I’ll approach this question from two angles: first, at a high level, and then from a more empirical perspective. My high-level observation is that most SEM accounts have too many keywords. There is an outdated notion that tons and tons of keywords drive lots of long-tail clicks – clicks with high relevance and low CPCs. While that may have once been true, I have consistently found that the long tail is mostly dead, killed largely by Google’s efforts to push more and more advertisers into competition among fewer and fewer keywords.
There are exceptions to this rule -- most notably travel and ecommerce, where dozens of destinations or SKUs can drive thousands of unique searches. For the most part, though, even in these categories, 95% of the accounts I review fall into what a friend coined “the uber Pareto,” where 95% of an account’s volume comes from less than 5% of the keywords. As a ballpark figure, this usually results in between 50 and 200 keywords that really matter.
Of course, you can build out keywords 201 to infinity, but at some point the time required to manage all of these tail terms, combined with the constant threat that you might get broad matched on a ridiculous query, makes the ROI of long-tail keyword investment shaky at best.
Many SEMs are nervous about simply slashing any keyword outside of the top 5%. What if one of those tail terms is a diamond in the rough, ready to become a top-performing head term? This is a valid worry, and it brings me to my more empirical argument: It’s always a good practice to continue to test keywords in hopes of finding a winner.
We do this through the Alpha Beta process, which puts experimental keywords on broad match modified and promotes winning queries to “alphas”: what we consider top performers. Within this system, it is impossible and imprudent to test more than a few hundred “betas” at once, but we usually find that a beta can get enough volume to be declared a winner or loser in a few weeks.
Put another way, if the average account has 50 to 200 keywords that are already declared “top performers” and another 100-200 that are being tested for inclusion in the top category, it’s reasonable to see an account with somewhere between 200-500 keywords live at any given time.
This number is no doubt much smaller than SEM accounts of yesteryear, where you could actually get a lot of quality volume by creating tens of thousands of keyword combinations. (In 2005, I concatenated the words “mortgage,” “home loan,” “refinance” and “HELOC” with all 22,000 towns in the U.S. and created 88,000 effective keywords overnight. This would have little value today).
The trend for fewer keywords may also make some SEMs nervous: If I can’t spend my entire day creating clever keywords, what is my value? Fear not. There’s plenty of SEM work to do today that doesn’t require massive bulk sheets of eight token queries. As keywords sets have become smaller, the value of each keyword increases, creating a need for extensive ad text testing, landing page optimization, and keyword-specific bidding.
I’ve seen accounts with 15 keywords, and one that had 250 million. The answer to “how much is enough” is somewhere in between those two extremes, and a lot closer to the former than the latter!