There’s much speculation that Google is experimenting with a new way to represent search volume in the Google Keyword Planner.
Over the past few days we’ve been noticing some fluctuations in search volume and, after doing a little digging, it appears this may be a result of Google lumping together the search volume of close variant keywords.
According to Jennifer Slegg at The SEM Post:
“…Google made a change to [Keyword Planner] where instead of showing individual keyword estimates for each keyword or keyword phrase, Google is now lumping in the data together, meaning Google will show identical [search volume] estimates for similar keywords or keyword phrases…”
This change is not yet universal, and with no official response from Google, the online SEO community is hotly debating whether or not it’s a bug or a permanent change to the way search volume is reported.
WHAT CONSTITUTES A CLOSE VARIANT?
While this change to the Keyword Planner continues to roll out, it’s not immediately clear how Google is classifying close variants. Based on The SEM Post and our spot testing to date, it seems to be:
- Plurals and non-plurals [juicer] [juicers]
- Acronyms and longhand [SEO] [search engine optimization]
- Stemming variants [design] [designing] [designed]
- Keywords with and without punctuation [men’s jeans] [mens jeans]
- Keywords with and without spaces [white paper] [whitepaper]
A Few INTERESTING EXAMPLES
Juicer and juicers
In a recent blog post on the SEO funnel from June 16, we pulled ranking data for 27 similar keywords. Below you can see that [juicer] had a regional search volume of 74,000, while [juicers] (plural) had a search volume of 18,100. You’ll also notice that [juicer] has a CPC of $0.66 and [juicers], a CPC of $0.84.
Today, the Keyword Planner shows both of those keywords as having a monthly search volume of 90,500, and the CPC is the same, sitting at $1.16. We can no longer see which keyword within this close variant set has the highest or lowest search volume, thus grossly overestimating the search volume of one — [juicers] has shot up from 18,100 to 90,500.
Texas A&M football
While most of the close variant keywords that we tested were grouped, many remain with their individual search volume intact, which either speaks to a bug or a slow rollout.
For example, we found that [Texas A&M football], [Texas A and M football], and [Texas A&M foot ball] — three variants clearly looking for the same thing — all have completely different search volumes and CPCs.
SCUBA and self-contained underwater breathing apparatus
Like most close variants, acronyms and their equivalent longhand keywords seem to be grouped; however, we did notice an interesting outlier in [scuba] and [self-contained underwater breathing apparatus], which still have different values.
Here is where search intent appears to be playing a bigger factor than close variance. Unlike [CMS] and [content management system], which have similar search intent, the intent behind [self-contained underwater breathing apparatus] would likely be different than that of [scuba].
A search for [scuba] is probably looking for the diving-related activity, whereas a search for its longhand term is probably interested in breathing apparatuses. In this case, it makes sense that Google would not apply the same search volume to both.
WHAT DOES THIS MEAN FOR STAT CLIENTS?
Since Google’s Keyword Planner is the primary source of Google search volume data for our industry, any application that relies on Google search volume data will be directly impacted by this change. Subsequently, STAT clients may see an inflation in search volume for some of their keywords.
Take, for example, a tag containing the two keywords [seo] and [search engine optimization]. Google appears to be treating them as though they’re one single search, and has assigned each of them the exact same search volume of 111,000. Because of this treatment, a tag containing the two keywords would therefore also have a total combined search volume of 111,000.
In STAT, the aggregate search volume for that tag will show as 222,000 — the combined total of each keyword’s new shared search volume. As a result, this will artificially inflate aggregates and averages that contain close variant keywords.
STAT clients may see this search volume adjustment throughout the application, including graphs, reporting, optional APIs, and — possibly most notable — in the competitive landscape, where share of voice calculations are weighted by search volume.
STAT’s next steps
After working on the problem yesterday, we have a solution in mind that will allow us to identify the close variant keywords that have been grouped together, and we’re figuring out a way to be able to triangulate their original search volume.
For now, our plan is to hold off until we get some confirmation that this is not a bug or a test. If it’s determined to be a permanent change, we’ll begin implementing our solution.
We’re watching this closely and as things evolve, we will continue to provide updates. In the meantime, if you have any questions or concerns, please reach out to us at support@getSTAT.com