I like the tone of this post refuting the widespread coverage of Eric’s click fraud comments. It feels….like Google being part of a conversation. I wish it came a bit quicker, but it’s worth a read.
Update: Donna Bogatin responds to Google’s response…she’s pissed….
12 thoughts on “Google: Uh, No”
Reading the blog post, this paragraph struck me:
The “let it happen” excerpt followed, in which he discusses the economic forces that can retard click fraud: “Eventually the price that the advertiser is willing to pay for the conversion will decline because the advertiser will realize that these are bad clicks. In other words, the value of the ad declines. So, over some amount of time, the system is, in fact, self-correcting. In fact, there is a perfect economic solution, which is to let it happen.”
I’ve heard this argument many, many times.. often on this very blog. But I’ve often wondered how true it really is? I’m a scientist, and not an economist, and not an advertiser. But it feels like it could be true under circumstances in which (1) fraud happened equally to all parties involved (i.e. 10% of everybody’s budget was eaten by fraud, equally) and (2) everyone had a fairly substantial bankroll, so that they could actually take advantage of this 10% expected value. i.e. if I only have enough budget for 5 clicks, and one of them is fraudulent, then I’m already at 20%.
But if you cannot guarantee that everyone is affected equally, like a flat tax, is it really true that the market will self-correct in its pricing? Won’t some advertisers who have been hit with 40% clickfraud just drop out, while other advertisers for the same keyword who have only been hit by 3% clickfraud be willing to pay the same current price, and feel no pressure to lower? And because the 3% people are just as happy to pay current market rates, won’t that further discourage those hit by 15% clickfraud, so that they also now decide to drop out, because thanks to the 3% folks the prices aren’t lowering to match their 15% experience?
In other words, how is the market supposed to correct under unequal conditions?
I honestly am asking this question. Is there someone (an economist?) that knows more about this stuff? Is there a situation elsewhere in the economy that is analogous, that we can drawn from and learn from, or that makes the argument plausible by reference?
the concept is true from a theoretical perspective, however real economics never works according to theory b/c one of the necessary conditions is that there is perfect information.. transparency to all parties.. so the costs can be observed and factored into the decisions of everyone comprising the market.
should google decide to offer up these data to advertisers they could factor that into their decisions and the invisible hand would, as schmidt suggests, work its magic (even further than with the rough estimates offered today)..
i’d love to see that.. and even think google would love to do that.. but it’s at the risk of the impact of the perception created in the minds of all those who aren’t savvy enough to see and account for the new info.. the probability that those customers would potentially move to, say, yahoo, if yahoo took the other track and elected not to disclose such information is why google has to walk the fine line on this for now.
disclosure of this type of fraud could be legally required so that it applies to all – that would help – but then the incentive changes from a company trying to do its best to detect click fraud to trying to do its best to not detect it but meet the regulatory req’s so that it looks better relative to peers…
standards could be applied, etc.. but the whole things becomes expensive and complicated. today we still have the invisible hand working, behind the information provided by journaists covering the issue and google, yahoo, etc to the extent they shed light with data, it’s just not perfect.
but in this case, right now at least, for reasons i mentioned i’d argue that this level of imperfection is appropriate to the current state of this evolving technology and market.
btw, i’m not a professional economist but hold a degree in it.
Schmidt’s comment is pretty scary because it shows he does not understand the dynamics of click fraud at all, much less economic theory. (I am a Google advertiser and an old econ major but neither of these things is particularly relevant). His argument is completely wrong for several reasons, only one of which I will highlight. One of the primary forms of click fraud is competitors clicking on links in order to drive up others’ costs. Under this scenario, if I am the victim, I am being subjected to the fraud and having my costs driven up while my competitor (perpetrator) is not being defrauded and is now able to bid less for his traffic as I eventually drop out. This is not a self-correcting system. In the extreme case, if the perpetrator can do this to all his competitors, he eventually must only pay the minimum bid. Not only is this bad for the perpetrators’ competitors and good for the perpetrator, but Google loses lots of money, too.
