Publishers Bought Millions of Website Visits They Found Out Were Fraudulent
buzzfeed.com
buzzfeed.com
That's a mind bogglingly big number, if true. I don't really find myself feeling sorry for advertisers, though.
If true fraud levels were exposed it could shift the digital economy in a massive way.
Say you have an unpopular dishwasher and it breaks. It's great to go to Youtube and find someone showing you how to fix the problem even for a relatively niche dishwasher. But if the ad money dries up for that content producer, that video won't be there the next time you look for it.
Or you will be forced to pay $1.00 to watch it.
Not everyone who makes YouTube videos is doing it for ad money, especially the sort of content you're referring to. A lot of people just do it to document things for themselves and use YT as a "convenient place to put videos"; the fact that everyone else can see is incidental.
But the only reason that YouTube is there as a platform that allows people to upload videos for free is that they expect to make money either off of that video through ads, or through other videos through ads.
Just like roads need to be created regardless of their profitability, I think it's important for society that we have means to share and view image and video content regardless of profitability.
Patreon works great when you enjoy someone content on a continual basis. These 1 off trouble shooting type of content producers need to find their own way.
If the advertisers thought they were paying $500 for 10000 clicks resulting in $1000 worth of visible revenue, the fact that they really only got 5000 "real" clicks resulting in the same $1000 worth of visible revenue doesn't really change the equation much.
Over the long run companies spending money on advertising are going to hone in on how much they can spend compared to how much revenue they get can trace back to that advertising spend.
The pay-per-click model and the current cost-per-click pricing is just an implementation detail of the current state of web advertising. Advertising still works when those stats aren't available, like in Podcasts.
They should have been able to dedicate half the money for the same revenue, or the same money for double the rev, in your scenario.
Put another way, the estimate is that they pay out $500 for $2k in revenue, then someone came in and dumped a bunch of bots in the way and ate away half their revenue. It's not like those empty clicks are people who just weren't compelled to buy what they were selling. They are bots intentionally disrupting commerce.
In some niches, there's an unending new supply of get-rich-quick newbies that will pay for underperforming ads.
There's really no incentive for Google's AdSense team to focus on fraud in these spaces. Because the suckers just keep coming. Google profits from it, so there's little incentive to fix the glitch.
It’s simple... ad value or die as a business. Is that easy? No, but why should it be?
You're getting paid with advertising dollars for posting on hacker news?
Surprise: you (and all the other posters on hacker news) are the content producers I love. We've never been paid for this nonsense, and I doubt we ever will be. Won't stop us posting though. I'm really not that worried if buzzfeed and medium go under.
Now, I‘m not sure YouTube itself will survive in its current form if ad money dries up.
Brands looking to build awareness are buying access to this fraudulent traffic.
Advertising agencies, ad tech companies, and publishers more times than not are incentivized to look the other way and not investigate what is going on too closely as long as it is profitable.
Digital display advertising has long been considered a channel of questionable effectiveness. But the current state of programmatic advertising has made things much worse and will ultimately lead to Advertisers pulling billions of dollars out of this space unless there are significant improvements in getting more transparency into their advertising investments.
That is believable to me. Larger than fraud in a lot of markets, but not particularly large, and everyone knows the market is full of bots, bad stats, and scams.
Edit: Oh did you mean advertising reports for sessions?
I may have forgotten to acknowledge that I never really cared I didn't get all the clicks -- I was doing a CPA campaign, so I just baked in the non-visits.
Depending on targeting and platform, there was almost always a 20-70% discrepancy between the clicks I paid for (reported by the ad platform), and the visits that landed on my site (reported by GA, server logs, piwik, and analytics.js).
Also interesting - AdWords was always within 5%.
I'm sure it differs case by case, but I would naively expect the latter is more common if you have competitors.
There's never a right answer to "am I charging the best price", it's always a guess.
The problem always is that it very difficult for an observer to tell the difference between an educated guess and just a, like, regular guess. So I don't fault anyone for interpreting it like that.
When I was shopping for lawyers, not that it was a consideration in my evaluations but I found that lawyers I recognized from subway or daytime TV ads were considerably more expensive than random selections from the phonebook. Their ads project images where you've been victimized, and you need a legal gladiator, willing to fight the evil empire to get all that settlement money that rightfully belongs to you. I found that the service was better with the random selects-- free consultations, quicker responses to inquiries, more services rendered, better aftercare, etc.
