Wednesday, June 17, 2009
Microsoft has gotten us one step closer to the hard-line Traffick stance of jail time for click fraud. Making money by directly yanking money out of advertiser pockets in industries like auto insurance must have felt very good to these alleged fraudsters: until the day they were slapped with a $750,000 lawsuit. Ouch.
Labels: click fraud
Saturday, March 28, 2009
Because my piece boiling down the whole click fraud controversy appeared elsewhere, many of you may not have seen it the first time (exactly two years ago)! My opinion hasn't changed. I still think that most of the discussions of click fraud are a sideshow to the serious concerns of marketers. The search engines' systems are not perfect, but major problems are rare and anomalous. This should be the subject of periodic audits, to be sure, but most of the time you're not going to find enough problems. Conclusion: much click fraud talk is driven by click fraud solution vendors. Sound familiar?
Anyway, other than savvy use of reporting and analytics in general, I find some tools to be handy for those periodic updates. PPC Assurance by Enquisite is one of them.
Rerun: Click Fraud: Sideshow vs. Reality
Excerpt: If you've owned a website for any length of time, you've probably received one of those spam e-mails offering you some scam search engine optimization service. The most recent twist on this kind of spam is: "Triple Your Google AdWords Results (By Reporting Click Fraud)".
This latter claims that advertisers are suffering "exponential losses of profit," and that if you submit a "periodic report to provider," the search engines "must comply promptly to your reports." Inane prose and poor grammar aside, what these claims are, most of all, is nonsense.
Labels: click fraud
Saturday, March 29, 2008
In case any of the naysayers are still listening, here's the scoop on how Google looks for patterns so it can do the best job of proactively filtering fraudulent clicks.
No, Google's post does not give away all the methods, of course, but if you engage him in conversation, you'll find that Googlers like Shuman Ghosemajumder are willing to go into a little more depth. For example, fraudsters know that clicking on a certain ad will create anomalies in the data that will trigger further investigation of their IP addresses, methods, or the targeting of a particular advertiser's ads. So they try to "even out" the anomalies by creating more impressions to balance out the clicks, and other attempts to make the fraud seem more "real." But all this does is dig them in deeper, creating other anomalies.
To sum up: we don't see much click fraud these days with our clients' accounts, due to meticulous campaign organization, careful attention to geotargeting, filtering methods on our campaigns, and a variety of other cautious settings, to say nothing of Google's proactive filtering.
To verify that an account is running without major fraud problems, I find a fourth-party audit service such as PPC Assurance helpful, as stated here before.
If you say that in a conference session on click fraud, the anomalously high number of bearishly inclined attendees who are anomalously more likely to fill out the comment form at the end, are unlikely to give the above analysis a high rating. They're unlikely to give Shuman a 5 out of 5 and say that his talk "kicked ass." It doesn't change the facts, though.
And that ties into a future post about the quality of speakers at conferences, and the weakness of evaluation stats. Highly polarizing speakers can be near the top of the list in terms of the quality of the information they provide, but the combination of 1's and 5's they receive can actually make them look "bland" at a 3.3 rating. Talk to you about this next week.
Labels: click fraud
Tuesday, October 02, 2007
Making recent headlines has been a heated debate between a click fraud auditing vendor, Tom Cuthbert of Click Forensics, and Shuman Ghosemajumder, head of click quality at Google. Who's giving us the straight goods? Cuthbert, who has recently spearheaded an industry group called the Click Quality Council, claims click fraud is growing, and continues to stick by numbers like 10% and 15%. Ghosemajumder has repeatedly presented much lower numbers.
Both parties get into a bit of a side debate about the unique gclid modifier attached to every Google paid click. But to an outside observer, this does little to illuminate the patterns going on inside individual accounts; especially around Google's claim that they are proactively refunding virtually all invalid clicks.
You almost feel like you need a fourth-party auditor to help you audit the independent auditors.
The data I am seeing show that Google may (still) be closer to telling the truth than Cuthbert is. Google does proactively refund clicks; clicks in many accounts appear quite normal a high percentage of the time; there are some gray areas.
Interestingly, the data also show there are ways of managing a campaign (professionally vs. haphazardly), and choices you can make about which parameters and techniques to use, that will run you into more or less trouble.
