July 15th, 2007
The Poor State Of Online Display Advertising Relevancy
I was on the New York Times site just now and was a bit freaked out when I glanced at the adjacent ad:
I live in Leesburg, and the ad knows this. What struck me is that I was “freaked out” even though I understand the geo-targeting technology that makes this kind of customized ad possible — it still felt like an invasion of privacy. I can only imagine how someone who doesn’t know anything about what’s going on behind the scenes must feel. Last year I researched behavioral targeting online, and ads from Tacoda and Revenue Science pursued me in completely irrelevant contexts for weeks.
It’s easy to find out who is responsible for this ad — just watch the browser status bar while the page is loading:
UDPATE: Well, I’ve learned it’s not so easy to figure out who is responsible for an ad — the folks from Tacoda told me that the dating ad is absolutely NOT from Tacoda’s ad network. The domain below that I saw in the my browser status bar was from Tacado’s audience insight tags, i.e. Tacoda collecting data to be used for behavioral ad targeting elsewhere, so Tacoda was NOT serving any of the low relevancy ads on these NYTimes.com pages.
What stuck me most is how poorly customized the ad actually is — the Tacoda ad server knows my network location, but it doesn’t know the crucial piece of data that would have informed the delivery of this ad — I’m happily married and have no interest in meeting singles. So while the ad got the geography right, it’s still utterly irrelevant.
I refreshed the page a few times, and here are some other ads that appeared:
I may be traveling in the near future, but probably not to any of those generic destinations. And I already graduated from college, thanks very much — crawling my LinkedIn profile could have squeezed out that data point.
Yahoo’s new SmartAd platform promises to cobble together more bits of information to produce customized display ads — it won’t be hard to improve on the current sad state of affairs.
Facebook is currently the king of irrelevant display ads — despite all the time I’ve spent on Facebook recently, I don’t think I’ve seen a single relevant ad. Let’s try a quick visit to Facebook and see what I get:
Come on guys, my birthday is my profile — you’ve got the data. Use it!
One additional note — it appears that NYTimes.com isn’t doing so well selling its blog inventory, since it appears to have Tacoda’s an ad network plugged into all of the blog pages, even for targeted niche topics like technology and fashion. That would seem more of a sales force failure than a lack of valuable inventory for direct ad sales.
After more than a decade of online advertising, you’d think we’d be farther along.
Powered by 
Design by 
poor state of online ad relevancy
[IMG]Following up on my post about the poor state of online ad relevancy, I’ve been digging deeper with some of the companies that are working to improve that state, which lead to the following interview with Larry Allen, SVP, Business Development and Marketing at TACODA. Believe it or not, we completed this interview
display advertising continues to lack relevancy
Scott Karp has an interesting post on his Publishing 2.0 blog today about how display advertising continues to lack relevancy over 10 years into it’s life. Being in the display advertising business, you’d think I’d argue, but I won’t. He’s right, display advertising is still largely way too irrelevant and there is still a ton of improvement that can be done to help
Interesting article on that by Scott Karp today
I recently set up a site and included an Amazon affiliate ad that shows stuff related to what I’ve been interested in at Amazon recently.
It’s weird cause I knew what it was, I put it up but it still felt invasive.
At least one of the ads (online degree) in the NYTimes.com rotation you used as an example appears to be “remnant-grade.” It isn’t even an attempt to target.
If it came out of Tacoda’s network, and not from a rotation between Tacoda and the NYT ad server, that would suggest Tacoda’s network isn’t exactly sold out of targetable campaigns.
Now, I know I’m connecting the dots from speculation to speculation, which is risky, but this would seem to back up the point that behavior targeting needs both very consistent means for applying targeting rules AND truly massive scale. If mighty NYTimes.com hasn’t the scale nor the addressable user profiles to do it, what hope is there for smaller online services?
Jay,
The problem with behavioral targeting is that no matter how large the network, the web is SO large that it’s easy for me to come to NYTimes.com without having visited any other Tacoda network sites, thus giving Tacoda NO data to use for targeting ads to me.
In such an instances, NYTimes.com might be better off running house ads than these crumby network ads and annoying me as a user — the CPM is so low, they might just hold the line and say we’re not going to run anything here unless we can sell it and make it worthy of the user experience.