Dentistry is the same. I've noticed a lot of billboards for dental services around my area. Most of them evoke Fisher-Price colors and softness, playing up to peoples' fears in how the experience doesn't have to be scary or painful. But when I did the legwork for quotes on significant dental work, I found that those that offices which advertised were much more expensive than random Yelp selections. The quotes varied by thousands of dollars for the same work.
Auto repair does things slightly differently. I live in a small town and don't have much choice in this space, but for lack of time I recently took a vehicle to a shop that bought ad placements simply because aside from that their reputation in town was better. Their quote was fair and accurate given the job and they did the work without playing any games or breaking anything else. But then they tried to upsell me on additional services (fluid changes, etc.) at absolutely obscene prices. Since auto ownership is a cargo cult where nobody knows even the basics of what goes into making them work, I'm sure this tactic works on plenty of other rubes.
Similar story-- I had a car break down on the side of the road on a long road trip. AAA towed it to the nearest repair shop, which happened to be a dealership (dealerships generally spend mad money on ads, and already have reputations for being overpriced). I know the failed part cost less than $300 and would take less than an hour to put in. They must have seen the out-of-state plates and the after-hours dropoff and banked on my presumed desperation and ignorance because their quote was $6000. I said no, had it towed to another garage and got it done for $450.
The presumption is that consumers don't (or in some cases, literally can't) shop around, are generally unintelligent and are easily swayed by appeals to emotion, so you play to these factors and charge as much as you can get away with-- not some paltry "(cost of goods + cost of labor + cost of advertising) + 10% markup" silliness.
My rule of thumb is that anything you see advertised is a sign of a hustle. If you don't know how the hustle works, you'll get got. This applies doubly so for anything marketed to the poor or minorities.
These business models seem like they’ve been modeled after real world JIT manufacturing logistics. Deliver X units of product A to site B on this schedule to meet these quotas and fulfill this growth pattern.
But real world manufacturing chains deal in real physical goods that can be easily verified at the point of use. But in the world of online content manufacturing, all the intermediary products are ephemeral electronic side effects of human action that are easily duplicated artificially.
I hate to say that this world is due for a bubble burst but I hope it happens sooner rather than later. Because if it’s working as envisioned currently then our world is going to get more and more unpleasant.
Why does it matter that millions of the ad impressions were fake?
It also assumes you have enough impressions from a single source to actually be able to make a statistically significant decision - if you only get 1000 impressions from a single site, not getting any conversions may not be out of the ordinary.
For car companies they track website sessions, chat/contact us forms/phone calls, and "vdps" (vehicle description pageviews, basically the product page for a car). There are services that claim to track back a car purchase to user's online habits/ads viewed but I've never had the chance to try it out.
For Coke/McDonalds you do brand lift studies which are expensive and difficult, but if you're doing branding advertising you can afford it. They basically test before and after how many people know your brand and how they feel about it. If your numbers are flat thats good, if they're up thats great.
Last touch credit isn't always the best, but its what everyone uses and makes sense most of the time as long as you track post-impression conversions for display and social and give them partial credit. (I'm a fan of time decay models, but since google analytics is last touch that what always gets used.)
But here's the thing: all of these things are fake-able. chat sessions are probably the least gameable metric since you'd need to target that specifically, whereas all the others can be done in a completely non-ad specific way - just browse the site in a statistically normal way, but how many impressions do you need to tell that the chat metric isn't tracking the other ones and become suspicious? And do you just not advertise on longer tail sites where you can't get that volume?
[EDIT]: I mean, click throughs definitely get faked, and post-click behaviour too. I don't know what chat rates you expect, but if it's small enough you might have people manually doing those chat sessions. Not all of these bots/fraudsters are naive.
The long tail is super important and you can't exclude it. How you can see those individual sites (by proxy) is through something like a supplier vendor, fold location, geo, etc.. so you can't see the site, but other factors you can bid will point to and show that fraud.
I think my point is that a skilled campaign manager can get rid of fraud, but the average marketing manager can't. That is a problem for most businesses, but also why agencies specialize in things like rtb.
