Chase had ads on 400k sites, then on just 5k, with same results

If so, the fact that they don't see an impact in # of impressions means more about the ad exchange's bidding system and inelastic demand (or lack thereof) for inventory on those whitelisted sites.

I don't think this has anything to do with fraud.

What makes you think that they reduced their spend by 99%? They simply limited their campaigns' targeting to a whitelist of 400 sites, there's not any indication that they were spending equal $ per placement.

I'm not sure why you believe that aggregating statistics is some sort of lie -- in AdWords, advertisers are able to easily see the placement for each one of their clicks. It's not some big secret hidden from advertisers.

You are also confusing why and how much advertisers care about the content of the sites. The site's content is only one signal as to the quality of an impression. It's a proxy for quality, because it's a good indication that a particular user might find the ad interesting and relevant. It's not as good of a signal as the characteristics of the user. Why would Chase be interested in only advertising on "banking" sites anyway? It's pretty silly to think that it's somehow a big scandal if the website content isn't directly connected to the business of the advertiser. Advertisers definitely don't want that. What they want is a quality user -- Chase wants someone who is actively shopping for banking services to see their ad everywhere, but they don't want to waste an impression on someone who isn't. If I were guessing, they were likely aggressively remarketing on the entire network, and now they're restricting to sites. In other words, they were following users across the entire internet, wherever they could find inventory, and now they'll just follow them to the most popular sites.

You're conflating two issues -- brand safety and ad spam. The first is the relevant issue: "I don't care how good the user is, I can't have the NYT printing that my brand name showed up next to a terrorist recruitment video."

You're talking about ad spam, or "you charged me for this click but it was fraud from the publisher." That's not at all what the story is about. You're jumping to a conclusion that the now excluded websites were fradulent to begin with. Do you find it unbelievable that there might be 400,000 legitimate websites in the set of available placements for Chase's ads, without spam? Why?

You are mixing terms, Adwords isn't an exchange -- AdWords has inventory from owned-and-operated sites (Google, YouTube) and network partners (AdSense, AdMob, Google Display Network, etc.) AdWords doesn't sell any inventory from ad exchanges, it has its own network, particularly because it wants to maintain the quality of the placements. Spam is actively and aggressively policed.

Again, that's not what the article is about -- none of the big tech co's have been doing much to prevent ads from showing against certain types of offensive content -- especially content like hate speech, fake news, etc., because it's really difficult to identify correctly at scale. The media and advertisers have started caring because it's the cultural moment we're in right now, I guess.

Finally, it's not the case that "garbage inventory" is spread around, subsidized by the good stuff, in a money making scheme. Such a scheme wouldn't make sense anyway. High-spending, sophisticated advertisers know the conversion rate of the advertising, and they have a sense of the return on ad spend that they'd like. Cost-per-click is just a shallow metric in the market, if the real value wasn't there then demand would decrease.

Does the industry advertise until they break even? Of course, that's the market dynamics, and it's economically rational. Advertisers should, and do, spend up to the breakeven point (or the discounted lifetime value of the new customer acquisition). They do so because one should accept growing one's business by a marginal customer even if it's a breakeven proposition.

That's not to say that advertisers don't reap value from advertising, of course they do in reality (or at least, they always try to). I'm just pointing out that click auctions are expensive because the competition is willing to spend -- not because you're subsidizing spam.

On the other hand this huge "fraud" is what makes Google free and has thus powered the information revolution that now looks like a pivotal turning point in human history. Hard to be all that mad that Google made all the worlds information freely searchable by ripping off dumb businesses...

Hang on: ad placement is not based on page content but the viewers profile (constructed from search and browsing history, facilitated by tracking)

Right?

The reason a Chase ad ends up on 400K sites is not because all of those sites are contextually relevant, it's because of retargeting. A user goes to chase.com, gets pixeled, and then shown retargeting ads on what ever websites they happen to visit that are hooked up to the exchanges. This could be either a good or bad thing.

This is something the Ad Contrarian hammers on a lot (http://adcontrarian.blogspot.com/). There seems to a lot of fraud and BS in the system, but it's to the benefit of most players in the system to ignore that. Even for many ad buyers, it'd be embarrassing for the CMO to admit to having thrown away tons of money on something that doesn't work nearly as well as it's said to.

You'll find all sorts of personalization opportunities and you get to cut out the middle-man.

2 - The exchange would likely make more because the increased auction pressure from the same amount of advertisers trying to buy the same amount of volume (same demand) on a smaller list of websites (lower supply) would lead to higher prices being paid to show ads.

It could be that neither display strategy was providing any value.

My guess is that prices would go up dramatically if you started selling on CPA - and the ad tech business model would change significantly.

Seems like the best way to do this would be to hand curate targeted advertising and then charge a huge premium.

Even if Google did provide a provision for blocking ads for certain keywords (and they do have one), who controls that system? Amazon is also a bad example since many 3rd parties affiliated with - and driving traffic to - Amazon buy those keywords.

I think their is so much input for rtb that it is prime for ML. Most exchanges have some basic ml auto-optimization, but eventually someone will automate almost the whole thing.

This is absolutely understood, and in fact deliberate. Amazon does not want another advertiser sitting above their top position in the organics. It weakens mind share, and risks losing a click.

Ad exchanges usually charge a fixed rate on revenue (like 15% or 20%).

If the spend is the same, then the ad exchanges don't make any more money.

There are other middle men, like the audit companies and the impression counters, but they end up making less money because they usually charge by the volume of impressions: Since the same budget will be spent over a smaller number of sites, the prices for those sites will raise and the number of impressions will go down.

The only people who win are the publishers of those sites who now have more money for the same users.

This is just complete mis-management of paid media. Fire the agency or internal people for stupidity..

The problem is there is two types of ad. The first is "buy this now", if the user clicks and buys it was a success, but that doesn't mean a failure to click is a failure because there is a second type of ad that I believe is more important: awareness. You advertise so that people know you exist - that way when they realize their need for your product you come to mine. I think the pay per click model that the web advertising has settled on is wrong because it doesn't well account for ads that you don't want clicked.

There is a reason McDonald's advertises on every radio and tv station they can every few minutes: it works. Some of those ads go to people who will never eat at McDonald's and if they could figure out how to save money by not advertising to those people it would be worth it. Overall though the return on those adds it large.

It invalidates the data? How so? Or do you mean conclusion?

Even so, there's nothing inherent with "a few days" that makes the analysis invalid due to sample size; that depends on the nature of the data. In my employment we are measuring changes hour to hour.

Go back and read old articles though and holy heck they were good. I feel like an old friend is gone.