- 1. What you need to consider
- 2. The most popular services for collecting semantics
- 3. Other ways of collecting semantics
- 4. Clustering requests
Targeted user requests are the main criterion for the result of contextual advertising. If one of them went to the site and is ready to buy a product or order a service, it means that advertising has paid for itself and can bring profit. In order for it to attract customers to a web resource, it is important to select the phrases that potential customers enter into the search engine as accurately as possible. In other words, to compose the semantic core. Almost all PPC campaigns begin with this important step.
Thus, the semantic core is a list of key phrases that initiate the display of an ad to an interested user. The success of contextual advertising depends precisely on understanding potential customers and on the ability to predict their requests. The availability of information on high-frequency queries will entail an influx of traffic to the site.
Benefits of well-composed semantics:
- budget savings (no inappropriate clicks);
- timely addition of unnecessary requests to stop words;
- maximum conversion effect.
1. What you need to consider
The most important parameter for collecting the core is the number of searches for a keyword. The frequency of a request depends on its coverage and subject, and the higher its rate per month, the more users are looking for this request.
So, the keys of the primary list of semantics will probably contain the name of the product/service, segment, action, price, urgency and location. With the help of the correctly selected parameters, it will turn out to show a more relevant ad to the target audience.
There are 4 stages of collecting semantics:
- Obtaining keywords using third-party services and search suggestions.
- Removing the frequency of keywords.
- Analysis of competitors’ results.
- Final clustering.
To compose the semantic core, it is better to use professional tools that show the actual requests of users. With their help, it will be possible to select advertising offers to get the maximum targeted conversions to the site.
2. The most popular services for collecting semantics
- «Keyword Planner» Google Ads ‒ It is located in the “Tools & Settings” section. You can pick up requests in it, either by entering them manually, or by specifying the site.
Fig. 1 – Keyword planner in the interface Google Ads
- Key Collector is a good tool for working with cores from 500 requests. It is not free software, but its functionality is wide enough. After getting the list from the program, you need to get rid of duplicates and remove inappropriate queries by placing them in the list of negative keywords.
Fig. 2 – Interface Key Collector
- Plugin for popular browsers ‒ Wordstat Assistant. Suitable for kernels up to 500 phrases. Free, allows you to select keywords from Wordstat in a semi-automatic mode.
Fig. 3 – Interface Wordstat Assistant
You can also use the Google Search Console Performance tool. This allows you to analyze queries by metrics such as Clicks, Impressions, and Positions.
3. Other ways to collect semantics
You can supplement the results obtained in the programs using search hints.
They are easy to spy on Google:
Fig. 4 – Google search suggestions for the query
In addition, keywords can be found in services like Serpstat, by going to Keyword Analysis → Search Suggestions.
Fig. 5 – Analysis of search suggestions using the Serpstat service
For a better understanding, you can analyze competitors of the same topic, paying attention to the requests for which they show their ads. You can also do this manually, or you can run the domain name in special programs and take into account the displayed search results. For example, use the Ahrefs service.
Fig. 6 – An example of analysis of competitors’ search queries by Ahrefs
It is possible to combine all received queries with each other in various combinations, for example, using PPC-help the phrase generator.
Fig. 7 – PPC-help interface start page
If there are too many requests, the first step is to cut off all low-frequency ones.
In addition, you can either put the keys in quotes and get traffic only for these words, or collect the maximum number of negative keywords and leave only frequency keywords to avoid clicks on non-targeted queries.
4. Clustering queries
By dividing phrases into thematic subgroups, you will carry out the so-called clustering. The goal is to prevent falling into the same group of queries, the promotion of which on one page will be ineffective.
The peculiarities of clustering for contextual advertising are that key phrases have a deeper and wider range. The frequency indicator in this case is given a lower value. The assembled semantic core is clustered based on which campaigns it will be divided into. It is the structure of the semantics of the account that influences ad management.
It is easy to differentiate clusters like this:
- high conversion;
- medium conversion;
- low conversion;
- transactional (with selling language);
- garbage (spam).
There are various options for clustering queries: hierarchical (for example, by location or by category) and non-hierarchical (by conversion, by negative keywords or the number of words in a phrase, etc.). You can break your campaigns into subcategories, or you can break them down into clusters of the ratio of profit per product to conversion.
- different bids for different campaigns – a strategy for stores / not stores (the second is suitable for launching any services);
- all keys are merged into one campaign, no preference, and often at the same rate – this method is suitable for test campaigns or for a single page site.
Clustering can be done manually, automatically, or a combination of both. The first method is the longest, but reliable, the second is the fastest, but not always correct. The third one is optimal, because it combines the advantages of the first two: it is better to start with automatic and end the check manually.
As a result, after receiving data on conversion in good clusters, it is worth expanding the semantic core, and in bad ones, accordingly, you need to clean up the semantics. Despite the length of the procedure and the need to constantly monitor changes, the work done pays off. It is optimal when each cluster has its own rate and strategy.
Based on the results of the collection, the semantic core for contextual advertising should be as complete as possible to include all the necessary queries, and at the same time exclude all words that turn out to be inappropriate clicks. It is in this case that the advertising campaign will be effective and bring the expected profit.