always pause keywords when a story breaks. There are times that a news story can propel sales as well as additional traffic. One must determine user intent for the search on those keywords to make an appropriate decision.
Back to the click fraud issue. While Yahoo has made a few public statements (and many in private at conferences), Google has stayed very quiet about how they track fraudulent clicks. Yahoo has mentioned a few points, however, this is one of the more specific published pieces of information to date:
Yahoo! Search Marketing’s click protection systems are in operation 24 hours a day, monitoring each click and filtering out those that are questionable or clearly unqualified. To do this, we track search-and-click patterns across more than 50 data points—including IP address, users’ session information and browser information, and pattern recognition—to help detect invalid clicks.
50 data points is quite a lot . I’ve hard Google engineers discuss in private that they do snapshots of browser versions, screen resolution, if cookies and java are enabled, etc – and put together information about a system to help determine if someone moving through anonymous proxies is in fact the same system, hence, it could be related to click fraud.
Onto the juicy part, Yahoo admits no one can detect every fraudulent click.
From the Yahoo Search Marketing Handbook:
We acknowledge that no system, no matter how sophisticated, can detect every invalid click. While bid prices will ultimately reflect the true quality of the traffic, Yahoo! Search Marketing continues to spend substantial time and money to make sure that our technology leads the industry. To further our efforts, we also rely on, and encourage, our advertisers to bring possible invalid clicks to our attention.
This is a very interesting acknowledgment. It’s also very true. Click fraud done randomly with sophisticated technology is nearly impossible (if not completely impossible) to detect.
While the click fraud department probably receives both many cases of fraudulent and legitimate clicks, there have been many stories about Yahoo just giving refunds when the traffic doesn’t ‘appear right’. This could be a case where a word had a sudden spike in traffic for no reason. Of course, Yahoo doesn’t state if they found illegitimate clicks, but they go ahead and refund the money anyway as the traffic just doesn’t adhere to a normal pattern.
I do believe the media has made more out of click fraud that actually exists. Often these stories are spurred on by ‘click fraud detection companies’. Since these companies receive more clients the more they can convince the public click fraud is a huge issue, of course they like the spin that click fraud makes up a large percentage of all visitors.
In any case, it is interesting to note what Yahoo has published in this new handbook (which is an excellent read if you use Yahoo Search Marketing PPC).
The general consensus of click fraud in 2006 looks like this:
- Click fraud will continue to be an issue.
- The search engines will continue to develop better technology in fighting it.
- Frauders will continue to develop better systems for perpetuating click fraud.
- Advertisers sit in the middle, monitor their clicks, and suffer the consequences of fraud.
Many of the media stories about click fraud leave the impression that advertisers are very helpless in the issue. However, many marketers have stated that they adjust bids by conversion and ROI (return on investment) so click fraud is already taken into account based on the various markets. In these instances, the total dollars available to the search engines don’t rise. If click fraud were less prevalent, these advertisers would just bid higher as their conversion rates would be even higher.
So who does click fraud really hurt? Actually, everyone. Due to the media stories, less advertisers try search marketing. The less marketers, the less search engines make. And don’t forget the searchers – they do like (and click) on these ads (the proof is in Google’s net worth and advertisers ROI).
It’s good to see Yahoo acknowledge click fraud. Only by search engines and advertisers working together will it be stopped. One group alone doesn’t have enough data to see all the search and click patterns. Without all of the data patterns, there is no way to tell if clicks are legitimate or fraudulent. If everyone starts to work together in data collection and sharing, then click fraud will start to be a battle that the PPC engines can win.
Google AdWords Guides from Google
December 31, 2005
I was poking around some of the new Google FAQs, and found a set of guides published in 2003 by Google.
One of the guides I had seen before, The Maximum Effect, which was excellent for it’s time.
The information in them is worth perusing, however, they are 2 years old.
• Managing your AdWords Account [pdf] updated Sept-03
• Keywords, Queries and Leads glossary of search terminology [pdf] updated June-03
• Conversion Tracking Guide [pdf] updated Oct-03
• The Maximum Effect [pdf] updated Dec-03
It looks like Google put together several in a short time span (updates September, October, December of 03) and then just stopped.
For those who like to read PPC PDFs, it is worth noting that Yahoo has re-released their workbook (PDF File) with the new Yahoo Search Marketing Brand.
The original Overture workbook was one of the best ones ever created by a PPC engine. The new one is fairly similar, however, it’s been updated with some newer information, and is definitely worth the read.
Conversion Analysis Needed Among Top Destinations Such as BestBuy
December 20, 2005
CNN is reporting that Amazon, Wal-Mart, and Apple are doing great this season, which isn’t surprising – their websites are easy to use.
This paragraph caught my eye as I’m a BestBuy user:
Meanwhile, customers leaving the BestBuy.com site are not heading to shipping sites in significant numbers, downstream traffic data show. This could be a measure of online sales weakness, mechanical differences in where Best Buy’s site directs customers for shipments, or even just the preference of its customers to buy in stores rather than online.
