It is with tears I write this article. I weep not solely for my loss, but yours as well. I am referring, of course, to what I see as the inevitable demise of a legend, an icon, the darling of all marketing land:
Joking aside, the keyword is an entity we should pause to appreciate more. In my recent book (Ponderings of a PPC Professional), I devoted an entire chapter to the remarkable nature of the search keyword. I am no enemy of the keyword, and yet, I find myself understanding the inner workings of Google search, marketing psychology, privacy and technology better than I ever have before.
This has led me to some reluctant (and some would consider, unfortunate) conclusions, one of them being that I do not expect us PPC advertisers to continue to have the ability to bid on individual keywords in the next few years.
Weeping aside, I find myself in the unenviable position of changing my viewpoint. “Changing” is a strong word, perhaps adjusting or tweaking, is a better term here.
The remarkable power of the keyword
Make no mistake, I still believe the keyword to be the most remarkable marketing tool known to humankind. A keyword is an advertiser’s means of selecting which search terms they want to advertise on.
The search term tells you exactly what individuals want, when they want it. This gives advertisers the opportunity to position their product or service in immediate response to that communication, and to even determine what value (the bid) that exchange carries. Absolutely remarkable!
Aside from someone physically walking up to a marketer and asking them a question (we typically call that “sales”), where else can this sort of immediate connection be seen in the world of marketing? This is why we PPCers have always battled changes to the keyword.
“Keep exact match, exact!” is an oft-repeated battle cry heard over the past few years on Twitter and in PPC conferences (for the youngsters: conferences were an impressive conglomeration of people who could actually communicate in person, back in the day).
But, and I say this knowing I will have pushback, have the times changed?
In the remainder of this article, I hope to re-open the PPCer’s traditional understanding of the keyword by presenting new information I believe should not be ignored.
Yet, I also want to walk a fine line: For anyone who has followed my writing over the past few years, you can relate to the fact that I am no Google shill, buying into their every edict with breathless anticipation. While I am happy to learn Smart Shopping, there has been no critic more vocal (on multiple platforms) about the alarming data loss and its harm to businesses and advertisers.
On the other hand, I want to be cautious not to criticize everything Google does immediately without all the facts. That, of course, can get difficult when the company withholds information from advertisers making it seem suspicious… But, sometimes I believe that is more to do with poor communication than the conspiratorial mutterers tend to admit.
So here we go. Here is my confession.
The inherent (newer) weakness in the keyword
As much as I love the keyword, and still believe it carries immense value, I have begun to question whether it is still telling us as much as we think it does.
Understanding this is important for our industry to move forward, because if the keyword is not telling us as much as we think it is, then we have an obligation to evolve with Google’s system for the good of the accounts we manage. After all, we’re here to grow our PPC accounts the best we can; that is our ultimate goal: to serve our clients or place of employment well.
What I would currently argue is that the combination of changes in user behavior, changes in Google Ads automation and targeting, and an increased number of privacy changes have made the keyword less of an intent-revealing ace-in-the-hole than it used to be.
Because of that, we must begin to rely more upon user-defined targeting in order to fill in the gaps of our contextual knowledge of searcher intent, whereas in the past, we would rely almost solely on the keyword.
Lest your eyes glaze over, let me get practical with an example before moving into the details: A user is interested in purchasing a new TV. They go to Google to type something into the computer. How should we bid on them?
Advertisers’ changing relationship with the keyword for targeting
Circa 2015 PPC: In the old days, we would build lists (I’m talking, LISTS) of detailed long-term keywords (exact match, of course!) based especially on modifiers (where we really could flesh out intent).
This might look like thousands of ad groups according to the following:
[best flat screen tv under 1000][65inch plasma tv above fireplace][samsung large lcd tv][samsung tv vs sony tv]
While we could use broader keywords for upper-funnel targeting, the truly savvy advertisers got focused and ultra-specific with what terms they would target. Long live the long-tail, purchase-intent revealing keyword!
PPC in 2021 and beyond:
Privacy regulations and changes are beginning to wreak havoc on the specificity of tracking. In addition, user behavior has been trained by Google personalized results to never spend more time than they have to typing (or speaking) into a phone (more on this to come below).
A few years ago, Rand Fishkin gathered data and presented some remarkable findings on search behavior trending towards shorter queries. He discovered that 46% of searches are one- or two-word search queries. Nearly one out of every two searches doesn’t even get to three words! Imagine what that is in 2021 (no, really, I’d love some data on this… I looked and asked, and I couldn’t find any).
Tim Soulo of AHREFs reminds us that long-tail keywords are not the same as multiple-word keywords (a good reminder) in his helpful and accessible article on the long-tail keyword. It is helpful to note his keyword length vs. monthly search volume chart within the article, which also demonstrates the significant number of fewer-word keywords in the head terms (high search volume) camp.
While users still search for a significant amount of new, long-tail keywords (regardless of word length within the term), my hypothesis is that user behavior has changed, along with platform changes, to result in a different PPC keyword landscape than has ever been seen before.
