Forum Moderators: martinibuster
So what ads showed up with AdSense? Nail fungus, nail fungus, nail fungus, and -- yep -- nail fungus (four in a row on a leaderboard). All of this because ONE person commented about being annoyed with getting dirt under her nails while camping.
This sort of poor targeting helps no one. I can't imagine that I'll get ANY clicks on that page, so Google loses, my visitors lose out, and certainly I get less revenue. In contrast, if Google had smartly shown camping-related ads (REI, camping equipment, campground reservations sites, etc.), I bet things would have been pretty hunky-dory.
* * *
This example strongly highlights why I hope that Google will eventually support what I call "Webmaster Assisted Targeting" -- or "WAT" for short.
I understand that there's room for abuse. In theory (and without good controls), someone could take a page with little content and tweak things to have 'lucrative' ads shown.
That's why I'd like your input on two critical issues:
1) Do you agree or disagree with the idea of WAT?
2) How might it best be controlled?
My stance is clear on the first question, and regarding the idea of controls, here's my idea:
Google likely has a handful of ideas of what any particular page is about. On this camping thread, for instance, Google may have had to weight three different possibilities... say, nail fungus medications, sleep aids, and general outdoor gear / camping stuff. As a Webmaster, I'd like the option to SUGGEST a particular focus, as in:
<meta adsense-focus = "camping, hiking, outdoors">
Of course, on a forum page, it's likely not easy to add a meta tag, but perhaps WAT could be tested initially with more static pages.
With my suggestion above, Webmasters wouldn't be able to fully dictate what ads showed up on any given page, but they'd be able to guide the AdSense algorithm to favor one possibility over another one. In this case, for instance, while 'nail fungus' ads may have initially seemed more lucrative to Google's automated system ("Hmm... nail fungus ads generally have a higher CTR and/or CPC!"), WAT would push Google to consider a different weighting.
I look forward to reading your thoughts, keeping in mind the critical goals of serving publishers, advertisers, and -- most importantly -- users.
Frankly, with the kind of sites I have been seeing lately with Adsense this would be exploited ad naseum.
Good idea, but I don't see Google implementing it as they are more willing to rely on technology than people, and this would require more people.
Google could go one step further by penalizing misuse of the helper keywords. Let's say a publisher has an e-commerce page on a low-bid topic such as toilet plungers and uses helper keywords like "Viagra" and "Mesothelioma" in hopes of getting high-EPC ads. The Mediabot would analyze the page, determine that the text has nothing to do with Viagra or Mesothelioma, and display PSAs instead of paying ads. Or it could automatically deduct a percentage of each click, thereby lowering the publisher's overall EPC and earnings.
Think about it, if the system currently thinks hair growth ads are well targeted for my classic car site (it does), why wouldn't it think Vi*gra ads were also a good fit?
I'd be more inclined to support negative keywords if the allotment was reasonable, and per page, or at least per channel.
Much as AdSense team members police sites to make sure there are no TOS infractions, I think it'd be easy for them to spot-check and make sure folks aren't abusing WAT. One way in which they could clearly combat potential fraud would be to have a VERY clearly articulated "one strike" rule; if any Webmaster is caught clearly attempting to mislead the mediabot, then their account would be cancelled, and all unpaid revenues would be forfeited. And, on a less-severe note, if the mediabot repeatedly notes mild-to-moderate WAT/media-bot-determined targeting discrepancies, the account could be flagged and -- if appropriate -- simply have its WAT privileges disabled.
And again, I'd like to emphasize that I envision WAT as a GUIDANCE, not determination system. I'm under the assumption that the mediabot already must decide from amongst a set of possible targetings per page, and I see WAT as a way of helping the mediabot decide... not telling the mediabot "Well, I recommend 'D' even though you see the possibilities of 'A,' 'B' and 'C'."