It would take too long to go into it, but under the current Google system, there is not one form of click fraud that would ever be self-correcting from an economic theory standpoint. In fact, the only scenario where click fraud would “work itself out” and be efficient, as Schmidt implies, would be if all advertisers are subject to the exact same amount of click fraud and the click fraud happens in an equal proportion on all publisher sites.
erik – not being an advertiser i can’t be certain i even know exactly how it works but given my understanding, where i’d maintain the theoretical validity of schmidt’s argument applies is in the calculation of cost-benefit of that advertising channel relative to another to that advertiser.
ie, if i pay $1 per click for an ad that uses keyword X (assuming my understanding of how it works is correct), and i get 10 clicks in said period, but 2 of them were known to be ‘fraudulent’ (like as you say made by a competitor trying to drive up my costs), and sales via this channel amounting to $20, then:
rather than calculating the cost effectiveness of this channel as earning $2 on each dollar spent, i instead calculate that i earned $20 on $8 spent (that led to the earnings) and lost $2 in a kinda ‘fraud tax’..
so while my earning potential is $2.5 (20/8) for each dollar spent the actual earnings taking in account the extent of click fraud is $2 on each dollar spent. With this information I can compare Google as an ad channel to Yahoo as an ad channel by comparing the earning potential and actual between both. I then make a decision based upon that information and calculation.
in a competitive market with perfect information (even if click fraud was not uniform) people will decide to advertize where they get the greatest ROI, and so as alternatives provide better ROIs, incentives go up for the losers to invest in battling the costs of that fraud for their customers to compete for marketshare, and this is how it is self-correcting.
this is also why until yahoo and msn (and aol?) have more compelling products with more competitive earning potential on dollars spent it will not self-correct, unless of course advertisers can calculate a greater roi on an entirely different medium for their ads, which as we all know is a problem with traditional ad channels anyway that makes search marketing more attractive.
a parallel to this notion of self-correction in the context of this industry is in spam.. as it impacted customers of email service providers, people left for more spam-free services. as the cost of losign that user increased (with the increasing value of keeping that user) companies invetsed more heavily in anti-span technologies.
in this scenario it’s not an issue of imperfect information, and it’s not needed b/c the calculation is simply a given user’s experience, but the idea is basically the same in terms of how a competitive market can self-correct according to intrinsic costs.
This is sad. It’s clear from the comments here, and the blog post by Donna Bogatin, that people still seem to not understand that the Google CEO was discussing this hypothetically. He was not saying that this is how Google views it. He was answering a question. Therefore simply quoting him with the “let it happen” statement without realizing that he was anwering a question, and making a hypothetical statement is taking it way out of context. It seems like people are just skimming, and not researching what’s really going on.
Here’s the question the google ceo was asked:”Recently there’s been some talk about click fraud being a potential threat to the entire advertising business model. I was just wondering what your thoughts on that were and if there’s an economic solution to it more than just technical solutions.”
This is how the Google Ceo began his answer:”Let’s imagine for purposes of argument that click fraud were not policed by Google and it were rampant …”
Simply pulling out the “let it happen” quote without pointing out this context is being misleading.
Eric, there is one flaw in your thinking. I certainly believe competitors are the leading cause in the small niche where I advertise. But Google uses a formula where click fraud can benefit the link that is being being hit by click fraud.
Under Google’s formula to determine who is top display in Adwords, it is a combination of bid AND click through rate. So when your competitors try to run your cost up they are increasing your click through rate. Especially in niches, this will dramatically improve your display position and – through Google’s formula – actually reduce your cost per click. I have hard numbers over two years of data to back this up.
