Friday, November 21, 2008
I was just going through a White Paper by an agency on the subject of paid search quality scores. I hate to be picky, but if your intent is to explain keyword auctions to your clients, then factual errors don't really sit well with me. In outlining the history, this paper essentially said that Google invented Quality Score to deal with relevancy issues in 2005, and prior to that everything was a pure auction a la GoTo.
As most of you know, CTR has been baked into Google's CPC search advertising auction since 2002!
Another problem with their very brief "analysis" was failing to explain the separate roles of keyword relevancy related quality, and landing page quality. It's a cool factoid to tell people that landing page load times matter (rarely, if they're really poor like above 12 seconds), but not if that gets people optimizing the heck out of already qualified, relevant landing pages, and misinterpreting the weighting of these factors in the ranking system. Google doesn't tell us the weightings, but anyone who's worked on a few dozen accounts (or actually paid attention to those accounts) knows that landing page issues create poor quality in rare cases, but "optimizing" an already good landing page is something you do, as the fitness magazines say, "for you" (for your own conversion rates)... the fun you have with landing pages shouldn't generally impact your ad rank too much as long as the page in question doesn't suck, isn't evil, etc.
So, this big agency's tiny "white paper" looks authoritative, but explains nothing in the end. Amazingly, one of this huge agency's specialties is buying clicks for large company clients.
But agencies aren't the only ones who fudge the facts on ad quality. In 2007, I saw a major international Google office once (no not Canada!) give a presentation that referred to the 2002 version of the formula. Guess they figured the ordinary schmoes "can't handle the truth"! (There may be something to that.)
At the end of the day, all of this fudging, and oversimplifying, and drawing of cartoons to show how easy it is to run an AdWords campaign results in rampant confusion and second-guessing. I talked with a prospect whose company has now been advertising on Google for four years. I noted the high costs on certain branded keywords, in part (I thought) due to certain relevancy and quality issues baked into the algorithm. I also noted that these were unjustified in the sense that the auction on those words wasn't particularly competitive; and moreover, the algorithm wasn't doing a great job on many of them -- there were wrong ads and bad ads from low quality competitors appearing against some of those terms, tickety-boo.
"But I thought it was a pure auction!," said the client side dude.
What? A pure auction? Not since 2002. All the cartoons and 138-word "white papers" in the world can't boil down the auction process into something super-simple. But let's go with this much: CTR is still super-important, and then there are a list of exceptions and quirks you should familiarize yourself with. And for new accounts, experts like David Szetela (I concur with his take) get into some of the subtleties of the ideal way to build out for maximum Quality Score friendliness.
Labels: ad quality, quality score
Tuesday, March 18, 2008
I had the pleasure of moderating the panel on Ads in a Quality Score World at SES New York today. Along with two advertiser-side speakers (Joel Lapp and Jon Kelly), Frederick Vallaeys of Google and David Miller of Yahoo weighed in.
Frederick pointed out that very long phrases and very low volume keywords well down the long tail are not necessarily an advantage to a marketer, as they don't reflect how "real users" normally search. The sweet spot of the long tail is 2-to-4-word phrase. 5-8 word phrases, not so much. Among other things, Google will have such limited data on these, they have no choice but to assign slightly worse quality scores to them.
I did notice this difference in one of our new accounts today. In a group of 2-4-word phrases, most started with initially Great or OK quality scores, but some 5-word phrases were marginally Poor. Also, there were subtle differences in meaning between some 4-word phrases and others. For example (fictitious example), "martian loyalty points offer" was OK, whereas "martian loyalty points program" was marginally Poor. Although the difference in meaning was not enough to deactivate either phrase, the latter may well have less predicted relevance to our offer, because it's more generic. Rather than an ecommerce transaction from a reseller, that user might be looking for an information page from the official Martian Loyalty website.
Basically, then, marketers need to stop asking the question "to long tail or not to long tail," but rather, even within the long tail, they should consider whether they're just getting too fine, or dumping too many irrelevant phrases into otherwise functional ad groups.
This may dovetail with a point made by David Miller of Yahoo Search Marketing. In his approach to explaining a successful campaign structure, he referred a couple of times to "thinking at the ad group level." Although in a formal sense keywords are evaluated for quality individually, we know that the whole picture matters. So at YSM, if you have a high-impression, loosely-relevant keyword generating a lot of the clicks or impressions of the ads in a given ad group, its lack of targeting could be "dragging down" the quality index for the whole group.
