False Ad Clicks Cost Google 1 Billion Dollars A Year 233
Meshach writes "There is an interesting story at CBC which claims that Google loses one billion dollars per year to fraudulent ad clicks. The article contains an interesting description how how the company determines if a click is false. 'The company explained that it determines which clicks are invalid through a three-stage system. Most of the illegitimate clicks are automatically detected analyzed and filtered out in the first stage ... The second part uses automatic and manual analysis of the AdSense network to weed out false clicks before they are logged to an advertiser's account.'"
Ledgerlines (Score:3, Interesting)
Actually there was some other article I read recently about how much Google probably makes off of that, but I can't find it now.
Google does not "lose" 1 Billion per year (Score:5, Interesting)
Google isn't out any cash for the fraud, it is people who **buy** Google ads and pay per click who potentially loose money to fraudulent clicks, not Google. And there no way that Google can catch all click fraud, so it is **inevitable** that at least some advertisers will be charged for fraudulent clicks.
Nice post. Way to make Google look like the victim when they aren't the ones who actually pay for fraudulent clicks.
A Merchant's Perspective & Article Critique. (Score:4, Interesting)
We merchants/advertizers are the ones screwed. Google says that 10 percent of clicks are fraudulent? I have zero idea if this is an over-estimate, under-estimate, or dead-on accurate. However, I do know that google has very little incentive to "mark down" my bill every month. My family runs a small business -- http://www.beadstore.com/ [beadstore.com] -- and sometimes advertise on google. How many of those clicks I pay for each month are fraudulent? Who knows. I certainly can't tell.
This isn't to say that I distrust Google. The fact is, that when we advertise, our sales go up. So something is working. Advertising on Google makes a bigger difference than any of our other venues. But those numbers suggesting that 30 percent of our advertising budget may be/once was/is potentially lost to fraud? That is truly scary.
Re:Bad math, bad logic. (Score:3, Interesting)
Google discards those clicks voluntarily. If they hadn't, they could be charging the advertisers for each such click, and would be making more money.
The logic is not bogus.
Re:Bad math, bad logic. (Score:2, Interesting)
"every percentage point of invalid clicks we throw out represents over $100 million [US per] year in potential revenue foregone,"
It is "potential revenue" because if they hadn't done the analysis, they could have charged advertizers an extra $1+ billion dollars. The + is for whatever extra money went to the website owners (zero if they weren't paid for the false clicks, + if they were). By detecting the false clicks, google quite directly makes less revenue than if they didn't detect the false clicks. Of course the problem is that advertizers won't renew the contract if google advertizing is ineffective, or if they feel cheated. So Google is saying "we could have made $1 billion more dollars, but we are good and threw out the bad clicks instead".
The interpretation by the writer is bogus as you say. The false clicks themselves only cost google extra work money, and whatever their reliable image is worth.
Re:Ledgerlines (Score:3, Interesting)
Re:Bad math, bad logic. (Score:3, Interesting)
For example suppose they can afford to audit a random 1% of clicks they detect are bogus and 1% of clicks they detect are legitimate. Assume 1% is statistically significant. They can compute with known accuracy how many clicks fall into each of these categories:
1) Legitimate clicks detected as legitimate.
2) Bogus clicks detected as bogus.
3) Legitimate clicks detected as bogus.
4) Bogus clicks detected as legitimate.
Type 1 clicks, obviously, are great. Type 2 clicks don't harm Google significantly. Type 3 clicks are the big problem for Google as these result in lost revenue. Type 4 clicks result in income for Google but harm their reputation. Ultimately, this might be reflected in lower ad rates or fewer customers.
But your basic argument is flawed. They may be able to detect bogus clicks very reliable with a method that's impractical for use on a high percentage of clicks.
Re:Bad math, bad logic. (Score:3, Interesting)
Thus, my criminal friend's fraud is costing me a million dollars. See the problem?
Fraud Is Bad, Made for AdSense is worse (Score:5, Interesting)
The AdSense block is under the header for each stage in the workflow, which suggests to unsophisticated Internet users that my ads ARE the next stage in the workflow. You might think I'd be happy about that, because it means a lot of users naively click on me thinking I'm the next step in the workflow, but ALL CLICKS ARE NOT EQUAL. As soon as somebody clicks on my ad, they get whisked to a completely different site and realize "Thats funny, something must have gone wrong". So they click back and I'm out nine cents. Repeat times a couple of hundred over the last 48 hours.
My CTR (click-through rate) for ads on other sites is in the general region of 1%. Thus, I can reasonably assume that about 1% of the audience reading content with my keywords is at least marginally interested in the product I sell. The CTR on ads on these pages which drew clicks by visual deception was in the teens. That means 15x the earnings for the owners of the deceptive pages. However, the conversion rate (percentage of folks who go on to download my free trial or buy from me) from customers with normal levels of interest (i.e. from other AdSense ads, for example) is about 20%. From these pages, it was less than 2%. Thus, the revenue split from a sale of my software goes from something like 40/40/20 advertiser-Google-me to 100/100/-100 advertiser/Google/me. (I am obviously hoping to tweak the campaign to the point where it is closer to 20/20/60, but even at 40/40/20 its still a positive return on investment.)
Anyhow, when you work out the math it had me paying something close to $25 to generate a fifty-fifty shot of selling a $25 piece of software. I've since banned the deceptive sites (you can manually choose to not allow your ads on certain domains or URLS), of course, but there are still advertisers getting screwed by them as we speak. And, looking through my logs, there are a LOT of sketchy sites in AdSense which would have cost just as much if they had been blasting through enough traffic. That really threatens the utility of the platform. If its 75% conmen to 25% upstanding sites like Mrs. Smith's Teaching Resources there is no reason for me to pay a single penny for the ads since I'll have to babysit the campaign every hour or get a negative ROI.
Re:Well, that's what you get... (Score:3, Interesting)
Yes, fraud does happen, but why should the advertisers get an undue break when its only on a small scale and not organized by the website operator? It seems awfully fishy that google seems to think that their system is that good.
From what the article says, I suspect that is exactly what is happening if only 0.02% get reported as falsely getting through the measures, you can be pretty sure that google is cheating the website operators. Cheating them by being overly conservative in which clicks to be counted and which ones to toss as being fraudulent. It would be quite surprising to me if the majority of that percentage was a legitimate case of fraud and not a few advertisers trying to cheat the system. 0.02 is unlikely to be larger than the margin of error in their analysis.
I remember the brief time I had one of their accounts, and it rarely worked right, the javascript was regularly broken and rarely actually showed any ads that weren't charity ads.
Re:Ledgerlines (Score:3, Interesting)
Well, "that much" is subjective, wouldn't you say?
Your point is that for GOOG to make ONE BILLION a year, they'd have to have 100MM accounts with $99.99 in them.
Not at all likely, for obvious reasons.
However, wouldn't you think that, oh, ONE HUNDRED MILLION would surpass "not that much"?
If so, they'd only need 20MM users with an average of $50 in the account.
That's still not all that likely, but it's certainly doable. And we're talking $275,000 a day in revenue (well, cap gains to be specific) for absolutely NOTHING in return.
And I don't get your point about needing to keep the amount "close to liquid." Do you think they create a special little box for each users money and when it hits $100 they take the money out of the box and send it out? That's just silly. Much like a bank, they'd keep, MAYBE, 10% of the cash in liquid assets. But really, I don't see why they'd need ANY of it liquid. When a check is written, it can come out of their "general fund." Just because it's coded into the AdSense GL account doesn't mean it would have to come out of a segregated bank account.