The wise PPCer will identify the ways in which adaptation is important, even while frustrated about losing search term data. We are stuck now, because the best-converting keywords to bid on in our accounts have likely shifted (in many, but certainly not every, instance) to look like this:[best tv 2021][led tv][samsung]
If we had bid on those three terms in 2015, our budgets would have resembled waste flushed down a toilet, and disappearing quickly. Now, those terms may be the core of our accounts.
Here is my assertion: We are seeing a shift in user behavior to intent-obfuscating search terms, which then impacts purchase behavior and our keyword targeting and bidding capabilities. This shift (along with Google Ads changes that may or may not be in reaction to this shift) is going to continue to force us to rethink our keyword targeting strategies.
Why is this happening?
This is a fairly remarkable change, and I think it’s worth considering why it has happened. If I might be allowed to ponder publicly (which runs the risk of me being wrong), I’d suggest four key reasons why this is happening, and it’s important to note they are all interconnected.
A brief caveat in case the above paragraph was ignored. These are observations formed to create an assumption. I could be wrong (but hey, you could be, too, right?) so I hope this at least encourages fruitful discussion.
Here is what I believe has provoked a change in how we should view keyword targeting in PPC. I should also note, internal conversation on our team has helped immensely with this as well. I have a brilliant team, and they’ve helped suggest or hone the specific points below.
1. A better product. I have to hand it to Google, the company has done a remarkable job with its search results. Think about it (including your own search behavior): When people go to Google now, they experientially know they don’t need to fill in all the gaps because Google personalized search seems to somehow just… know them. People have been increasingly trained by Google (without knowing all the details of their previous behavior being tracked to build that product) to get lazier in their searching because Google’s results will still be personalized to them. That’s pretty remarkable, and Google has indeed built an impressive search product.
Let’s say Person A is in the market for a mattress and has spent weeks investigating numerous brands online, researched local mattress shops, clicked on specific shopping ads to view individual products they’re interested in. Google has all of this data, and the next time they go to Google ready to purchase, they simply type [mattress] and Google knows exactly which Shopping Ads (dynamic remarketing) or local stores they should show to Person A.
Person B, on the other hand, is just beginning their shopping experience online, so when they type [mattress], Google changes the results to a higher-funnel, more informational SERP.
Same keyword, dramatically different results, amazing product!
2. Autocomplete (plus constant improvements). Along with the first point, consider how Google autocomplete plays into a change in user search behavior as well. I would go out on a limb and say I now use autocomplete more than not. Autocomplete really is remarkable in its accuracy (though, humorously not at times) and this goes along with the better search product mentioned above.
I’ve even gone to Google at times not knowing exactly what I will type (and not caring that I don’t know), since I know Google autocomplete will figure it out. Again, credit to the impressive product Google has built, but autocomplete certainly impacts how people search in bulk.
One of the things Tim Soulo brought up in his article (that I previously linked to), is the surprising number of many-word terms searched for in head term searches. According to Ahref’s data, 9% of all searches with 10,001+ monthly searches have four or more words! That is remarkable, but I wonder how much of that is influenced less by synchronization, and more by autocomplete.
In fact, while writing this article, I did a search on Google and used autocomplete out of habit and ended up clicking on a five- or six-word term. If Google senses trends in specific phrases and raises those in autocomplete, it stands to reason that act will, in itself, increase the number of searches done for that exact phrase.
That changes the nature of how people search and the value of certain keywords for bidding within Google Ads.
3. Mobile device impact. What would a change over the past years in user search behavior be without the impact of mobile? What would an article about digital marketing be without mentioning mobile?
If you remember such declarations as “Mobilegeddon” a few years ago, there was a shift of Google searches from desktop to mobile. This never fully transitioned as proponents of the concept suggested, simply because there are still a healthy number of people who utilize (and will likely for decades to come, in my opinion) desktop for things like work.
Yet, mobile devices have an impact on search behavior, in part, due to practical considerations such as limited screen size and inefficient touchscreen keyboards (I must be getting old, because typing on an iPhone keyboard makes me want to throw my phone into a blender).
One dataset I have never seen, but would like to, is whether the number of words in a search term tends to decrease based on device. I would be fairly shocked if it didn’t due to the aspects I noted above. It is also possible more-word terms tend to be more voice-centered while fewer-word terms tend to be input via touch screen. This is admittedly a calculated assumption based on known behavior rather than fact based on gathered data.
4. Google Ads close variants. Finally, we get to the Google Ads product itself. I purposefully left this as the last point, because I think we PPCers have tended to under-emphasize the influences of the factors mentioned above on changing keyword targeting.
However, we would be remiss in not acknowledging the changes within the Ads platform itself on keyword targeting — one of those being close variants. In this way, Google has made exact matching of long-tail (as well as more-word) search terms to specific advertiser-selected keywords more difficult as contextual and close variants are now considered. Negative keywords may assist in filtering out some of the unwanted variations into a preferred ad group (as some advertisers have developed into “query filtering” or “query funneling” strategies); however, traffic availability may not always allow for this… it can sometimes actually prevent your overly-filtered keywords from serving, rather than successfully filtering the term into the correct keyword within your structure.
