SEOKeyword Clustering for Maximum Search Profitability

Keyword Clustering for Maximum Search Profitability

An in-depth look at how to group your list of keywords –keeping in mind your short-, medium- and long-term goals –ensure the highest ROI potential per sale by looking at your competition and search volume, and the tools you need to get it done.

dr-wordlove

In “How to Find Profitable Keywords For Your Website,” we discussed methods for researching and logging your keywords, determining the relative profitability per phrase and touched on the need to schedule your targeting to help produce a faster ROI.

The approach we used is to select long-term keywords but focus on complementary, long-tail phrases early on. The logic? It’s easier to spend the money you’re making than the money you’re not – so a path to faster ROI, even if it delays slightly the time frame for hitting maximum monthly revenue, is still desirable.

We ended the article last month touching on, but not getting into, the need to do some basic competitor research to prioritize your phrases. While we may know the profit-per-sale, this is only useful if we can also determine the estimated competition for the phrases we’re looking at.

We also have to avoid analysis paralysis. With so much data available you could spend months looking into all the details of what’s possible. Rather than put you through that, the goal is to help you make bulk decisions based on grouping data and making some educated assumptions.

Thus, you may not be right 100 percent of the time in all cases, but you’ll have gained many hours/days/weeks of productivity – so where you are right (the majority of the time) you’ll win much bigger. Better to be top 3 for 10 phrases in 3 months and top 3 for 15 phrases in 7 months, especially given that you can now reinvest your profit in time and effort, and get those other 5 phrases faster.

Keyword Clustering

Group your phrases further and cluster them by common attributes. By “clustering” keywords, I mean grouping the keywords into logical groups, such that the work towards one will positively impact the results of another. The profit example from the first article (linked to above) is a good one (done by instinct as opposed to intention) and so let’s begin there.

We found in our research the following two phrases (the search volume is as reported by Google’s AdWords Tool):

  • Marzocchi 44 Rlo – 210 estimated search/mth
  • Marzocchi 44 TST2 – 170 estimated searches/mth

To fully understand the girth of data we’ll be looking at, we’ll have to imagine the full hierarchy of the site and all the products our mountain biking enthusiast has to deal with. In a simple form it would look like:

Home > Product Category > Product Sub-Category > Brand > Product > Accessories

To use an example from above, the Marzocchi 44 TST2 would be at:

Home > Bike Parts > Suspension Forms > Marzocchi > Marzocchi 44 TST2 > (any applicable accessories)

So imagine that each of these breadcrumbs represents and slough of different options (There are a dozen brands of suspension shocks alone, each with a large number of models). There is no way to target every possible phrase and further, there is no way to target each cluster at once.

Clustering starts at the lowers level of the hierarchy and works its way up. Since accessories are low-profit sale items, unless they’re add-ons (in this case), we’ll bypass that level and head up to the product level.

What we need to know is which individual products have the highest potential profit using the equations set out previously. Further (just to make it a bit more difficult), we need to take into account the traffic volumes at higher points because while we’re setting our short-term goals, we need to know what our mid- and long-range ambitions are.

So, for our purposes here let’s assume that in the suspensions forks category, that Marzocchi, is tied with Rockshox for search volume on just the brand search but that it’s higher ROI per sale. This means that if “suspension forks” is one of the main clusters we want to focus on, Marzocchi is a great brand to start with. But this assumes that suspension forks are a great place to start. To determine if this is true we need to compare the profitability of forks with the other sub-categories of the “Bike Parts” section.

Going through the lists in this way helps us isolate our major groupings. We’ll have a solid understanding of each of the levels of our hierarchy, how the brands compare by potential profit, and from there we can build forward.

Why You Need to Know How the Various Clusters of Your Site Work

If “downhill mountain bikes” is the highest search generic phrase and “Marzocchi 44 TST2” is a great product then why don’t I just go for both right away?” The reason actually falls back on your site structure (and the structure of most well-built ecommerce sites). The internal linking structure of a well-built ecommerce site favors similar products and categories.

Consider the breadcrumb navigation. The breadcrumbs on the Marzocchi 44 TST2 page would look similar to:

Home > Bike Parts > Suspension Forks > Marzocchi > Marzocchi 44 TST2

This gives us prominent internal links to the Marzocchi brand page, the suspension forks page, the parts page and the homepage. Any strength coming into this page is going to be directed more strongly to the pages within this group than on any other site.

Further, you’d likely have links to accessories and similar products (likely from the same manufacturer but certainly within the same category). All of these links are going to pass weight internally to the clusters within a group, strengthening them, which will in turn strengthen other pages within that cluster.

In essence, selecting your phrases in part based on their positions within the clusters will make advancing on all of them easier based on the simple principle of the way weight passes within a website.

Search Competition

For the ease of example let’s select two groupings, the first will obviously be the suspension forks and the second will be frames.

Now to divide the top queries by their hierarchy placement. we’ll look at the top two terms for each cluster to help this page not rank for bike parts.

