Keyword clustering for smartbidding marketing campaign management

Ready Made

REACH-2021-READYMADE-JOT_1

JOT

JOT Internet Media is part of Cube Ventures, one of the leading digital groups in Europe. Based in Madrid and with offices in Germany, Mexico, Brazil and Italy and more than 280 employees, the group is focused on lead generation, services monetization, digital marketing, media and investments.

The goal of the challenge is the development of a service enabling the selection of the most relevant keywords based on financial criteria and the generation of a semantically related keywords list.

The impact and profit of the digital marketing campaigns depends on the relation between the society interests and the content displayed in the ads. In addition, search ad platforms have evolved the optimization process from a Quality and CPC approach to a smartbidding one, where the publisher determines the general budget and the expected return of investment but has no direct control on the price per click, which is directly determined by the search platform algorithm.
In this scenario it is very important to have knowledge about the expected profitability of the keywords as well as the generation of a homogenous group of keywords to facilitate and improve the price calculation. If this is not the case, it can happen that the publisher invests a lot of money in a low quality keyword because it is embedded in the same ad group as a top one, which will reduce the overall profit of the campaign.
Thanks to the development of an intelligent service enabling the identification of the top keywords and the generation of keyword groups based on semantics and financial criteria will improve the quality of the campaign, being positioned in the top places in the ranking at lower prices.

Expected results is a SaaS prediction service enabling:

• Classification of keywords based on multiple criteria
• Algorithm should work in English, Spanish, French, German and Portuguese
• Train the model with millions of entries
• Models and classificatory with errors R2>0,75 to be integrated in production