Although the specifics are bound to change, the concept of tight targeting, and making advertisers pay a premium for experiments with loose targeting, appears to be here to stay. There is nothing that says that long tail keywords are always particularly well targeted to a given ad or landing page. Often, they are detritus mucking up the rest of the group - and potentially, your forecasting and tracking efforts.
Related: Keyword Intent: Tidy Campaigns Avoid the Dump and Chase
Labels: ad quality
Tuesday, February 26, 2008
I just learned that Yahoo Search Marketing is introducing reserve pricing in their keyword auction. Although the Panama system already included a quality index, this was for ranking purposes only. Now, similar to Google's auction, Yahoo can raise minimum bids to dissuade low-quality advertisers further. It can also, presumably (like Google) use this system to optimize (i.e. increase) its own revenues, concurrent with improving the user experience. Win-win, n'est-ce pas?
In light of these sophisticated systems, talk of a fresh comScore report that shows Google's total US paid clicks have leveled off seems like just noise, although the stock market seems to take it quite seriously. When you sharply raise minimum bids on many keyword buys, as part of a quality initiative, it only makes sense that click volume will take a hit. But will Google's short and long term profit benefit? That's the real question.
Labels: ad quality
Wednesday, October 03, 2007
Followup to my recent article on Google AdWords' website quality policies. Although the majority of rank and file advertisers I chatted with favor Google's stances against, for example, "arbitrage sites that are designed for the sole purpose of showing ads," not all go along with the Google take on things.
One respondent, CEO of a midsized technology company, missed my deadline but took the trouble to call and leave a detailed voice message. His position explores the case for being "pro-arbitrage," on several counts:
- It's worrisome that our rights as advertisers to try different business models can shrink not because Google cares about users, but because Google is acting anti-competitively. What can be an official curb on "sites that are designed for the sole purpose of showing ads" today could in future bleed into banning "sites that show ads that Google just doesn't like, or competes with."
- So-called arbitrage sites are at the leading edge of user testing. Often, they convert better to a sale than so-called high quality sites, albeit requiring an extra click. In essence, arbitrage sites are the purest form of exploiting inefficiencies in the worth of media exposure. Remove this from the equation, and only less efficient forms of exploitation remain in the mix. This potentially weakens the rest of the herd as it is now being helped by enforcement as opposed to economic superiority.
The solution proposed by the observer taking the pro-arbitrage position? I'm not sure. It seems like more of a general reminder that Google's positions can be one-sided, and that they only selectively protect users from negative experiences.
- Google directly benefits from the ads showing on things like parked domains, many of whom show nothing but ad links. So Google listens to user complaints about being directed to such sites from a paid ad, but then again, they aren't above directly earning revenue from such sites through partnerships (DomainSense). It's a question of mixed messages, and also a holier-than-thou message in the sense that companies other than Google, who like Google profit from sites that pretty much just show ad links, will be hurt by the negative rhetoric surrounding "arbitrage" while Google, in fact, continues to earn revenue from stumble-in traffic to sites that look just like the ones they are supposedly protecting us from.
Labels: ad quality, click arbitrage
Thursday, August 16, 2007
...but it helps!
That's the t-shirt many paid search auction-grapplers have been mentally wearing for years.
In a few weeks, Google will be going live with a slightly altered version of the AdWords auction. Group Business Manager for Ads Quality, Nick Fox, was kind enough to go into considerable detail outlining the change, which as usual has been misinterpreted somewhat here in the ventosphere. The most common dismissal of any change Google has made to AdWords is "another cash grab by Google"... which I guess looks like a clever dig in pixel form, but I think the changes are definitely more interesting than that implies. Many of those comments tend to emanate from sources who still think all clicks should be free, so take them for what they're worth.
