BUHI Supply Co. — Conversion Optimization
Python EDA • CPA Reduction • Conversion Rate Analytics
Conversion rate increase
15.9%
vs. inform bid baseline
CPA reduction
12.98
While perserving volume
Phrase vs broad CR
21%
Higher at similar bid levels
Bid strategy winner
Manual
Outperformed automated
OVERVIEW
The second phase of the BUHI Supply Co. performance optimization strategy — shifting the objective from maximizing clicks to maximizing conversion rate, a more downstream and business-critical metric in the digital marketing funnel. Python was integrated into the analysis pipeline to evaluate bidding scenarios, identify high-ROI keyword opportunities, and reduce inefficient spend tied to low-converting keywords.
This project builds directly on the Click Optimization case study — together they represent a full-funnel campaign analytics workflow from traffic acquisition to conversion.
WHAT I DID
Data exploration & preparation
Loaded and cleaned the Stukent conversion dataset using Pandas. Standardized data types, removed null values, and grouped keywords by intent and performance metrics.
Conversion rate & key metrics
Computed conversion rate (conversions ÷ clicks) across keyword groups, match types, and ad settings. Calculated CPA, CTR, and conversion yield by match type for comparative analysis.
Exploratory data analysis
Compared match type performance (exact, phrase, broad) on conversion rate, CPA, and profitability. Investigated underperforming keywords with high CPC and low conversion rates.
Optimization & visualization
Built charts showing match type vs conversion rate, CPA by keyword, and cumulative conversion volume. Compared manual vs automated bid strategies to identify the most stable approach.
OUTCOMES
-
Increased conversion rate by 15.9% through optimized bidding strategies and keyword match type selection
-
Lowered average CPA by 12.8% while preserving or improving overall conversion volume
-
Phrase match keywords outperformed broad match by 21% higher conversion rate at similar bid levels
-
Manual bid adjustments produced more stable conversion gains than automated bidding on this dataset
-
Identified and paused low-converting, high-CPC keywords — reducing wasted spend and improving overall ROAS
KEY LEARNING
-
Conversion rate is a more meaningful optimization target than clicks — traffic volume without conversion efficiency is wasted spend
-
Phrase match consistently outperforms broad match for conversion-focused campaigns — specificity drives intent alignment
-
Manual bid strategy provided more predictable and stable results than automated bidding in this controlled simulation environment
TOOLS USED
Full code and analysis avaliable on GitHub