Optimizing Google Keyword Performance for Psquared
Our Client
PSquared is a Canadian digital media agency that operates at the intersection of buy-side and sell-side advertising. They have developed a SaaS platform that allows media buyers to buy and sell advertising inventory at scale and generate revenue on the margin. In the ever changing and competitive landscape that is advertising, PSquared invests heavily in technology and research in order to support data-driven decision-making and to maintain their market edge.
Our Impact
- Provided a foundation for informed decision-making by exposing relationships between keyword features and their performance.
- Transformed insights into a set of comprehensive recommendations that boosted revenue by over 15%.
- Designed a machine learning model to predict keyword performance to an accuracy of 80%
- Developed a framework for deploying and iteratively testing the model in production.
The Challenge
PSquared’s SaaS platform allows media buyers to mass deploy Google Search campaigns but, until recently, keywords were being designed based on instinct. A desire to optimize keywords in order to maximize revenue emerged. The sheer amount of data and the need for sophisticated data processing prevented the problem from being solved in house.
Our Solution
By leveraging PSquared’s extensive historical data, our approach first aimed to uncover correlations between keyword features and revenue. In line with this strategy, the details of our approach included:
- An NLP-driven analysis to understand how the composition of a keyword impacted revenue.
Eg: Best 2024 SUV vs. Top 10 best value SUVs in 2024 Montreal - Analysis and identification of optimal keyword syntaxes to maximize revenue.
Eg: [noun][verb][noun][non][noun] vs. [adjective][noun][proper noun] - An industry-specific analysis to determine the most profitable industries and assess how various features impact keyword performance within each industry.
Eg: Best 2024 SUV vs. Cheap bahamas cruise April - An analysis of historic Google Keyword Planner (GKP) metrics to identify optimal ranges (e.g. CPC, average monthly searches) when designing new keywords.
This comprehensive exploration of historical data served as the foundation for our strategy. We identified intricate patterns and trends that likely would have been elusive through traditional analysis methods. We then leveraged the relevant features to design and train a machine learning model that accurately predicts keyword performance and can be used to design and deploy the most profitable keywords possible.