For the most up to date information, please view our comprehensive keyword research guide.
Pump Up Those Pages
For this particular study, we analyzed 989 keywords for a fitness supplement site we are trying to rank.
The seed keyword we used to collect all of the different keywords in our list is "protein powder".
The input of this B.Y.O.D. report is an export of the keyword magic tool from Semrush.
Our objective is to demonstrate how we can use Keyword Cupid's functionality to automatically pull metrics for each of the keywords we are going after and allow the algorithm to create groupings and clusters that will unravel how Google associates them on their end.
The initial keyword list we built had 2,500 keywords.
Using our Smart Filtering Algorithm©, we analyzed multiple features like low impression volume, many competitive sites in the results, low-value words in the phrase, etc. to decide whether or not the keyword is worth passing to the final dataset.
The final dataset ended up having 989 keywords.
So Many "Reps"
One could easily collect 989 target keywords through various free tools.
However, this large of a list proves challenging to organize.
More often than not, there are several questions one needs to answer before they are genuinely ready to sow the fruits of their keyword labor.
Should I go after 'best whey protein' or 'best vegan protein' first?
How do I properly organize them to avoid pages ranking for the same terms?
Should I create separate posts for 'banana protein shake' and 'chocolate peanut butter protein shake' or put them on the same page?
And of course the biggest question of all...
WHY choose Silo 'mass building protein for women' instead of Silo 'lean protein for women'?
Keyword Matchmaking From The "Google Heavens"
Keyword Cupid is here to help you answer all of your questions with only one objective, to help you rank in the eyes of Google.
Our hypothesis is simple.
then "Best Protein For Bodybuilding" and "Cheapest Protein For Bodybuilding" must be related.
- Keyword Cupid
We don't care to answer why or how they are relevant, nor are we trying to reverse engineer the NLP algorithm that produced that association.
Our objective is to place on the same page/silo keywords that Google associates with each other.
This way, we can speak the same language as the Google NLP algorithm.
We solved half a dozen optimization problems and eventually created an ensemble of machine learning models using k-fold cross-validation to optimize the homogeneity and the completeness of the clusters we produced.
Let's go see the results, shall we?
Get Your Page "Summer-Ready"
All of the subsequent parent-level clusters have to do with diet powder, diet drinks, weight loss or even a cluster on how much is a scoop of protein.
Arguably, one of the significant intents in the fitness and supplements industry is diet.
Who doesn't want to lose those extra "love handles", tone up and get ready to hit the beaches?
This intent manifests in a very pure form in the cluster "protein shakes for weight loss" that we see in the image below.
All of the subsequent parent-level clusters have to do with diet powder, diet drinks, weight loss, or even a group on how much protein is in a scoop.
Each cluster shows aggregate metrics on volume and difficulty for all its children.
Following down the trail of the highest impression volume cluster, we will further analyze the cluster "weight loss protein powder" which has a total aggregate impression volume of 8,550 views per month.
The keywords in the cluster exhibit interesting intent properties.
Most of them have to do with "protein powder for weightloss", "protein powder for diet" etc.
One not so obvious association is that the algorithm clusters together "fat burning protein powder" with "weight loss powder drinks".
Fat burners, mostly in the forms of capsules, have very different substances and formulations than protein powders, which target protein synthesis.
Interestingly, there are 250 searches per month for protein powders with fat-burning capabilities.
In the two images below, you can see that the two queries have 5/10 SERPs in the first page to be the same.
Our models analyze more than 50 results for each query and exponentially weigh the associations in later pages to accentuate the importance of the associations on the first pages.
We can utilize this information to have an excerpt in our content about how lean proteins are different than fat burners and other ways the audience can boost their metabolic rate to lose weight.
Mirror Mirror On The Wall,
Which Is The Best Vegan Brand Of Them All?
Dietary preferences and restrictions are promising content verticals because the respective audience is passionate about the subject.
One of the parent-level topics that emerge from our inspection is related to do with "best clean protein powder" keywords.
The following themes are very informative as they include subjects like "protein powder safe for breastfeeding", "no sugar protein powders", "organic protein powders", "vegan protein powders" and so forth.
The generated groupings are all about clean sources of protein powder, creating a well-formed content strategy.