“BestBuy.com shows a stronger indication that it is being used as a research tool” to check prices before visiting the company’s stores, Tancer said. Traffic from BestBuy is especially heavy to in-store promotional advertisements.
Anyone skilled in website analysis could tell BestBuy its not just about the traffic, it’s about usability.
Consider the purchases I made just a couple days ago.
I went to BestBuy to ship out two gift cards. The gift cards were easy enough to find. However, the cart page is cluttered and difficult to navigate – chance of lower conversion – not making it clean and easy to follow the checkout process.
When arriving at the checkout page, it asks me for my BestBuy ID. I didn’t have one, so thought this would be a great time to make one. However, there was no way to create one at this step (at least that I could easily see). At this point in time, let me also just create a BestBuy ID. It not only gives BestBuy my info, it also keeps me vested in the checkout process.
After realizing I can’t make an ID, the next page that was presented is probably my most hated aspect of forms – too many fields. If they know my zip code, why can’t they fill in my city and state for me? Making someone fill out additional information is another step in losing a customer.
Next, the items were going to different addresses, and I couldn’t ship them independently (like Amazon, which lets me ship every package to a different address). So, I had to go back into the shopping cart, delete one of the gift cards, and then re-navigate back to this screen. More conversion loss in those additional steps. It also meant I was going to have to check out twice. Many people wouldn’t want to go through these steps – even more chance at conversion loss.
The credit card field doesn’t allow for spaces or dashes. It’s a very easy database change to have the system remove non standard characters and not have this be an error. If someone is trying to give you their credit card – LET THEM. Don’t have silly database errors stand between you and a paying customer.
Next, it asked me again to create an account. This should have been done much earlier in the process which was talked about above. I did make an id because I wanted to turn around and have another card shipped immediately after this one. However, this is when it turned ugly. I made an ID, and the system errored out. I tried to make another ID, and it just refreshed the page and didn’t take me anywhere. HUGE conversion loss. I had to delete my cookies, and then re-navigate to the page to checkout. This was two days ago, and I’m still not sure if I have a BestBuy ID. I’m assuming not as I didn’t receive a ‘Welcome to BestBuy email’, but I’m not positive.
From arriving at the Home page, adding a single gift card to my cart, navigating through the checkout process (without creating an account) took 12 page views. This number could be cut down by a few page views. Every time a user must click the ‘continue’ button (this is for any website), there is a chance at conversion loss. The easier to navigate, fill out forms, and check out – the higher conversion percentages will be. Of course, websites do have a minimum amount of information they must collect, but there are many ways of collecting the same information.
When considering conversion analysis, increasing conversions, determining why your site isn’t selling, don’t just make assumptions. Use focus studies, have people try to buy from you. Listen to the feedback. Make improvements.
There were some good things on the site as well. On the page which asked for credit card information was a prominent link to the privacy policy. The checkout pages did have the Verisign seal on it (although, it could have been placed better). The shipping date was estimated, and when I changed my shipping options, the date changed as well. Some simple things that users look for (shipping price and dates are very necessary this time of year).
Some simple changes suggested in this article should increase BestBuy’s conversion percentage. At even a 0.5-1% increase (often increases are higher than this, however, even at these low levels of increases) BestBuy would see a significant growth in online sales. If one considers that conversion percentage is the only statistic used in every marketing analysis reports, it starts to become clear just how important a statistic it is.
If a site had 1 million visitors a month, a 1% increase in conversions is 10,000 more sales. I would guess BestBuy’s average checkout is at least $50 (and could be much higher with plasma TVs, gaming consoles, multiple DVD purchases, etc). 10,000 new sales at $50 is $500,000! That’s half a million in revenue.
When one considers how much top level conversion analysts cost. How much a highly targeted PPC account interacts with conversion percentage. Their is no reason to make this number even higher.
BestBuy would see almost an immediate return with a few changes.
Of course, information like this goes far beyond BestBuy – it applies to every eCommerce site on the web.
If you took an internet user, any one of millions (except for the select few who work for your company) and asked them to buy something they want from your website. What would they find:
- How hard is that process?
- How easy is it to find information?
- How easy is it to checkout?
- Do error pages blame the user or the website?
- Are forms intuitive and helpful?
- Is the cart easy to find?
- Is it easy to get lost within the shopping cart?
- Are you measuring shopping cart abandonment?
- Are you measuring the actual page where abandonment happened?
- Are you measuring conversions by organic traffic, type-in traffic (i.e. direct navigation), PPC traffic, and by various banner based traffic?
- Many, Many more questions to ask.