In other words, like in my camping example above, if someone attempted to use 'viagra' as a meta-adsense keyword... the bot would look at the content of the page and say (er, in botspeak) "uh, bub, I don't think so! Try this one more time and I'll flag your account to have a rep look over it with a fine tooth comb."
Frankly, with such safeguards (and threats) in place, I see WAT being much more of a benefit than a risk to Google, its advertisers, and its users.
Oh, and one other thing: This could be a very cool way for AdSense to learn of existing targeting problems. If they repeatedly are made aware of certain accounts that experience bad targeting (hair ads on a car site, for instance!), they may then be able to take that info in the aggregate and improve the algorithms.
I KNOW that Google's aim overall is to automate processes intelligently. I see WAT as not only something to help folks in the short term, but also a 'reporting' mechanism of sort to help flag recurrent targeting problems, alert Google to problems, and provide a better foundation for them to address the issues.
And again, I'd like to emphasize that I envision WAT as a GUIDANCE, not determination system. I'm under the assumption that the mediabot already must decide from amongst a set of possible targetings per page, and I see WAT as a way of helping the mediabot decide.
If done right it would help adsense target better. Most of my pages come out targeted pretty well but I just added adsense to a small site and there are two pages where the targeting is ridiculous. One is about decorating your home and including good light both natural and artificial. It would do great with any kinds of home decorating ads, even lamps and such. but instead it has all lightbulb ads including bulbs for projectors. The other page is about learning at home and it serves mostly ads about knowledge base software. I did some searching and there appears to be plenty of AdWords on home study. I haven't put alternative ads on these pages as sometimes the matching improves after a week or so. If it continues I'll have to remove AdSense but it's frustrating when I know there are good AdWord matches out there.
I'd be more inclined to support negative keywords
That would help a little, like I could stop "knowledge base" and "light bulbs" but it still may not bring in better matched ads. I know if I write to Google they will just say to filter out these ads but the same sort of ads could keep coming on and on and I am not about to dedicate a great deal of my filter quota to a couple of individual pages.
This connects to a discussion elswhere where someone asked if excluding advertiser URLs could generate more money (as opposed to just using it to screen out competitors). The answer, of course, is yes, and the same goes for excluding keywords. Even granted that CPC can't rise if you exclude advertisers or keywords, the camping page would be better off with $.05 camping ads than $1 nail fungus ads.
Incidentally, I have a similar problem with some "forum" pages of mine. The pages start with some general blather about wanting to get a good conversation going, and then procede on to a specialized academic topic. But I get ads about conversations, chatting, etc. instead of ads about the topic in question.
If AdSense stopped working well enough to make sufficient profits, I suspect they would jettison it and do something else before enhancing it with features that require human interaction.
We know that WAT (to use thatadamguy's acronym) is open to massive abuse and that AdSense would want to be able to deal with such abuse in a scalable manner.
We also know that robots and algorithms sometimes get things wrong... which might have scary implications if something similar to europeforvisitor's suggestion were adopted and webmasters were financially penalised if the mediabot concluded that abuse was occurring when in fact it wasn't.
What we need here is not so much for webmasters to be able to tell AdSense what their page is about, but rather to tell AdSense what their page is NOT about (exactly as suidas said).
So why not have a webmaster login section on the page that comes up when you click on the "Adverts by Google" link which then takes us through to a very short page which says something like:
"Google believes this page is about:"
and then gives you three or four checkboxes, all ticked, with a topic name next to each.
You could untick any or some of the boxes which might not have any immediate effect, but it might give mediabot some more info to work on the next time it comes around and visits the page.
"Google believes this page is about:"
Well, Google doesn't get it wrong across thousands of pages - it only mismatches adverts once in a while.
Exactly. And for short-lived pages (such as news stories), mismatches may not be worth fixing. But on evergreen pages that get traffic day in and day out, being able to prevent those "once in a whiles" could result in a nice chunk of extra revenue for the publisher and Google over the course of a year.
and junior scammaster will sit there and hit NO NO No NO No NO until mesothelioma shows up.... even if it takes all day.