But really, when you are a large corporation that advertises through hundreds of channels there is no way to attribute conversion to particular ads. You are paying to have your ads seen by real people to create brand awareness and get the multiple touches that will result in conversions. If those ads aren't getting seen then obviously they are not effective.
A channel is say Display vs PPC or Organic - I think you meant campaign.
I am talking about large corporations like JPMorgan Chase (the subject of the article) that are on the internet, tv, radio, billboards, print, sponsoring conferences, sports teams, lobbying governments, thousands of physical locations and ATMs that people see every day, etc, etc, etc.
Even with a PPC ad that results in a conversion can you really attribute that sale to that particular ad?
For a large corporation no, because you can't tell what part of the sale is due to the corporation's existing brand that is built up through exposing the consumer to that brand thousands of times across multiple media.
Conversion rates for digital ads are only meaningful relative to the conversion rates for other digital ads. They do not not allow you to attribute a value to digital advertising relative to the rest of your ad spending.
If there are fake impressions then your actual ROI is higher than stated, so it's likely there is another strategy that is less effective that you may switch to without realizing it.
Sophisticated advertising campaigns involve looking at the actions of each source of traffic separately, because there will always be bogus sources.
http://adage.com/article/cmo-strategy/p-g-slashe/309936/
Plus, the money lost to fraud could have been better spent elsewhere.
Edit: Here's the part I am referring to:
> Publishers often buy traffic at the end of the month or quarter to ‘make its numbers.’ Traffic sellers often promise the publisher that the traffic is human and will pass through all ad fraud detection filters.
So I am not questioning advertising but blatantly buying traffic.
Edit: In response to OP's edit. "Buying traffic" is literally just advertising. You spend x dollars and that drives y traffic. Its worded like that so it sounds bad.
The problem is when you're paying per impression in advertising, as that's the easiest to game - the number of shitty impressions I've seen coming from small, not-well-known data centers has taught me that cost per impression is so prone to advertising fraud as to not be worth it.
The most egregious fraudulent traffic came from at least a dozen different no-name data centers, and it was actually easier for us to cut out the fraudsters selling us the traffic than to blacklist the traffic itself (even though it was several steps away from the original source the traffic was purchased from).
The traffic itself was actually rather easy to identify independently of its actual network source, because they were using Selenium locked to a specific, outdated version of Firefox that also happened to consistently throw some errors that our standard infrastructure monitoring caught and alerted us to.
I still think that cost per impression is a difficult traffic source to properly verify (especially if you end up with a traffic source that is botting from compromised PCs) and as such should be treated with skepticism. Cost per conversion/install/etc. generally seem to be slightly less gameable (excluding, of course, the farms spread around economically disadvantaged areas where workers have hundreds of actual devices on which they are installing apps to game the cost per install ad traffic)
Go out in the street and give people $2 each to view the website?
In the print world, it would be like WaPo selling its advertisers on 100K copies circulated to ABC1's, but actually handing a quarter of the papers out for free on Skid Row.
Publishers who buy traffic should absolutely know that the traffic is unqualified, unknown viewers... the least valuable. That means the publisher is selling ads based on their best qualified customers but delivering a fraction of dregs along with them. The publisher plays an arbitrage game with their reputation.
That practice can't help but make the advertisers discount the value of digital ads.
Not inconceivable. This past Thanksgiving I got the local paper's black-Friday deals edition delivered for free to my doorstep. I am not nor have every been a customer of that particular paper.
For a while, Best Buy was handing out free Sports Illustrated subscriptions any costumer with a pulse who'd give them their address.
Also consumers are getting smarter by using ad blockers.
The only thing left is these schemes making the illusion that real people are watching.
You can "reach" thousands of people for less than $200 online, whereas conventional advertising is substantially more expensive.
This has to be one of the underappreciated reasons that traditional media has been savaged so badly by the internet - the price difference is stark, especially on a CPM level. But that price difference is also jacked up by massive fraud everwhere - for example Facebook has told advertisers they have access to more people in the US, than live in the US.
That's not to say that traditional media is clean in how they measure their audience, but at least with a traditional ad you can turn on the TV and watch it, or pick up a newspaper and read it, or drive past it. How many digital ads never get seen by a human?
Fraud seems to happen more frequently in situations of distress. Like Madoff, who had to expand his fraudulent operation to hide his past wrongdoing.
No idea if anyone ever bothered to look, I just moved on.