Introduction
The study below tells a little about the work done with a Pareto client over a few months analyzing and restructuring Google Ads campaigns. The result was a very well-defined process, which today can be implemented in just a few weeks.
Clustered Keyword Management takes 3 factors into account:
- The fundamentals of the current Google auction;
- The user's journey to conversion;
- The main types of statistical optimization in the Search Network.
The Case Study - Segment: Education
In the first phase, Foundation, the aim was to raise the quality of the entire account structure. It turned out that the account was very fragmented. Not only in the distribution of the budget, but also in the huge number of segmentations. This made optimization much more time-consuming from a statistical point of view. In addition, the keywords were grouped into ad groups according to their semantic proximity, which mixed up different user profiles.
An example: the keywords "cursinho", "cursinho pré-vestibular" and "cursinho para enem" - all in the same ad group because they contain the word "cursinho" in their composition.
Even if different matches were used, it wouldn't be possible to optimize the account by funnel stage and use ads accordingly, because there are different levels of purchase intent and number of terms, all in the same group. Consequently, it was also not possible to optimize the ad groups by targeting different KPIs and budgets.
New STAG Structure
To change this scenario, a new structure was created, taking into account the entire user journey and the composition of keywords. An extensive study was carried out on historical data and, as a result, a unique account structuring system was created called STAG - Single Term Ad Group. This system, although it has the same acronym as the Single Themed Ad Group, has a very different structure - it comes from SKAG (Single Keyword Ad Group) and has proven to be more efficient and with better results for optimization. It uses a single keyword per ad group, allowing for highly efficient micro-optimization. The table below shows comparisons between the two structure models:
By considering two correspondences - broad modified and exact - it has also made it possible to accumulate data more quickly, to carry out cross-negativities between ad groups and to personalize ads to a high degree. When it comes to Machine Learning and paid media, faster data accumulation translates directly into budget savings, and can only be achieved through the use of automation tools. This leaner structure has made this possible.
The new structure, by stages of the sales funnel and using STAG, can be summarized as follows:
This separation was validated by historical performance in dozens of Pareto accounts, i.e. each cluster showed clear characteristics of the stage of the funnel in which it was allocated.
With an account format focused on the funnel stages, fewer campaigns were needed and fewer keywords were used. Then came the second and third stages of the strategy, which focused on the customer journey and account optimization through automation applied by Pareto.io.
Specifically for the top of the funnel, the new configuration made it possible to create new campaigns for this stage, working not only on sales, but also on lead generation and traffic. Having new campaigns 100% applied at another point in the consumer's journey was one of the key points for creating rapport with the user, increasing volume and the conversion rate at the bottom of the funnel.
Thus, for Awareness, CPM was set as the optimization indicator and, as success KPIs, the number of sessions, time spent on the page, average position, impression share, bounce rate and, most importantly, conversion assistance.
Algorithm performance
Every day, the Pareto.io algorithm automatically scanned and adjusted all these variables, trying to optimize each campaign for its objective. Something that couldn't be done manually.
In addition, the Quantic Sculptorwas developed, the automation responsible for analyzing the roots, affixes and suffixes of keywords and then adjusting the bid. An NLP System with Stemization and Lemmatization was created. Using it, it became possible to isolate the performance of each verb, preposition, adverb and noun in the account, using singular treatment via n-gram for more accurate dynamic negations, as well as enabling the "bidding" process by affinity clusters.
Such innovation has resulted in a process that feeds back on itself, increasing the level of optimization in the account and maintaining the granularity of adjustments at keyword level.
In this way, the increase in investment at peak times - which was as much as R$40,000 per day in the Awareness campaigns - was regulated and maintained intelligently. This allowed us to keep the top-of-the-funnel words in the best positions, generating high traffic and even a high volume of leads, especially in the periods closest to ENEM, where terms like "enem" before the exam and "gabarito" after the exam are extremely popular.
Conclusion of the study
Although the account structuring process was extremely important, it is known that the response time for results can be a little long. To increase performance in the short term, a differentiator in Pareto's optimization process was to compensate with remarketing and Google Analytics lists in Search, i.e. to bid differently for more qualified users.
Finally, this new arrangement of methods, structure and optimizations enabled the Search Network to act in a more integrated way with other Awareness campaigns used at the same time in the account, such as YouTube or Display campaigns. Using Pareto HotSpot - a tool for cross-referencing online and offline actions - it was possible to monitor the impact of Awareness campaigns on Search performance.
WOULD YOU LIKE TO HAVE ALL THIS IN YOUR ACCOUNT? TALK TO THE EXPERTS AT PARETO!