The other pay per click engines are based only on your bid. Click fraud only hurts in these engines. Google’s formula that is designed to maximize it’s revenue also benefits the advertiser inadvertantly.
the interpretation of schmidts comments and the discussion around the economic theory behind it are two different, parallel subjects.. what was interesting to me in this thread was not the former, which as you mention is no biggie and has recently been blown up in light of missing context, but the latter, which started in response to JG’s question 🙂
libicki – I believe Schmidt was making the point that Google as a closed system would self correct, not relative to other ad channels. I think you’re correct in that case, but I didn’t think that was his point.
or – did you actually read the comments here or just skim them? I believe it’s quite clear from all the comments on this post that everyone took Schmidt’s comment as one for purposes of argument. My point was that this argument was false and demonstrated a lack of understanding of both click fraud and economics. No one on this thread was under the impression that Google is considering just letting it happen.
Duke – I know exactly what you’re talking about but I believe your scenario only works under very specific circumstances that would be very unlikely to occur if click fraud was not policed at all (you’re in a narrow band where the cost of extra clicks is compensated for by the profit from higher rank/increased clicks). Imagine, for example, that 90% of the clicks you were charged for were fraud but your competitor had no fraud. Even if that moved you up to the #1 position the gain from more clicks at the same price would probably not make up for the fact that each click on average is now only worth 10% of what it used to be.
libicki, you argue that “that this level of imperfection is appropriate to the current state of this evolving technology and market.” I disagree, because this imperfection could have been avoided to a great extent by the introduction of alternative methods of charging, such as CPA (which is now being introduced), or fixed fees, which I have argued in the past offers the least risk of click fraud for advertisers. (There is of course a risk of overpaying for insufficient clicks, but this is not a consequence of fraud; rather, it is poor gauging of ROI.)
It has always been my opinion that the risks of click fraud were readily apparent with CPC and CPM models. There is an extensive body of literature, well known to people in the Internet technical community, of the ease at which traffic can be fraudulently generated and sent to web servers. The types of fraud are by no means limited to most of what is discussed publicly, such as repeated attacks from the same IP address. There are ways of perpetrating fraud that are impossible to detect, because they appear as any nonconverting clickstream would. I have always wondered why AdWords/AdSense was permitted to go live without at least some tools and options (such as CPA) available to advertisers to limit their exposure to fraudulent clicks.
Personally, I didn’t find Donna Bogatin’s comments out of context. I’m sure others will disagree with me, and some will flame me. Life goes on. However … given the fact that Google has been served with click fraud lawsuits, and that the CEO’s comments on matters that affect his business will draw attention, I would think the CEO would be a bit more circumspect in his remarks.
And now the Google blog has linked to the actual ZDNet post; and the rest of the world still doesn’t care about a petty blog war. Donna Bogatin should have stepped back and taken a day before posting that article.
Google CEO on click fraud: ‘let it happen’ is perfect economic solution
The scenario went something like this:
What would have been just an unoticed Blog posts was DIGG-ed because of the convenient JavaScipt DIGG link added on the blog – That automatically fills in the URL and Title on Digg. IT MADE THE DIGG FRONT PAGE
Then of course it was submitted to REDDIT which most Digg Front page stories are – IT MADE REDDIT FRONT PAGE
Then of course it became popular on DELICIOUS being SAVED by a number of people as TWO seperate Articled to make their “Popular” Section
It was also picked up by Battlemedia, WebProNews, as well as several other popular SEO NEWS BLOGS – each giving their “synopsis” of the ZDNET story….. And don’t forget the SEO forums posts with hundreds of pageviews.
So as we can see PROVOCATIVE titles can be very misleading – even if the stories does detail other facts. With so much news information being PUSHED – sometimes, many items just get one cursory reading – but the title has already biased the reader about what to expect. ( Including the comments made by Search Engines Web on a blog post.)
And, of course, for the tens of thousands who just visited the Digg Front page withOUT reading the story – That is what they will remember – that misleading title.
However, without a provocative title many stories would NOT pique curiousities enough to read.
In the ZD net Blog post this ONE phrase was there – but many may have just glossed over it
Schmidt’s scenario for what would happen if Google did not police click fraud and it was “rampant”:
That ‘scenario’ word was the key – but how many readers picked it up??????