PPC-Man Drowning… Too… Many… Keywords…

The keyword long tail is dead. Long live advertising.

The first shot (heard ’round most of the search marketing world) in this war against busywork was actually fired in 2002 by Google, when they introduced powerful matching options for AdWords. Too bad some folks haven’t yet heard that shot.

This may be an exaggeration for effect. But if you’re like me, an experienced paid search pilot with somewhere close to your 10,000 hours in flight time, you’ve probably had enough of bloated ad groups containing hundreds of pointless, longshot, low-to-no-volume keywords.

Some experts in the field are suddenly joining me, jumping on this Long Tail Keyword List Critique bandwagon. One, who used to work in business development for a campaign management platform which shall remain nameless, up until recently sung the praises of the Long Tail to the sky. Until he changed jobs. Now, he says that’s so 2005. The top 10% of your keywords create 99% of sales volume. Or if you make really long keyword lists, the top 1% of your keywords create 99% of your sales volume. Let’s get real.

If I had my choice, this would be my shortest column ever! Maybe I’ve been tweeting too much. Forgetting prepositions. PPC superhero tired. Many keywords. Dragging him down.

First, the requisite disclaimers. Accounts vary. Campaigns can benefit from a lot of keywords, sometimes, if managed correctly.

But the situation I ran across recently was typical. Let’s say the product (name changed to disguise client identity) is “balsa wood sailboats.”

Inside the ad group someone had diligently built years ago, I found hundreds of phrases: [green balsa wood sailboats]. [balsa wood sailboats midwest]. [how to make a balsa wood sailboat]. And my particular favorite: [pinkbalsawoodsailboat]—perhaps a translation from the original German. And hundreds of others. You might think that being so thorough is good campaign management. I don’t think so.

Not to overcomplicate my reaction: I have a pretty good feel for these things. And these pages and pages of keywords were really starting to piss me off.

Part the reason for this visceral reaction boils down to keyword intent. Googler Nick Fox has recently been quoted as saying that focusing Quality Score so heavily on CTR is kind of like using the wisdom of crowds to tell you what is or isn’t relevant. CTR isn’t the only relevancy factor, but if there is a huge disparity from e-commerce industry norms, that should be speaking volumes. So the high-intent words were generating really good CTR’s, in the 3% range. And some of the words that might be more informational or DIY in nature, were eking along at 0.7% or so.

Perhaps rather than asking for each and every low-volume keyword, whether its presence might help a campaign, ask if there is much downside to getting rid of it. If it’s very low volume, and the same search might be covered by a broad or phrase match in the campaign (adding appropriate negatives when time allows), then hyper-efficiency isn’t really doing you much good. And indeed, the presence of such keywords can be hurting the account more than helping (though the impact should be marginal either way): they drag down overall Quality Score, and they could be making the job of analysis just a bit more difficult.

Granularity is good, of course. The ability to tightly target very specific search intents is why we paid search marketers get off on what we do. But you can overdo it. By letting these “passenger” keywords take up too much mental bandwidth, you forget that your job is to make the bold moves—to get the big important things right.

The historical data for the past year of this particular case study campaign backed up my theory.

Only the top 80 keywords or so had been clicked even once or more in the past year. Out of 116 conversions, only one came from the bottom 30 of those; in other words, from an economic standpoint we would have been OK if we’d cut the list off at 50.

104 of 116 conversions came from the top 15 keyphrases.

A few of those farther down the list that seemed promising, like [organic balsa wood sailboats], received few clicks and zero conversions for the year. The “organic” keyword got 7 clicks in the year, but didn’t convert. The CTR was adequate, at 2.5%. I guess I’ll let it run for another year. Of course, by then, it’ll only have around 14 clicks—far from statistical significance—but it’s less annoying than some of the others (the ones with zero clicks, especially).

After my deletions, there remain 200 active keywords in the group. I’d like to see another 100 or so eliminated.

Perhaps I’ve been doing this too long, if certain keywords irritate me so much that I want to squash them under my foot like ants. If you prefer to treat hundreds of stray, low-volume keywords as precious larvae just waiting to become beautiful, conversion-boosting butterflies, that’s your call. They just happen to make me ornery.

My approach might be altered slightly if attribution models were more forgiving. Some of these could be counted as “research” keywords that contribute to further searches down the road, and eventually, sales. But given that we’re generally evaluated based on strict ROAS criteria, and Quality Score punishes us for low CTR, that’s a nice theory that doesn’t reward campaign managers today. Performance that takes more than a few days to be measurable isn’t something that most clients and bosses are able to reward, especially if they’ve drunk the AdWords-as-Direct-Marketing Kool-Aid.

Beyond a certain point, more keywords won’t lead to more sales. Remember the purpose of testing and improving paid search campaigns: better CPA’s, and increased volume and total profit. If the long tail stuff isn’t helping with that, isn’t it time to focus back on what will help you achieve those objectives?

You’re not off the hook quite yet. Although the long, shot-in-the-dark, or overly specific phrases were not helping the account, that doesn’t diminish the importance of keyword planning and keyword discovery efforts. At a high level, there are probably concepts you (and maybe the tools you use) are missing.

If there are some regional variations on the name for “balsa wood” (hypothetical example: “scatterwood”), or different ways for saying “sailboat” (obviously there are), then you’ll need to incorporate keywords using those concepts, perhaps by testing them in different groups with their own ad copy. It’s a matter of feel: how many significant variations in concept are worth representing in the same ad group, or different ad groups?

But do refer back to the core point here as you do this: the excessive granularity often counseled by busywork-loving “best-practices-robot-people” can actually throw a wrench in the response testing process, requiring you to wait ever longer for a statistically significant conclusion. Just as there is such a thing as too many ads in an ad group, there is such a thing as too many campaigns and too many ad groups. And as I’ve tried to show, too many keywords.

That’s all for this month. Gotta run. Spidey senses tingling. PPC-man, away! Off on another important mission.