Multicriterial parametric modeling of semantic phrases for online advertising: filtering and ranking algorithms

Context and Relevance. In online advertising, the success of campaigns largely depends on the quality of keyword selection. However, traditional approaches often rely on a limited set of metrics (such as frequency and competition), overlooking the complexity of goals like reach, budgeting, and localization. This paper proposes a scientifically grounded method of multi-criteria parametric modeling for filtering and ranking semantic phrases, aimed at optimizing advertising campaigns. Objective. To develop and demonstrate an algorithm for filtering and prioritizing keyword phrases in contextual advertising, taking into account multiple criteria: demand, competition, specificity, geo-dependence, and commercial value. Hypothesis. The use of a composite index that aggregates normalized indicators of popularity, competition, specificity, and cost-per-click enables the selection of more relevant and effective keyword phrases compared to traditional methods. Methods and Materials. The study was conducted on the semantic core of a furniture e-commerce website. Approximately 500 keyword phrases were collected, for which various metrics were calculated (search frequency, KEI, specificity index , CPC, geographic factor). A stepwise algorithm was applied, including filtering (removal of irrelevant, overly broad or rare queries, and phrases with high CPC) and ranking via a parametric model. An A/B test was conducted to validate the prioritization approach. Results. The final set of 120 keyword phrases covered approximately 85% of relevant search traffic while reducing the projected budget by 25%. The test confirmed that high-ranking phrases generated 1.8 times more clicks and 2.1 times more conversions for the same budget. Incorporating geo-dependence, specificity, and competition metrics significantly improved the accuracy of keyword prioritization. Conclusions. The proposed method effectively solves the task of semantic core optimization by increasing return on ad spend without losing relevant reach. The methodology is recommended for use in small and medium-scale online marketing. It is adaptable to other domains and can be extended to include conversion rate data.

1. Ampler, N., Lehmann‑Zschunke, N., & Olbrich, R. (2025). How to Design Keywords in Search Engine Advertising: A Multi‑group Comparison Based on the Search Volume of the Product Type. Review of Marketing Science. https://doi.org/10.1515/roms-2024-0020
2. Manning, C. D., Raghavan, P., & Schütze, H. (2008). Introduction to Information Retrieval. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9780511809071