KW Explorer is Working to Disambiguate Keywords Google Merges Together

[randfish]

Hey gang,

Russ Jones from Moz has been doing a ton of heavy lifting work to try to get around the new problem posed by Google AdWords recent change to merged-keyword volume data. But, we're fighting back against this obfuscation in Keyword Explorer. I'm sharing two emails (slightly edited) from Russ about what we're doing here:

Introduction to the Problem:

Google Adwords Keyword Planner is the primary source for keyword search volume (how often a keyword is searched monthly on Google) for much of the search marketing industry. While Google has grouped together highly-similar terms for a while (especially misspellings), in June of 2016 they dramatically increased this keyword-grouping. This means similar phrases like "keyword rank", "keyword ranking" and "keyword rankings" would all be reported as having the same, combined search volume, rather than their individual search volumes. If you were to take Google's numbers at face value, you might think there are 3,000 searches per month for these 3 terms, when in reality there is only 1,000, divided amongst the 3 terms.

How we are addressing it:

Moz's Keyword Explorer uses a blend of data sources, not just Keyword Planner, to build our volume metrics. This gives us a distinct advantage in that we can adjust the volume of words that deviate dramatically in one data set verses another. Take for example the phrases "keyword rank", "keyword ranking", and "keyword rankings". While Google Keyword Planner might report all of these as having 1,000 searches per month, Moz Keyword Explorer can detect that these numbers are significantly higher than what our models would predict given our other data sets. We can then adjust the volume accordingly. Moreover, given our huge keyword data set, we can also identify grouped phrases (like these 3) and divide the volume proportionally to what we see in our other data sets. Thus, we address the grouping problem from multiple directions.

Here's email #2 from Russ, detailing more of how we're attacking this:

I have been working pretty much non-stop on this keyword volume disambiguation problem (finding the real search volume of individual keywords when Google clumps several together). I think I have settled on a pretty good solution and am working on getting it all in. For example...

Google Keyword Volume for the phrases "briefcase for women" and "briefcases for women" are both at 3600 because they have been lumped together. My disambiguation script says the singular (briefcase for women) should be 2731 and the plural should be 869. Google Trends roughly agrees with this, showing that the singular is searched more than 2x the plural: https://www.google.com/trends/explore#q=briefcase%20for%20women%2C%20briefcases%20for%20women&cmpt=q&tz=Etc%2FGMT%2B4

Basically, Keyword Explorer should already be providing some more accurate/segmented numbers than AdWords, and in the future, we'll get even better thanks to our clickstream data and our evolving models.

Any questions, let us know!

[bradycorp]

Thanks for touching upon this topic which is seriously impacting all marketers who rely on google data. Even the core definition of "Search Volume" is now NOT correct. It's THAT bad as pointed out here http://bit.ly/29MvvvK. Thanks for your efforts.

Search-Volume-Data-Merged_2016June.png

[bradycorp]

This has a HUGE HISTORICAL impact.

Search-Volume-Data-Merged_Definition-Incorrect.png

[94501]

Thanks Russ and Rand! I've used many tools for getting to the bottom of estimated volume and have high hopes for your evolving KW Explorer. It's definitely needed.

[rjonesx. 0]

You're welcome for our efforts. I am really excited as we roll out this volume fix. I think it is going to offer a real improvement for SEOs everywhere.

[rjonesx. 0]

Thanks!

[Igor.Go]

You're working in the right direction, guys. It seems a little scammy on the part of search engines to show these search volumes, as if they prompt you to bid on those keywords because they have a high search volume. While they do show the same SERP for this keyword, I believe a page optimized with "Buy a car in Orlando" will have a better score and a lower bid than a page optimized for "Buys a cars in Orlandos".

[iREP]

Thank you, Rand and Russ!

I found a weird result. "bag" in US is reported to have 0-10, while "bags" in US is 11.5k-30.3k.

Considering KW planner and Google trend,  "bag" in US seems to be outlier.

[rjonesx. 0]

That is a weird bug, because our database is showing 18k+ for that term! I will investigate!

[rjonesx. 0]

The bug has been fixed but it will take 24 hours for the caches to clear! Thanks again!

[AndyKuiper]

Thanks for taking the time to come up with a 'solution' to the problem

[iREP]

Thank you, Russ!

Google confirmed me that this change is due to an update, it won't be rolled back and there's no way to fetch data for each variant individually. Details are here.

Your tool will help us a lot.