Junior Scammaster won't have to hit "NO NO NO NO" all day, because he can save a lot of time by simply creating a page about mesothelioma.
Fact is, Google does need help. Take a guide to cruising in singles bars: I don't know what the page should be advertising (a book titled "100 best pickup lines"? condoms?), but ads for Carnival Cruises or Cunard would definitely be off-topic. Until AdSense ads are truly contextual (not just related to keywords), "helper keywords" (whether positive, negative, or both) are the best way to prevent chronic mismatches on evergreen pages.
<meta adsense-focus = "camping, hiking, outdoors">
I definitely agree. WAT would be a meta data classification scheme similar to what is successfully employed by <snip> where statistics show that 99 % of all webmaster classifications are good or acceptable. A webmaster suggested AdSense classification may also be complemented with other AdSense AI targeting algorithms to verify the classification, but by trusting the webmaster with the initial classification a lot of ground can be covered.
AdSense targeting has degraded significantly on the sites where I use it and the current choice of algorithms is clearly not optimized for all kinds of web sites. Maybe a webmaster driven classification can be one component in addressing this problem.
[edited by: Jenstar at 6:00 am (utc) on June 4, 2004]
[edit reason] NO URLS as per TOS, please! [/edit]
"Adsense News: What's WAT?"
and "WAT's UP"
and so on :D
and junior scammaster will sit there and hit NO NO No NO No NO until mesothelioma shows up.... even if it takes all day.
Ah. That's not quite how I envisaged it.
What I envisaged was that if you untick (say) all four boxes, you get PSAs.
Like I said, AdSense may get it slightly wrong sometimes, but if it believes your page is about one or two of possibly four things and you say it's not about any of those four, the chances are you're making things up.
Re the concept of WAT ... sounds good, but not the easiest way for webmasters of multi-topic sites. If your site covers 10 major topics, with 20 sub-topics each -- all with their own set of advertisers -- it will be such a pain to set "suggested topics" for your 2000+ pages.
It will be interesting to see if G will allow webmasters to suggest keywords. I think they are more inclined to rely on their algo and simply work to refine their targeting algo. If and when they allow WAT, I think they would implement it alongside penalties for "over optimization" -- a site abusing this feature will get PSAs and no targeted ads.
Re the concept of WAT ... sounds good, but not the easiest way for webmasters of multi-topic sites. If your site covers 10 major topics, with 20 sub-topics each -- all with their own set of advertisers -- it will be such a pain to set "suggested topics" for your 2000+ pages.
That wouldn't be necessary in most cases, because mistargeting is generally confined to a few pages. WAT would be useful mainly for publishers of "evergreen" content who became aware of consistent targeting problems on some pages (e.g., B2B ads for ATM equipment and supplies on a consumer article about using ATMs in Europe).
I think they [Google] are more inclined to rely on their algo and simply work to refine their targeting algo.
IMHO, it probably isn't realistic for Google to expend the kind of computing power that would be required to perform a detailed contextual analysis (including identification of the target audience, as in my ATM example) for every page on the Web that has an "Ads by Google" box. If, say, only 40 out of my roughly 4,000 pages are showing blatantly mistargeted ads, is it worth performing a detailed contextual analysis on the other 3,960 pages? Probably not.
But if Google were to see "meta=wat" statements on the pages that were having problems, it could use those helper keywords to prevent mistargeting. Over time, the addititional revenue from properly targeted ads could add up to a respectable sum for both Google and the publisher.
... except that if and when this feature is introduced, I think webmasters would want to sufficiently cover their bases and make sure that every page has helper keywords. We know that a page can show targeted ads today, and fail the next. Rather than risk getting far-from-targeted ads, I would surmise that webmasters would use this feature throughout their site -- probably beginning with the most worrisome pages -- if only to alleviate the concern that the pages may begin to show mistargeted ads.
It would be interesting to see how G will implement this, if at all.
