PAGE ZERO ADVISOR NEWSLETTER
[ Premium Subscriber Update ]
Contents © 2006 Andrew Goodman, Page Zero Media
-------------------------------------------------------------------------------------------------------------------
VOL. 5 NO. 4 | OCTOBER 2006
Google AdWords Relevancy
By Andrew Goodman
Just one article this month - on the relevancy
component of Google AdWords Quality Score. Keyword Relevance, Torso
Words, And Quality Score.
In Google's documentation -- you can look it up -- you can ascertain
that your ad rank on any given keyword is determined by your bid
multiplied by Quality Score. We've gone over the fact that Quality
Score replaces CTR (clickthrough rate) numerous times here, and
that according to Google, CTR is predominant in the formula to
determine ad quality.
We've also explained that historical CTR starts to take over once
you have an account history, and other "predictive" relevancy
factors take a relative back seat to historical CTR. Bottom line?
If you build up a consistent CTR of 7% over time, that's going
to give you a lot of leeway and a nice, high QS.
That being said, I also posited the idea that if certain factors
in your offer are "red flags," then you may find even
high-CTR keywords hampered with high minimum bids. In essence Google's
worst pet peeves have a certain veto power, formulaically speaking.
If they (or the algorithm intending to reflect editorial policies
and user and competitor complaints) don't like something about
your business, landing page, etc., then it can be a struggle. It's
like editorial oversight but less overt because they don't have
to tell you what's going on. Squeeze pages, scam offers, and arbitrage
may be in Google's sights, depending on unseen criteria. Clear
violations of user experience guidelines (popups, nondisclosure
of business information) may also factor into the equation.
Let's face it: I don't know the exact algorithm. I haven't seen
inside, and neither will you. But a useful way to get closer to
the truth about how predictive relevancy scoring is done - beyond
listening to the info Google explicitly provides -- is to look
at commonalities in what people are seeing, in what engineers at
both Google and Yahoo are saying about relevance, and common threads
that date back to strategies that always worked when the formula
was just a bit simpler.
Tight Loop Concept
First off, let's listen to what both Google and now Yahoo (for
their forthcoming Panama algorithm) are saying about keyword relevancy.
Essentially their algorithms will look at how well the keyword
relates to the ad, and how well these relate to the landing page.
Formerly, enforcement of that relevancy requirement was loose,
inconsistent, and cumbersome - Yahoo had editors who would "pick
at" your landing page relevancy and reject keywords. Now,
both Google and Yahoo are going to be looking -- algorithmically
wherever possible -- at the "fit" amongst all the elements
of your offer. More generic ads will tend to struggle to stay at
the top of the listings, when up against a really tightly targeted
ad and landing page. Guess what: that's been the reality with AdWords
for awhile, given how tight targeting and granularity improve CTR's
in most cases. The difference now is that data is not only collected
historically, it's used predictively to provide disincentives for
new advertisers coming in and just lazily throwing stuff in front
of the user's face, to the detriment of advertisers who are already
there. That advertiser who is still in dark ages and thinks they
can turn this into a "big important media buy" will have
much worse ROI than the ones who believe in the concept of tight
targeting and iterative campaign growth and testing.
Old News - The Poor-CTR Penalty
When Google has built up a hefty file on you getting poor CTR's
for a given keyword, your QS might really get hammered. The method
of deactivating you is a bit different from the old way. Your minimum
bid gets set to $1.00 or $5.00, effectively deactivating your keyword
depending on what you're willing to bid. This low-CTR penalty is
very similar to the way it was previously, and it still makes a
lot of sense all around. Let's take an example that's disguised
slightly to protect the innocent (sorry if that makes it slightly
nonsensical). The broad match for the word "repair" would
not work very well in a lot of contexts. Let's say it has worked
well for a particular campaign because of the relevancy of the
ad and the content it led to. By contrast, the plural word "repairs" gets
slapped with a $5.00 minimum bid, and you can't figure out why.
Isn't that keyword just about as relevant? No -- as it turns out,
it's getting a click only about every 800 impressions! Let's come
up with a potential reason for that: ambiguity. The word repairs
could overlap with many very irrelevant queries -- perhaps bridge
repairs, road repairs or something else -- research or news driven
searches. Maybe it's just a quirk that "repair" did OK.
Now, how to work around the dead "repairs" word? Well,
a couple of points. One, do what I've been counseling for some
time. Go for the "torso" of semi-common queries by loading
up an ad group with two-word broad matches that include the word
repairs (as long as they're relevant). Drain repairs, plumbing
repairs, floor repairs, etc. I know some of these phrases may sound
nonsensical but with expanded broad match you will get some relevant
traffic out of a group of 200 or so two-word phrases like this.