And even when it comes to well-managed, normally running accounts, things can go awry. It's important to have some kind of monitoring tool for this - although you'd get a good cut at it with some of the other (non-click-audit-oriented, campaign management oriented) third party campaign management tools. A hard-working analyst using good web analytics could gain strong indications based on a granular assessment of "bounce rates" or time spent on site by keyword or ad group, but this methodology is weak and only really indicates that traffic is untargeted, potentially. Better than "bounce rates," as counseled by John Marshall ex of ClickTracks, is to look at "very short visits" (not the same as "bounce rate").
To help sort through the claims, I have been involved in beta testing with PPC Assurance, a product being developed by Richard Zwicky's team at Enquisite, a startup in the search analytics space.
In their interface, various screen types are available that illustrate your campaign patterns intuitively using a color coding scheme.
- Then, there is "green." Clicks you paid for and that fall into the terms of service you agreed to. The vast majority of clicks in any account fall into this category. There may be some irritating gamesmanship (competitors manually clicking, etc.) and some poor quality traffic inside that green area, to be sure, but Google is saying they also try to filter for that stuff. Ultimately the ROI on your campaign will tell you if "green" is putting enough green in your trousers.
- "Red" is bad. These are clicks you paid for, and that fell outside the terms of service you agreed to with Google. For some reason, even on this simple definition, many accounts have between 1% and 10% of this type of traffic. Even if this is getting up close to 4-5% you may need to look for a refund. But more importantly, you can use a tool like PPC Assurance to see when spikes occurred, on what keywords, from what geographic locales or problem IP's, etc. The information is so well packaged in their interface already, says Zwicky, that soon you'll be able to send a refund request with associated data, with a single click.
My next point should be reassuring to anyone who manages campaigns for a living. We compared a professionally managed campaign, one we have been working on for three years for a UK retailer, with a well-meaning, but amateurishly managed campaign. See screen shots below. (These were not hand-picked to make this point -- they were just two early sites in the PPC Assurance beta.)
- "Yellow" is central to this whole debate. These are clicks that fell outside the terms of service you agreed to, but that Google (Yahoo is coming soon in PPC Assurance, Zwicky assures us) did not charge you for. The first key to the yellow area is that you're going to be getting fairly accurate information that seems to dovetail with Google's own claims -- in fact, they are proactively refunding a lot of questionable clicks. But another thing you can do is to gain insight into click fraud patterns generally, without much effort. By looking at the "yellow" click data click by click (if you have time), you can see what kind of wacky behavior is going on out there. But no, you didn't pay for it.
The two screen shots below show first a "normal" campaign that had some click quality problems. Some underlying reasons for this include poor keyword selection and misunderstanding geographic targeting. It may also include reckless use of content targeting. Setting geographic targeting very tightly also places a difficult onus on the provider of the clicks, so campaigns that are local in nature can often run into apparent click quality problems because by definition you're asking for something the provider cannot deliver as accurately.
The next shot shows a "perfect ppc" or at least optimized paid search campaign. In these cases, the campaign was organized on sound principles of granularity, long tail, testing and managing to ROI objectives, careful control of content bidding, and an understanding of basic parameters and settings.
As you can see, the picture as shown by the second chart is not too bad, and unsurprisingly, the "perfect ppc" campaigns have made this client a lot of money over the past three years.
Disclosure: I have never been paid to talk about Enquisite or PPC Assurance. Their sister services firm, a search marketing firm called Metamend, is functionally separate from Enquisite, but Zwicky founded Metamend and remains involved. My firm, Page Zero, focuses on optimizing paid search campaigns, sometimes refers "SEO" business to Metamend, and vice-versa, and in some cases, we have collaborated on client projects. Metamend staff have imbibed with members of my company at the lobby of the San Jose Fairmont (I won't tell you what they drink -- it's embarrassing.) If I continue to believe that PPC Assurance is the best click auditing solution for clients, in future I will consider reselling the product for commission.
Labels: click fraud
Thursday, March 01, 2007
For some odd reason, I expected the Washington Post to improve its coverage of the search engine industry. Nope. It seems to have gotten worse.
Sara Goo picks up the click fraud topic with this helpful effort:
Maybe I should back up here. What is click fraud and why should you care? Google makes money every time people click on the "sponsored links" places next to search results, which are really text ads paid for by advertisers.
Wow, oh wow. It appears we'll have to get ready for every article for the next five years being mostly taken over by the boilerplate "here is what a search engine is" preface, so readers never actually get to learn anything. Hard to blame the author here. Her editors made her do it.
These ads appear not only next to search results but also as text ads that Google places on blogs or other Web sites, labeled "Ads by Google." (I never click on these ads but apparently, a lot of people do because it provides Google with a significant chuck of its revenue.)
