Step 2

Step 2

Jesper Hansen - May 2, 2018




The next step is to determine which words and phrases are more important to the users, that our website should target. This is done by breaking down the most highly searched terms into single words, and assigning them a score that indicates their importance to the users.

From our listed search terms, individual words are extracted, comprising of each unique word that appears within these terms. We’ll of course have many terms gathered featuring ‘dress’: ‘wedding dress’, ‘summer dress’, ‘evening dress’, for example. There are six individual words in these three terms, but only four unique words: ‘wedding’, ‘summer’, ‘evening’ and ‘dress’. So this step essentially is removing duplicates, and breaking down all the terms and phrases into many single keywords.

Every word is allocated a score that indicates the importance in regards to usage by the users. The score is found simply by adding the search volumes of every search term phrase in which the word appears.



The resulting list of single words is then sorted into descending order of ‘importance score’, which gives us the first view of which keywords to use on the website in order to increase the relevancy – because this list consists of the words that users really use when they search, that we find relevant to the products or services that we supply.



But the list in itself is both long and contains information sometimes of little to no value for our purpose. So what we want is to translate the long list of single words into functional information that reflects the mind of the user. We do that by again categorizing the keywords into meaningful groups. These groups are decided upon based on visual inspection of our unique words and a knowledge of the industry, identifying associations between keywords with respect to common meaning and intent. We may also add groupings as we go along, if we find words that associate with each other that we didn’t expect to see.

For example, the list may include keywords like cheap, offer, price, discount and sale. These words all relate to pricing. When categorized according to user intent, therefore, the underlying theme of this group of words could be ‘price’.

Input from the website owner can be invaluable at this stage, due to their expertise in recognizing associations within their range of products or services, in order to form the groups. For example, in the field of fashion, ‘tights’ can mean ‘hosiery’ but can also be related to how loose an item is, so having industry input will ensure we don’t miss out words that could be assigned to particular groups, that we otherwise wouldn’t have realised. The words showing up in the keyword analysis may be quite different, but we should be able to fit them into one of these categories:


Some words are very central to the search endeavor and it’s impossible to deduce any specific meaning from the keyword itself – i.e. ‘dress’ or ‘jacket’, so these words we put into a group called ‘primary keywords’. Be aware though, that we should keep the amount of words in this category to a bare minimum, as we get no extra information from them.


Another big group of keywords fall into a group we call non-semantic. This group contains words that in and of themselves carry no contextual value, i.e. ‘it’, ‘and’, ‘if’.


This is where we form the real value of the analysis and add the crossfeed information between the user’s intentions and our knowledge of the ‘market space’ or industry field of the analysis.
Take care to be as accurate as possible, but also keep in mind that this is an iterative process and we can always regroup, make new groups or remove groups if it makes more sense at a later time.



It’s important to note that we should set some rules for how to group the keywords. Our experience shows that one good rule is that any given keyword should only be assigned into one group. If a keyword seems be able to fit to multiple groups, then that might be an indication of us not having set the grouping optimally, so we could then decide whether to add a further group.
It’s also worth noting that the groups that we create are not webpages, but potential subjects. So any given topic can be approached on more than one landing page at a time – this analysis will just show us how important users finds the various subjects and potentially how they relate to each other.



Having divided the most relevant keywords into groups, the subjects within those groups are then prioritized to gain further understanding of the user mindset.

By totaling the scores of all the keywords included in each group, and comparing these totals to the full total, relative importance is proportionally assigned to each group by percentage.

We can use the score percentage as an indicator of whether the groups we have made are satisfying our need for segmentation, i.e. if one single group has a relatively high share of the score (maybe 40% for one out of 10 groups), then that’s an indication of us needing to split that group into more groups.
On the other hand, we might have one group of 10 with a score percentage of 2% or lower, and that would indicate that either the group is lacks significance, or we should go and find more search terms that relate to that group.