And I guarantee you -- Google's projected relevancy for most of
these terms is likely to be considerably better than for the one-word
broad match. Another thing to keep in mind is that if you're using
torso terms (two-word broad matches, let's say) for either the
singular or plural version of "repair," Google's expanded
broad matching might show your ad against any user query that is
similar enough to your phrase to warrant an impression.
A lot of people use a different approach to this, sticking with
the one-word broad match and adding a lot of negative terms. This
should be almost as good, but if you didn't do it in the beginning,
you may be stuck bidding the high amount to get your terms back.
So use broad terms sparingly in the early going UNLESS you pay
close attention to negatives.
New Problem - Classes of Keyword They Don't Like
It seems that big brand keywords and individuals' names are among
keyword classes that have poor predicted relevancy and are hard
to bid on in new campaigns. Combine that factor with the lack of
granularity (bid on just Best Buy or Mcdonalds?) and you're looking
at $1.00 or $5.00 minimum bids. Don't lose heart completely, though.
Granular approaches here might well work, if there's some honest
user response to be expected. For example, I'm asked to bid $5.00
on the term Home Depot. But they let me bid $0.50 on Home Depot
Canada. And Home Depot Toronto has a minimum bid of $0.07. As it
turns out, the Home Depot Canada projection by Google was even
a little bit lenient. The ad I ran got very low CTR's on that term,
and so eventually the QS will get worse, and that will be bumped
up to $5.00. The "local" term, however, is doing quite
well, and I want to add, it converts well too. That's because,
on the site in question, consumers are actually writing reviews
of local Home Depot stores. Arguably, placing the same ads against
generic terms would have been annoying to consumers, and deceptive
enough to cause the Home Depots of the world to rage against Google,
legally or rhetorically. So it all sort of makes sense.
Obviously, you are probably not wise to bid on the word "Kenwood" --
but "Kenwood Vacuum Cleaner" has a clear business purpose
to it and a clear user intent. As a retailer, you have every right
to use that phrase in many cases, but perhaps more importantly,
Google expects user response and thus user satisfaction to be higher
than if your ad showed up next to something like "kenwood
lawsuit," "kenwood kentucky restaurants," or whatever.
One-word broad matches are high-risk plays for Google so they want
to get rid of that risk by making you pay every penny of their
risk for showing overly generic ads, and then some.
The watchword, still, is GRANULARITY. A well-organized campaign
that takes users to relevant offers still works. If that sounds
like a bit of a cliche, consider trying to build out your "torso
words" (between the head and the tail) of two-word broad matches
(long lists of same) and see if you feel like that's a cliche or
whether there isn't more room for improvement in your campaign.
Then, try ten intelligently-targeted landing pages where you used
to have one. Not only is it going to give you better economics
by way of better quality scores, over time, but your conversion
rates should rise too.
To really "grok" this, probably the best book you can
read is "Waiting for Your Cat to Bark," by Bryan and
Jeff Eisenberg. Or listen to Bryan talk.
Every time I start to whine a bit about high minimum bids on keywords...
or at least half the time, anyway, I have to remind myself: "dude,
admit it -- you're bidding on irrelevant keywords!" Yeah,
it's cut down on a few tricks, but we can replace much of that
inventory with more diligent pursuit of "torso words."
As a search engine user, do you really want the engine to leave
incentives around for clever advertisers like you and me to bottom-fish
for five-cent clicks on irrelevant ads? For every one user who
stumbles on something they'd like, there are 5,000 who would get
annoyed.
And now, Yahoo has come to fully understand this. It has *vastly*
changed how they will operate their ad auction. New advertisers
will be migrating onto the new Panama system -- much easier to
use -- for the remainder of 2006. But the "marketplace design" (the
new auction format) won't be live until approximately March, when
the last advertiser is migrated over. As a Yahoo advertiser, you'll
have to be ready for this completely new environment. My colleague,
Mona, author of the Yahoo Search Marketing Handbook, is now working
on a completely new version of the Handbook, so you'll be ready.
The nice thing about Panama is, well, it works a lot like AdWords.
It supports the granularity you require by using the modern format
that includes campaigns, and ad groups, and lets you test multiple
ads against those ad groups.
Until next time!
Andrew
P.S. Did you check out the free audio roundtable following up
on my trip to Burbank to preview the new YSM? I interview Barry,
Mona, Catherine, and Jen about their impressions from the two days
we spent at Yahoo HQ. |