1. The Simple Explanation... Past vs. Present
a. OK, so the explanation isn't that simple if you've been doing this for awhile but not paying much attention, because Google has changed the way they sell the top-position "premium" ad slots several times in the past few years. Several years ago, the first major change occurred when Google started, then stopped, selling those ad positions completely separately on a costly high-CPM basis, using a dedicated sales team, 6-month insertion orders, and the whole nine yards. That was convoluted to administer -- and more importantly, divorced that type of ad buy from Google's traditional AdWords formula that focused on ad relevance and quality -- so they discontinued the practice. The top slots more or less joined the regular auction, but they remained distinct in look and feel, and were great to achieve due to high CTR's, high volume, and sometimes, higher conversion rates.
b. However, Google then began to experiment with how many ads it would show in those positions, and by and large, placed a higher threshold of quality [prior to August 2005, CTR (click-through rate) was the only measure of quality] and CPC (cost-per-click) before it would elevate an ad to the premium area. The exact thresholds were never publicized, but typically, ads with actual CPC's below .75 had trouble making the grade for premium positioning. Ads with CTR's below the averages for the keyword or industry sector had trouble making the premium spots too. I'm guessing Google tweaked the formula several times.
As Nick Fox told me today, because (1) "these ads get more attention from users," and (2) "push organic results lower on the page," to say nothing of (3) "being particularly valuable for many advertisers," it's important for Google to find the right balance here. Above all, the user response to the ads is paramount. Users must not become turned off by the overall perception of quality in their day-to-day use of the search engine.
So until now, the function that determined whether an ad would show in a premium position or be placed in the right-hand margin was weighted heavily towards quality score, but also included actual cost-per-click. That made it different from the rest of the auction, which went on the basis of quality score and max bid. Max bid can often be much higher than your actual CPC because of the discounter built into the pricing method.
c. Now, there will be a small change to the formula. The downside of the "actual CPC" part of this was that it stopped certain advertisers from getting in premium slot due to a lack of what Nick Fox called "auction pressure." Other advertisers simply might not have been present or bidding high enough to push the actual CPC up enough for the leading advertisers to qualify for premium spots. Worse, I'm guessing that this might mean that slightly lower-quality ads could creep into premium positions, because they *did* pay more on an actual basis. They wouldn't be *poor* quality ads, but they might not be the highest quality. This actually reveals a flaw in the previous formula because it unwittingly relaxed quality standards slightly in the area where it should have been heightened.
My interpretation here is that advertiser control and overall user satisfaction rise with this change, but yes, it does seem that some advertisers might experience price increases.
2. Unforeseen complexity.
Unpredictable effects might occur to some advertisers who find they suddenly get promoted to premium position, or who have been bidding very high and changes in the auction dynamic cause other advertisers' positions in the auction to change.
Will you pay more in those cases? Probably, if your actual CPC has in the past been coming in below what the new minimum price for premium placement is set at.
What, did I just say "new minimum price for premium placement?" I don't want to put words in anyone's mouth. There is no set reserve price that Google is implementing here, but rather, a dynamic minimum that is a little bid difficult to grasp at this stage.
I guess the throbbing feeling I have in my brain about now comes from the fact that by adding this wrinkle, Google appears to be adding a second "minimum bid" - I'd liken it to a larger rung partway down the auction ladder. If you qualify to be in the premium slots (let's call this the treehouse), then you can stay up there, but you'll potentially pay more. The rope ladder down to the ground has the usual auction dynamic, but if you fall below that familiar minimum bid, the bottom rung on the rope ladder breaks and you fall into a big pile of leaves on the ground (out of the auction entirely). Potentially, your best friend's dog, Scrappy, licks your face as consolation, while you muster the courage to "improve quality or bid higher."
The minimum bid to be on the rope ladder (active in the auction at all) is *published* right next to the keywords in your AdWords account. But it doesn't look like your price of admission to the treehouse (the premium positions) is going to be published. One minimum bid, disclosed; the other, not. And a real dilemma for Google as to how they could possibly publish that given the limitations of the AdWords user interface as complicated as it now is.
So while the change may be minor in practice, advertisers are counseled to watch their bids, particularly if they've mucked about with high maximum CPC's (aka bids) in low actual CPC zones. You might begin paying more than that low level you've been accustomed to, in those cases. You should always be prepared to pay the full amount of the max bids you set, so it's never a good idea to bid wildly high, counting on a lack of auction pressure to make up for your laziness in bid management.
Let's ditch the passive voice, just to be clear. I counsel you to watch your bids!
Labels: ad quality, google adwords, nick fox
View Posts by Category