Step 1

Step 1

Jesper Hansen - May 2, 2018



The only real touch point we have between the users and the website, where the users make clear indications of their intent, is in their search behavior. It’s well known that the better we are at matching the message of our results to the intent of the users, the more traffic we get – because search engines are so good at matching the intent of the user with the content of a website. So ultimately, we need to get closer to the user intent, in order to lift the relevancy of the messages on the website.

The first step in understanding the user behavior is to define the landscape in which our potential visitors can be found – thus we need to identify the more relevant search terms that are relevant to the website objective. Simply put, we need to list the search terms we expect the users to use, when they’re searching for the products or services our website provides.

The search term research is quite simple and aims to define the scope of the ambition for the website, filtering out irrelevant information in the process, to leave only the most relevant terms.



The first step is to gather as many search terms as possible that are relevant to the products or services our website offers, and thereby lay out a broad user landscape. The output of the first step is to have a two column list – one column with our long list of all possibly relevant search terms, and one column with the average monthly search volume of each term.

A good starting point is to describe our product or service offering, supplying essential personal insight into our specific market that adds to general industry knowledge. This step also usefully reveals the overall ambition of the website, through which search terms we deem as relevant. For example, one fashion vendor will select ‘dress’ as a potential search term, whereas another might find it too ambitious and competitive, and therefore select ‘flowered dress’.

A full list of words and phrases is eventually compiled that describes the products or services, their usage, specific features, etc.

The following steps elaborate and add to this initial list and there are various sources to find relevant search terms – we will address some of them here, but there are no specific limits to the sources you could use.

To expand on the list of search terms, we should take a look at what we have of visibility today and two good sources are available and highly relevant: Google Adwords and Google Search Console are both highly relevant.

The next step would be to use Google Keyword Planner. Having the massive data resources of Google at its disposal, this tool is probably the best tool available for gathering search phrases related to our given product or service. The tool is very good at finding top and mid-level generic search terms, as well as adding some branded searches into the mix, but it’s not very good at harvesting at the long tail spectrum of search terms.

So we need to turn to other sources for further expanding the list of search terms, particularly long tail search terms, using tools like, or other similar keyword research tools.

More specialized tools similar to are also valuable as they provide insights into user intentions in several different formats, that may add to the keyword landscape being created.
Finally, we have semantic sources, that can help us address other areas of the landscape, i.e. WordNet from Princeton University ( Similar tools can be found for various languages, such as for Danish (see a compiled list of wordnet websites here: The strength of these tools is that they return information based on the word alone, i.e. which words it’s contained in, the parents and children of the word, and similar information.

It quickly becomes clear that these sources will deliver a very high volume of search terms, so we’ll need to handle the fine tuning of relevancy at the next stage – at this stage the goal is to gather as much raw data as possible for further analysis.

As previously mentioned, the final output of this step should be a two column list, with the search term in one column and the average monthly volume in the second. So we need to return to Google Keyword Planner since we’ve harvested so many terms from various sources, to get the average monthly search volume for each unique listed term. We need to take our list, remove duplicates, and pull the average monthly searches back out from Google Planner.




So now we have a list of probably thousands of search terms (the number you gather will depend on the type of product and industry). We need a way to quickly categorize and analyze the search terms according to their expected ability to convert. This relevance sorting helps to grade the terms in order of significance, in preparation for the next steps of the keyword analysis. It involves creating search term categories and ‘tagging’ each term to attribute it to said categories.

We can create as many search term categories as we wish, but here we’ll just build four types to illustrate the point of the method. The search term categories should include, at minimum:

Search terms that include the company name, product brands or sub brands, or trademarked services.

Low funnel search terms that are directly related to the product or service. We expect them to have a relatively high CTR and CVR. They specify the main function of the product or service, without its branding. And if they also have a high volume, we refer to them as ‘money keywords’ or ‘primary keywords’.
Examples: To a fashion vendor, direct search terms could be: ‘buy dress online’, ‘blue dress with buttons down the front’, ‘jeans sale’.

Search terms that are further up the funnel, with lower CTR but still relevant to the product or service. This is not to be confused with long tail searches, as our related are in the field of our product or service, but don’t particularly specify that they are, compared to the direct search terms, of which some could be long tail.
Examples: To the same fashion vendor, related search terms could be: ‘fashion’, ‘what to wear’, ‘women’s apparel’.

Words that appear in the list as search terms but have no relevance to the target objective. These are terms that would be inserted into a Google AdWords campaign as negative keywords.
Examples: To our fashion vendor, we might add ‘table’ or ‘maxi’ as a negative, because we might have found search terms similar to ‘how to dress the table’ or for ‘maxi dresses’, which we might not sell.
The idea of filtering them out at this step is that if we know of them, we can easily add a new batch of search terms, without having to sift through and exclude these again.

Brand names are better at attracting and converting clicks than general, non-branded terms. Direct terms convert better than related search terms and, of course, negative words being the least significant, do not convert at all.

Remember that more categories can be added. You may want to add additional product brands, for example a department store may need this category to list all of the brands they sell, as this is separate to their own brand name. We are using four categories here as there’s a clear distinction on metric variation between these categories.

It would be very time consuming to attribute every word on our list separately, so we use a function to automatically tag them for us. For example, if we attribute ‘dress’ as direct, all search terms with ‘dress’ will auto populate as being tagged as direct. However, we might have the search term ‘wedding dress’ which should sit in the related category. So we can add ‘wedding’ to related, and all search terms that contain ‘wedding dress’, or ‘wedding’, will then be over-ridden and be categorized as related. Similarly, if we use our previous example, we might want to exclude any terms mentioning ‘maxi’, since we don’t sell maxi dresses. So if we add ‘maxi’ to our negative column, all terms containing ‘maxi’ will be tagged as negative keywords, even ‘maxi dress’ which was previously in direct because we tagged ‘dress’ as direct. This method means we don’t have to categorise thousands of search terms manually. It usually only takes around the first one hundred to be attributed into our categories for the rest to be categorized by association.