This has certainly had an impact on the changing nature of keyword targeting. Google argues that this is to allow the advertiser to match up to the many searches (mentioned earlier) that are contextually the same as the keyword selected, but that do not match it exactly and would thus be missed in the advertiser account. If it were up to me, I would prefer broad and phrase match to allow for close variants while leaving exact match a level of exact specificity, as this would (I believe) unite the role of machine and human oversight in keyword targeting in a healthy manner, especially for budget and ad creative control to core terms.
But, here is where I think it’s crucial to re-assert the purpose of this article: I would point to the first three reasons above as to the cause of a shift in user search behavior. We PPCers can be so focused on the data points within our ads system, that we miss the human aspect here, which also drives our targeting capabilities.
Yes, the close variant matching has had an impact on our ability to target as specifically as we were able to target in the past, but I would argue we are seeing a shift in user behavior to intent-obfuscating search terms which then impacts purchase behavior and our keyword targeting and bidding capabilities. This shift (along with Google Ads changes that may or may not be in reaction to this shift) is going to continue to force us to rethink our keyword targeting strategies.
Bonus: 5. Privacy. I would have failed if I didn’t at least throw a passing nod to a core aspect of the search environment that is in constant flux: privacy.
As Google begins to eliminate the third-party cookie and to institute FLoC and the Privacy Sandbox for tracking (along with first-party cookie products, of course), users are certainly more aware of data, tracking and their own privacy than they have been in the past. I do not have anything to point to quantifiably here, other than to suggest something as socially contextual as user privacy likely plays a role here in changing user search behavior.
Practically, more people than ever are shifting to privacy-encouraging browsers such as Brave, and news articles such as Google being sued for tracking users within their Incognito browser window suggest the average internet user is more aware of, and savvier regarding privacy decisions. It isn’t too much of a stretch to assume that this could work its way into search behavior as well. However, I don’t think it is as significant as the first four, so I only noted it less as a root cause of a clear user shift in behavior, and more as something to be aware of and monitor.
So what do we do about this?
I think the key is to evolve with this change in user behavior and the Google Ads platform changes by accepting a more user-targeted system, and shifting strategies and tactics accordingly.
Think about it like this: If 1,000 people are shown your ads for “samsung tv,” and in this paradigm of new user behavior that I have outlined above, 20% of those people are ready to purchase within the next seven days, 20% are in the education and shopping phase, and 60% are too high in the funnel to determine a strong purchase intent, then don’t we need additional context in order to capture the 20% that are ready to purchase now?
Here is where the additional user signals (that Google alone has access to) become essential. A frustrated PPCer may exclaim that Google shouldn’t be the only entity with access to that data (I am in agreement with this concern as I have published previously), but that is an argument for a different time since I am trying to pragmatically explain how I believe PPCers need to evolve within a changing system based in large part by a change in user search behavior.
Patrick Gilbert, in his book Join or Die, notes that there are over a million signals utilized by Google to determine bids for an advertiser within a single auction (this is supported within this Google help document).
How does Google know whether that person typing in a generic two-word query is closer to purchase or not? Because they have a mind-blowing number of data points they have collected on that user based upon things like previous search behavior, sites visited, affinities and interests and a host of other behaviors we humans simply do not have access to.
Legitimate concerns about closed systems aside, this is one of the strongest arguments I have heard for embracing more automation within Google. I simply don’t have access to that level of data in a real-time auction scenario, and with decreased intent revelations based on the keyword itself, I also don’t have a lot of other options. In the words of Patrick, the PPCer of 2021 and beyond must “Join or Die.” That’s not a threat (at least from me or Patrick), it’s a warning. A plea, even.
What does this mean for the keyword itself?
I believe it was my friend, Aaron Levy, who once referred to the keyword’s future as suspect simply because the idea of the keyword has already changed. “Think of the keyword more as topics or categories,” Levy noted (I paraphrased). I think Aaron has nailed it, and this supports what I have been outlining above. I believe the wise PPCer will begin to prepare for a world in which the anonymized and aggregated data points on individual users (which only Google has access to and can utilize) are the primary signals used, and the keyword is simply one part of this system (albeit, one in which significant weight should be placed based on the nature of its intent-wielding power, as outlined in the first part of this article).
However, weep as we may over the potential loss of the keyword, if you consider my previous points of user behavior itself changing, perhaps this isn’t simply mean ole’ Google removing targeting to line their own coffers (or, I mean… it could be both). Perhaps, it’s simply a necessary evolutionary step as privacy regulations creep forward, users give less specific data points to advertisers and Google’s incredibly high and well-guarded walled garden holds the only keys to the kingdom of search advertising.
Regardless of what the future holds for us, I believe in my core that the advertisers who shift more quickly to understand and manage automation will be those more likely to survive.
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.