Frames Page:

  • mountain bike frames – 1,000 estimated search/month
  • mountain bike frame – 590 estimated search/month

Frames > Brand Page:

  • surly frames – 140 estimated search/month
  • surly frame – 91 estimated search/month

Frames > Brand > Product Page:

  • banshee spitfire – 320 estimated search/month
  • (no other term high enough to list)

Forks Page:

  • mountain bike forks – 880 estimated search/month
  • mountain bike fork – 260 estimated search/month

Forks > Brand Page:

  • manitou forks – 880 estimated search/month
  • marzocchi forks – 390 estimated search/month

Forks > Brand > Product Page:

  • Marzocchi 44 Rlo – 210 estimated search/month
  • Marzocchi 44 TST2 – 170 estimated searches/month

What we’re seeing is that if we compare the meta-clusters, there’s more overall traffic to the forks vein than to the frames. Frames may sell for significantly more but with the larger total number of products in the suspension forks category overall and higher search volume for the lower clusters, it will win (if we only concern ourselves with the potential profit based on ranking). But which can we rank faster?

Assess Competitor Websites

You’ve compared and grouped your clusters by their potential ROI, you have an idea of how much traffic is there once you rank, now we need to figure out how hard it’s going to be to get there.

First, let’s compare the top-level phrases. To do this quickly I like to use a few tools. We’ll start with SEOQuake or a similar tool to quickly pull the PageRank of the top ranking sites and get an average.

For “mountain bike frames” there are:

  • 3 PR 4’s
  • 3 PR 3’s
  • 1 PR 2
  • 1 PR 1
  • 1 PR 0
  • 1 NA

For “mountain bike forks” there are:

  • 1 PR 5
  • 0 PR 4’s
  • 4 PR 3’s
  • 0 PR 2’s
  • 2 PR 1’s
  • 1 PR 0
  • 2 NA’s

While there is a PR5 for the forks phrase there are as many at the PR1 and below point as there are above, making it seemingly easier to get onto the first page. We’ll leave it at that for now and move to the next cluster.

Now let’s see what’s occurring in the branded clusters.

For “surly frames” there are:

  • 1 PR 6
  • 1 PR 5
  • 3 PR 4’s
  • 0 PR 3’s
  • 2 PR 2’s
  • 1 PR 1
  • 2 NA’s

Because of the significantly higher number of product-specific phrases in the forks cluster for “marzocchi” I’m going to focus in there as the cluster I choose to start an SEO project – where high search volumes and ROI possibilities across all the levels for the faster longtail ranking are available.

For “marzocchi forks” there are:

  • 1 PR 5
  • 0 PR 4’s
  • 1 PR 3
  • 2 PR 2’s
  • 4 PR 1’s
  • 2 NA’s

As with the higher level up, we’re seeing significantly lower competition in the forks meta-cluster. But to know for sure we have to go one level further to the product-specific phrases.

For “banshee spitfire” there are:

(Note: because of the lack of solid product phrases for “surly” we’re having to switch brands at the product level. It would take significant incentive in the form of extremely low competition to make this a good choice.)

  • 1 PR 4
  • 2 PR 3’s
  • 3 PR 2’s
  • 0 PR 1’s
  • 1 PR 0
  • 3 NA’s

For “marzocchi 44 tst2” there are:

  • 1 PR 4
  • 0 PR 3’s
  • 0 PR 2’s
  • 1 PR 1
  • 3 PR 0’s
  • 5 NA’s

Across the board the “forks” meta-cluster is lower in competition if we’re looking at PageRank. In this instance the competition is so consistently lower that it becomes optional to look to the next stage in quick competition analysis – that is to review both the homepage PageRank of the ranking sites in a similar fashion to what we’ve done above and also to use a tool like Majestic SEO or SEOmoz’s Site Explorer to review the linking domain counts for the ranking domains as well as the ranking pages. I prefer linking domains as opposed to backlinks as this naturally filters out aspects of links that can create misleading conclusions such as run-of-site links.

In performing this we would create lists as we have above (though on a large scale you should use databases or spreadsheets) with the alternative strength signals. On a large-scale this can get extremely time-consuming, so spot-checking a few of the higher and lower PageRank domains should suffice.

If your findings in the linking domain counts fall contrary to the strength indicated by the phrase-based PageRank breakdowns, and you can’t see anything awry in the backlinks of your competitors, either look for alternative clusters with more conclusive results or, if that’s not available, assume that the linking domains are more authoritative. Or, best of all, dig deeper into the backlinks looking for signals as to why the discrepancy exists (better use of anchor text for example).

What Can You Do With This Cluster Data Now?

In the example above it’s clear that the forks meta-cluster is the least competitive and with the search volumes and potential ROI possibilities, it becomes the superior of those looked at here. Now that we’ve locked in on what should be the main target, the next step is to actually do it (i.e. optimize the pages and build the links). Of course, it can’t be that easy.

A lot of this article has focused on the idea of meta-clusters and the strength inherent therein. We covered the “why” of that briefly above with the breadcrumb example however, before proceeding forward, a true understanding of how internal linking works, why clustering promotes the strength of sections of a website providing global ranking benefits faster and how to build the links to your clusters needs to be explored.

Resources

The 2023 B2B Superpowers Index

whitepaper | Analytics The 2023 B2B Superpowers Index

8m
Data Analytics in Marketing

whitepaper | Analytics Data Analytics in Marketing

10m
The Third-Party Data Deprecation Playbook

whitepaper | Digital Marketing The Third-Party Data Deprecation Playbook

1y
Utilizing Email To Stop Fraud-eCommerce Client Fraud Case Study

whitepaper | Digital Marketing Utilizing Email To Stop Fraud-eCommerce Client Fraud Case Study

1y