[…] não elucidam como deviam (a maioria não faz a mínima ideia, suspeito), podiam começar por ler The Poor State Of Online Display Advertising Relevancy. De que serve comprar 10.000 milhões de pageviews, endereços de e-mail, ou perfis no Hi5, se não […]
In the Google model, advertising follows editorial. If I’m searching for “lawn fertilizer” or reading a gardening site, I’ll probably get a simple but fairly relevant text ad from Adwords or Adsense. Most ads are unviewed, uninteresting and unclicked. But the clicks add up to +$10B for Google.
Large media sites can’t live on the the CPMs this haphazard model produces. The banner ads on most content sites are uniformily uninteresting, irrelevant or worse, not there. I popped over to CNET. Found an article on “auto tech”. No ads the page. This would have been a perfect place for BMW or Sirius or LoJack to place a banner. Instead nothing.
The solution is for CNET to have a preexisting relationship with BMW et. al to allow them to place their ad next to a story when deemed appropriate by the ad salesman/engineer. Compensation would be based on CTR not CPM.
Naturally if this model is embraced by the NYTs and CNETs of the world, we’ll see more advertising-friendly editorial. Well duh. If subscription revenues disappear as they will, content will follow advertising.
Contextual targeting, geo-targeting, behaviorial targeting… s0 many people think it’s “the next best thing,” but as it backfires more and more sites could lose visitors and ad dollars. Publishers want this technology because CPMs, CPWhatevers can be increased.
http://www.businesscommonsense.com/story/story.bsp?sid=56176&var=story
[…] Scott Karp has an interesting post today on the Publishing 2.0 blog about the state of relevancy (or lack there of) when it comes to display advertising. […]
The “Leesburg” dating ad looks to be geo-targeted but will not be targeted on any “behavioural” activity. They are reckoning that if they at least target it geographically they have a better chance of getting a response - even if it goes to everyone not just the 20-30% who are singles. I agree this is still a pretty poor job - c. 70-80% wastage.
As other have stated - the rest in this series of ads looks to be “distressed” or “remnant” ads. This could come from the lower-CPM end of the networks such but it looks to be pretty low-grade stuff.
As you point out - it is striking that no one that I know of in the Ad Management / Ad Serving / Ad Network space has successfully hooked up ad serving technology to user data held by the publishers/site owners. I guess this come next. Given the lack of usability/functionality with most ad management software this is hardly surprising.
Scott, The technology is far from perfect, but by condemning behavioral targeting in this specific case you are perhaps painting the canvas with too broad a brushstroke. More likely it’s a combination of an even less perfect ad server deciding to hand off an impression to a network that hands it off to another and another, ad nauseam. Your geo data is in your user registration profile, and not necessarily in your behavioral profile. As you frequently and eloquently point out, there is much room for improvement in the entire online ad ecosystem. While I am biased (3 years at Tacoda prior to working with multiple BT providers on the Right Media Exchange), I believe - and have seen numerous times - that this form of targeting can greatly enhance relevance. Let’s not throw the baby out with the bath water ….
Bennett,
I mistakenly attributed this to BT after seeing the Tacoda domain in my status bar, which turned out to be audience data collection not ad serving on the page. I corrected that in an update to the post.
Lifting it above the level of who is responsible, it seems unbelievable at this point for a site like NYTimes.com to be serving up these kinds of ads — I’m a (more or less) desirable advertising target in a high-end context. Why should the NYTimes.com be reduced to serving junk? I realize that “junk” pays more than house ads, but as a user, I have higher expectations.
Well I got Hoodia diet patch. NYT does not know my age so its not guessing BabyBoomerBelly.
All this guessing based on click-stream, age, zip seems to be a game of diminishing returns. By the time the media firm figures out who you are - by (unknown to readers) amassing tons of data and slicing it through increasingly sophisticated software - readers will get spooked by the results.
Search does reveal intention and that can be monetized and nobody gets pissed.
[…] up on my post about the poor state of online ad relevancy, I’ve been digging deeper with some of the companies that are working to improve that state, […]