BrightonSEO 2024 presentation

April 24, 2024

News

Scaling keyword research and making sense of the data

Voxels CEO Daniel Axelsson had the opportunity to speak about keyword research and keyword analysis at scale at BrightonSEO. In this blog post we’ll cover his talk.

Read about the talk below or download the PDF.

First thing first, it’s really easy to collect a lot of keywords – simply exporting keywords from search consoles API will give you tens of thousands of keywords, when we include keywords from other sources as well it tend to stack up and become quite overwhelming.

Naturally, having a lot of keywords will make it harder to make sense of them and find valuable insights from them but also presents challenges on how you analyze and present this amount of data so it makes sense.

Making sense of big data sets is challenging for different reasons:

Information overload
Lack of context
Ambiguity

And when we add multiple data sources to the equation we complicate it even further since this:

Compromises accuracy
Increases complexity
Hampers decision making
Limits understanding of audience behavior

But ultimately it reduces confidence in the keyword analysis which is the last thing we want.

This is more common than you would like to think, and I was guilty as charged as you can see in the slides above, comparing apples to oranges, to put it mildly. This led me on a journey to standardize our agenesis internal workflows which went on to become “Pixel Workflow” and quickly became an everyday tool for the consultants at Pixel and if we fast forward a few years more Pixel Workflow became Voxel when we decided to open up our internal tool to the public.

So this is the reason why we designed and built a tool like Voxel in the first place. Voxel allows you to unify your keywords from different sources (and source new keywords ofc) and streamlines the process in which you clean, de-duplicate, organize, cluster and analyze your search data.

As you can see in the image above Voxel is a necessity if you which to simplify you keyword research and keyword analysis process since it helps you organize and visualize your data to help you understand patterns and relationships between keywords in an easy 3 step process:

Organize – Break down the keywords into smaller groups.
Visualize – Charts, graphs, and other visual aids can help you interpret.
Context – Understand patterns and relationships between keywords.

Insights & visualizations from Voxels Keyword analysis

We’ve decided to divide the keyword analysis into the following sections:

Market demand
Search behavior
Site insights
Site potential
Competitive intelligence

We start with breaking down the overall market demand to better understand the scope of the keyword analysis and what it entails.

Starting with the amount of keywords in the project [16,147 keywords] and then we introduce the demand these keywords represent [23,823,190 monthly searches] and lastly we strive to take it down a notch and concretize these numbers by showing a potential traffic you could achieve by investing in SEO [2,340,390 traffic potential]. The traffic potential is dynamic and you can adjust it to accommodate your needs, I went with 2,3 million because I looked at the estimated traffic for the best performing domain on this set of keywords and added roughly 10% to their current performance.

Next we are looking at the demand over time to help us understand seasonal aspects within the dataset. As we can see in this specific case we have the lowest demand in February and the highest demand in November but overall it’s quite evenly distributed throughout the year.

If we zoom out even further and look at how the search demand has evolved over the last 3 years, we can identify market trends such as +9% growth in demand over a 3-year trend. In this case we can see that the search demand in United Kingdom on Google for fashion related keywords is on positive trend and that the fashion industry is an growing market.

When we’ve a solid understanding of the market demand we can start looking at search behavior to understand the demand at an even deeper level.

We’ve already established that this keyword analysis is for fashion but more specifically it’s for product related fashion keywords. And when I conducted the keyword research I decided to focus on product categories and relevant combinations to these. As seen in the data above we can see a demand breakdown of different topics that indicates how popular these topics are in combination with product categories.

Color is a clear winner with 9,1 million In search demand followed by profile at 7,7 million and Material at 4,7 million. Then we can find pattern related keywords at 1,1 million in demand followed by seasons, sizes, occasions, power words and sustainability which all have below 1 million search per month, on average.

This slide shows the most frequent combinations within the dataset in terms of amount of keywords and amount of search volume. One clear takeaway is that [colour + products] related keywords are the most prominent both in terms of the amount of keywords (13%) but also the in terms of demand (22%). Another interested finding is the combination [Product + Profile] which is the second biggest combination in terms of demand, accounting for 17% of the total demand but only 4% of the amount of keywords which indicates that the product + profile is made up by head keywords.

If we zoom in at our topics we can easily identify the underlying sub topics and the demand each sub topic is associated with. If we look at the topics colour we can see that it accounts for 38% of the total demand with the following sub topics and their demand:

Black [2,7 mil]
White [2,1 mil]
Pink [700k]
Red [660k]
Green [630k]
Blue [590k]
Gold [360]
…and 11 more colours…

If we do the same exercise here we can see the percentage of the demand each of the topics represent and the top volume keywords within each topic. To keep it consistent let’s look at the top keywords within colours and their search volume:

White dress [110k]
Black dress [110k]
White trainers for women [91k]
Red dress [74k]
Men’s white trainers [74k]
Pink dress [74k]
White trainers men [74k]
…and 7,475 more…

As always we can drill down even further if we like as seen in this case where we’ve drilled into the sub topics [Colour] on the left and [Material] to the right in the image above. As you can see we can see the different sub topics with their associated keywords, and the search volume for those keywords.