Indeed, a significant chuck. Like 97% of $11 billion per year. Sara Goo, are you trying to rile me up with that old "I never click on the ads" thing? Hey guess what Sara: I never buy anything from the advertisers who put their ads in the Washington Post! Tit for tat! The "I never click on the ads" gambit is fine and expected from the average person in cocktail party conversation, but you're a tech journalist so it seems disingenuous to my ears.
Each time someone clicks on the ads, the advertiser pays Google and the Web site displaying the ad. But fraud occurs when people repeatedly click on ads with no intention of really viewing the ad; rather, they want to drive up the cost for the advertiser and increase revenue for the Web site owner. It's more sophisticated than it seems. People create sophisticated "click robots" to make it look like authetic clicking when it's really not. People have also created networks of clickers, or people paid to click on ads to evade Google's click fraud filters.
It's an industry-wide problem--not just for Google--and one that has been compared to the days before newspapers and television had any accountability to advertisers to reveal true numbres of how many readers or viewers they were delivering to advertisers. Similarly for the online advertisers, there is no independent organization now monitoring how many clicks are fraudulent and many advertisers don't have the technology to know the difference.
Still clinging to "wild West" mythology vis-a-vis the Internet? Claiming that newspapers and television are bastions of accountability? It's getting hard to take this. Well, at least you're making it plain why we won't be seeing fair and balanced coverage from you anytime soon.Also, you made a second major spelling error.
And finally, I'm resentful because you just have the word "Goo" under your photo. I'm Goo. It's been my nickname since university.
Anyway, the rest of your piece is pretty OK, but I see we're always in jeopardy of sliding back into old-media-vs.-new-media rhetoric. It will be good to see some independent auditing of clicks, and wait a minute, did I hear you say it's Google's competitors (other than perhaps Microsoft) who would be in more trouble if that happened? Google stands to lose a lot less from independent auditing of clicks precisely because a much lower proportion of fraudulent clicks are being charged to advertisers. In any case, all of the top four firms can afford to "step up" to allow more scrutiny of clicks by third parties, and I look forward to the day. Financially, it will hurt Yahoo and Ask the most.
FOLLOWUP: Google has now posted an extensive report on the Inside AdWords blog.
Labels: click fraud
Wednesday, February 28, 2007
CBS Marketwatch is ahead of the curve with this item. Google is releasing more detailed information about click fraud. Last time around, they were somewhat transparent in their approach as they chatted with Andy Beal. However, some stats never got fully examined - so now Google is redoubling its efforts to explain both the magnitude of the proactively refunded clicks (it amounts to over $1 billion per year in forgone revenue), but also the relative tininess of the proportion of clicks that get refunded by request in investigations after that (below 0.1%, well below). Recently I had an opportunity to speak at length with Shuman Ghosemajumder to discuss some of the stats and also Google's fraud detection methods. IMHO they have some pretty interesting tricks up their sleeve on the detection front.
Having gone over some of their info so far and reflected on how it squares with current opinion in the industry, I just completed an audio interview (a PZCast to be exact, hosted as always by Mona) with some detailed comments on the click fraud controversy. The MP3 file will be posted to the member area for Page Zero Advisor subscribers by noon Thursday. If you listen, remember as always, the initial 90 second throat-clearing banter is only outdone in its inanity by the official banter lasting three full minutes. Fast forward past both if you prefer content to yuks.
Labels: click fraud, google adwords, page zero, podcast, shuman ghosemajumder
Monday, February 05, 2007
In a densely-written post about click quality on the Inside AdWords blog, Google provides a surprising amount of information in a short space. They combine an informative approach to different types of click fraud and the steps advertisers need to take to document it, with a new, consistent approach to communications that includes a dedicated contact form.
Google has been opening up its communications in several areas. For example, a member of that team took the trouble to chat at length with Andy Beal about the invalid click detection process. And Nick Fox of the Ads Quality team has commented directly on this blog on issues relating to arbitrage and the adwords ranking algorithm.
Permit us one small note of churlishness however, by way of defending conventions of ordinary human conversation the way we remember them pre-Google. "Meet the Click Quality Team" as a post title gives the impression we'll be hearing attributed commentary from identifiable individuals. We might even get names, or a group photo. Here, we're getting nothing of the sort, so we haven't in fact "met" anyone, although information has been disseminated. Also, there is a big stylized "open quotation" mark, and the use of the first person (singular as well as plural), but the quote is never closed and the person(s) is/are never identified. Am I the only one confused by this?
Labels: click fraud, google adwords
View Posts by Category