Then think about all the scenarios that could be applied:
- A grandmother looking to buy her grandson an iPod accessory. She doesn’t use the web much except for email. She still doesn’t understand most of the web lingo, and has no desire to learn. She just wants to find a fun gift for her grandson. Are there reviews, information telling her what is hot, some sort of suggestion for her to follow? Her main concern is privacy, credit card usage on the web, and shipping. S
- A mother looking to buy her son an XBox game. She doesn’t know the difference between FPS, RPG, etc. She wants to know what other boys her sons age are playing (and what’s appropriate). Many would tell her to buy a gift certificate, but with the correct information, she wants to try and put a gift under the tree. Her main concern is finding the gift. Her second is the exchange policy. She wants to know that if she chooses wrong that her son can exchange it for something he wants. This is the best of both worlds – a gift under the tree, and exchange if she gets it wrong so everyone is happy. Are exchange policies clearly stated?
- A new father who is hip to the internet lingo, shopping online, etc. However, he now has a 6 year old daughter and wants to see educational games. Being a MORPH gamer, he understands games, but has no idea what is appropriate for a 6 year old. Are ratings clearly marked? Is there a gift suggestion for this age group? etc.
- The same new father wants a gift for himself as well – a 50 inch Samsung HDTV. He knows what he wants. He’s already done the research. What he wants to know is if he can get the great BestBuy warranty online. He also wants BestBuy to ship it AND set it up. Does the website allow for him to find out delivery and setup options? Does it make suggestions for an audio system that compliments that TV set? If he’s about to drop $3k on a TV, a $1k audio system might be a worthwhile investment for the entire home theater experience. He wants BestBuy’s expert suggestion as well as have the entire audio system installed as well. Is the information there?
Many more scenarios could be devised, the real questions are:
- Is there information there?
- Can it be found?
- Is there an 800 number for clarification?
It’s not always the advertising. It’s not always the website. It is about the user experience and expectations.
Meet the expectations. Guide the users through your website. Remove barriers between visitor and buyer. Watch your sales increase.
Yahoo to Change Look of Search Results
December 16, 2005
In a move that has been anticipated for quite some time, Yahoo announced yesterday that they will change the look of their paid results (those from Yahoo Search Marketing) beginning mid-January.
A new look is coming to the Yahoo! search results pages that
will translate into more clicks for your listings. On January
18th, Yahoo! will debut a streamlined design that will make
the search results displayed on Yahoo! even easier for
consumers to read. Our research has shown that by improving
the search experience in this way, advertisers can generally
expect to see an increase in clicks, while maintaining their
conversion rates.How this change impacts your listings:
* Yahoo! will display shorter descriptions for Sponsored
Search listings
* You don’t have to make any changes to your listings; they’ll
be automatically shortened for you when displayed on Yahoo!
* If you’d like to optimize your listings for Yahoo!, begin
your description with one short sentence that includes your
keyword and focuses on your most important information in
the first 70 characters
* Over time, we will fine tune the exact character count that
we believe works best for advertisers and search users
* Most of our partners, including MSN, CNN, ESPN and Infospace,
will still display longer descriptions for your Sponsored
Search listings, though the exact length may vary from
partner to partnerYahoo! is taking this step to improve the search experience
for its users. By continuously focusing on delivering highly
relevant search results in a user-friendly format, Yahoo!
also gives you the best possible platform for reaching customers
interested in what your business provides.
At present, on a 800×600 screen resolution (still used by a decent percentage of the population) Yahoo results are all paid except for the top organic listing. If Yahoo has a few special offerings for that result, or it has a decidedly local flavor, it’s possible that no organics from Yahoo will show, but instead local.yahoo.com links will be in that single spot.
What will make this difficult for advertisers is making determining exactly how your listings look at the 70 and 190 character limit. It might be best to build a 70 character ad for Yahoo; and then layer on an additional 120 characters for their partners.
The paid results on Yahoo are still based around the original pay per click engine launched in 1997 (full Yahoo Search Marketing Timeline here).
Several marketing studies, search usability studies, user experience studies, etc, has led them down this path. Interestingly enough, these new limits are strikingly similar to Google. One of the major differences between the search experience on Yahoo and Google is additional integrated offerings.
While both Yahoo and Google have properties that include classifieds, images, news, local, etc. Yahoo has many offerings for small businesses such as hosting, offerings for romantics such as personals, offerings for travelers such as Yahoo travel, etc. How Yahoo maintains exposure of it’s own properties while giving searchers a wide selection of options for finding information will be interesting to see.
Hopefully, they’ll finally give in and include Mindset results.
It has yet to be seen how this will affect CTRs (click through rates) of advertisements on Yahoo and how much this will help their profitability versus the user experience.
I’ve been expecting this, and several other major changes from Yahoo, in Q of 06. Stay tuned for more, this won’t be the only significant change Yahoo makes in the next few months. They’re starting to work on their magic formula for success, and not likely to slow down anytime soon.
Increasing the Effectiveness of AdSense Ads through Section Targeting
December 13, 2005
One of the largest problems with AdSense is often the ads are irrelevant to your content.
This is especially true of blogs where the ads might be about blogging, or the rest of your navigation, and not about the specific topic on that page.
The more relevant an ad is to that exact topic, the higher the relevancy to your reader. The higher the relevancy to your reader – the higher the CTR (click through rate).
Alternately, the worse the targeting is, the lower Read more




ShareThis