To wrap up our search behavior section we look at the specificity among the keywords in our dataset. This give us a good indication how our keywords are divided between:

Head keywords (1-2 words)
Body keywords (3 words)
Long tail keywords (4+ words)

We know that this is an oversimplification of head, body and long tail but it still gives a good idea about how the audience is searching and the demand associated withing each bracket. One takeaway from here is that the long tail volume comes in at 2,3 million searches per month (i.e. 4 or more words per keyword). A lot of SEO’s underestimate the value a solid long tail strategy can bring and here we can see, black on white, that it’s entirely possible to build a fashion business on long tail keywords alone.

When we have a good understanding about the overall market demand and the search behavior associated to that demand we will move on to site insights to better understand how your domain perform across these keywords. For this talk I selected H&M since they are a Swedish brand with global presence.

First we look at a position spread pie chart to better understand how the H&M domain are performing as a whole. We can quickly see that they rank on 8% of the keywords within top 3, 13% between position 4-10 and 14% between position 11-20. In other words this means that 65% of the keywords are outside top 20.

If we drill down and look at the position spread within the same position intervals as before but across different topics we get deep understanding about how H&M are performing across colour, profile, material, pattern, seasons, sizes, occasion, power words and sustainability related keywords.

As the graph clearly shows we can see that H&M has the best visibility within pattern related keywords and the worst visibility within sustainability related keywords (chocker right).

Here we simply plot out the rankings on top volume keywords within each topic. The colour coding gives a good indication of how H&M performs on the most important keywords. Green is good, Red is bad and by the looks of it H&M has an average to good performance.

Here we tie the whole keyword analysis together and reestablish the narrative Voxel has through it’s SEO insights by yet again introducing keywords, monthly searches and estimated traffic keeping the same concept we introduced the whole keyword analysis around within market demand.

But this time we don’t look at the overall market potential, instead we look at how H&M stack up towards the overall market potential. And as we can see in the graph above H&M rank on 22% of the keywords (3,489 of 16,147), they have a visibility coverage of 30% (7,175,440 of 23,823,190 monthly searches) and they get 735,988 estimated clicks our of 2,340,390 which shows that H&M have lots of potential.

When we’ve established the current performance we move on to the potential to help stakeholders better understand the value an SEO investment could yield in.

Here we kick off lying out how much traffic potential each topic has and by declaring how much estimated traffic we currently have so we can easily identify important topic for H&M to focus on.

Next we want to understand the traffic potential across different positions. We have the sum of volume on the left, ranking positions in the bottom and the number of keywords to the right. Ideally we want to find a big gap between the bars (# of keywords) and the line (sum of volume) in the graph for different positions. As we can see at position 18 where we have approx 200 keywords with a combined search volume of 400k.

Keep in mind that we are looking at aggregated insights across a keyword portfolio consisting of 16,147 keywords. So, to better understand the potential tied to position 18 for H&M I browsed through our different topics.

When I arrived at the topic occasion we could identify a clear outlier at position 18. For H&M that means 16 keywords with a combined volume of 95k search per month, indicating huge potential! Naturally we klick on the bar for position 18 to find out more.

When we click at the bar Voxel takes us to a pre filtered view in the table within the keyword section.

The table is filtered for H&M’s rankings on position 18 and the topic occasion.
We can see that 95% of the potential is tied to the keyword “Party dresses” which has 90,500 searches per month on average.
Voxel automatically declares the peak quarter for each keyword (If there is one) and lets you know which month the keyword has its peak demand.
And if we look to the right we can see that the keyword party dresses has a quarter of a million searches in the UK in November.

As seen in the bubbles above Voxel plots out monthly searches and the potential traffic. We get both the current estimated clicks and the potential. Let’s assume a conversion rate of 1,4% and add that in and we get…

Great, now we can see the amount of conversions we would get with our current estimated traffic and also good understanding of the conversions we would be getting if we improved our organic traffic. Let’s take it one step further and add in an average order value of $31(assumed) and…

Amazing, now we have a proper business case of what a proper SEO investment could mean in actual numbers.

Sometimes you need to present data over time to better communicate what the journey would look like and Voxel allows you to simply plot your potential out over time and compare it to current state labeled “business as usual” to illustrate the growth potential over time both in terms of traffic and revenue.

To wrap the keyword analysis in Voxel up we look into the competitive landscape to better understand how we stack up against the competition.

We introduce a classic scatter plot chart with estimated monthly traffic on the left hand side and the amount of keywords at the bottom. At this graph we can easily see how H&M stack up against the competition and we can also see that Asos is best in class driving over 2 million in estimated traffic, per month.

An alternative way to present this data is to plot it out as we introduced it at the very beginning within market demand illustrating how many keywords each competitor rank on, followed by how much visibility they cover and wrapping it up with the estimated traffic each competitor gets.

Out of all the insights, charts, graphs and visualizations we’ve gone through in the blog post this is my absolute favorite, the color scale is red (bad) to green (good). The numbers represent the estimated traffic each domain is getting across each topic.

Looking at the competitive intelligence this way provides so much insights and depths.