Vegan diets are appealing both from the physiological standpoint, the benefits for longevity,
as well as the ethical up sights as they promote animal rights, conservation of resources, clean sources of diet, etc.
We decided to investigate the cluster "the best vegan protein powder".
The results are illuminating as the algorithm creates sub-silos for all the vegan protein brands automatically, without any supervision.
The keyword groupings capture the domain knowledge we lack by further aggregating brands in the field of protein powders with vegan ingredients.
- "aloha protein powder reviews"
- "arbonne protein powder"
- "olly protein powder reviews"
- "toneitup protein powder"
- "vega protein powder"
- "vegan protein powder whole foods"
This is not magic as it is working because Google has done the heavy lifting for us by analyzing the intent and the semantic relevance of each query.
We are just happy to piggy-back on their results and reverse engineer the clusters we need to support our content.
Cheap Isn't Always Best
... but maybe it is
We can all agree that the intent "best" and the intent "cheap" are not the same, or at least it is subjective to who you ask.
In some niches where the consumers are not aware of the distinctions between brands, especially when the products have inherently similar characteristics, "best" and "cheapest" tend to be correlated.
When you search for a car, seemingly small differences can play a huge role in your choice.
In contrast, when you purchase protein powders, you may only be interested in two criteria, the flavor and the protein content.
Due to federal regulations, the ingredient list of each supplement is publicly available.
For instance, the vast majority of Whey proteins have around 20g of protein per scoop.
These standardizations allow consumers to focus more on price than anything else.
Therefore, it is reasonable that "best protein" and "cheapest protein" are correlated consumers' minds.
So far, our hypothesis intuitively makes sense, but how can we test it?
And most importantly, how can we find on the right questions to bridge the gap with more established and knowledgeable brands?
In the two images below, we see that "best vegan protein powder bodybuilding" and "cheap vegan protein" have 3/10 matching results on page 1 alone.
Our algorithm analyzes up to the first ten pages of each keyword to calculate the SERP overlaps.
Therefore there are potentially even more connections between these keywords in later pages.
Capture ALL The Intents
There are other topics we identified by skimming through these clusters.
One of the parent topics is associated with "weight gainer protein shakes".
The parent cluster has 9,200 impressions per month and an average difficulty of 13/100.
Several sub-topics stand out, such as "weight gain shakes for women".
Within that cluster, we find another parallel intent "healthy weight gain shakes".
The word "healthy" is a notable distinction because many athletes associate weight gain and bulking with eating "cheat food" (pizza, burgergs, etc).
Creating a content strategy that targets women who want to gain weight by eating healthy food could set your site apart!
A vanilla keyword research strategy would rely on lemmatization and stemming to find the roots of each word and assess similarity.
A more advanced approach would use NLP and LSI to find similar words using word embeddings like Word2Vec or G.L.o.V.e.
Our approach goes one step further also to encapsulate the intent that these words contain as they appear in SERPs.
The cluster below shows how the algorithm has "understood" that "stevia" and "sucralose" are types of artificial sweeteners.
Another keyword in the cluster is "ghost protein powder".
As we can read in this source, Ghost contains sucralose. Still, no other sweeteners like acesulfame potassium, which is present in a large number of protein powders, tend to be more vilified than sucralose.
The results of this study illuminate many benefits of using Keyword Cupid for your keyword research.
Our similar clustering of "best" and "cheapest" signified similar intent between the two verticals, which is contradictory to common thinking.
The clustering algorithm we employed singled out healthy weight gain as a possible perspective when writing about women's weight gains.
Finally, our keyword groupings captured domain-specific knowledge pertaining to brands that meet the underlying cluster theme's specifications.
For instance, it accurately grouped six vegan protein brands under "the best vegan protein powder" cluster.
Without Keyword Cupid all of these findings would have required countless hours of manual research and critical thinking.
It is also possible that the resulting conclusions of such manual research would have been tainted by user bias.
Spread The Love To YOUR Project
Are you interested in applying our methodologies to your own project?
Do you want to learn how your custom keyword list is grouped in the eyes of Google and how to use them in your content to increase their semantic relevancy?
We are ready to help!
Just point our cupid arrows in the right direction!