Summary
-
Customer feedback was fragmented across multiple systems.
-
Teams needed timely insights within privacy and compliance requirements.
-
I developed a Python pipeline to analyze more than 100,000 customer responses per quarter.
-
Quarterly analysis fell from four to five weeks to fewer than three days, freeing analysts for complex investigations and supporting faster cross-team decisions.





Methodology & Approach
Working with the UX, Business Intelligence, and Sales teams, I developed a Python-based analytical pipeline to process large volumes of unstructured customer feedback.
I structured and cleaned incoming text, automated classification by issue type and urgency, and applied sentiment analysis, thematic clustering, and association mining to uncover recurring concerns, behavioural patterns, and product-demand signals. I examined behavioural evidence alongside the qualitative findings to distinguish isolated complaints from recurring problems and identify opportunities requiring further investigation.

.png)
Integrating data from multiple departments required navigating differences in systems, operational practices, and reporting expectations. I worked within these constraints and Medavie's privacy and compliance requirements to establish a consistent analytical framework, ensuring that findings were traceable, defensible, and useful to stakeholders with different responsibilities.
​
To accommodate both strategic and operational decision-making, I organized my standardized reporting into two levels. The quarterly overview (above, left) provided leadership with a consolidated view of feedback volume, comment types, urgency, and recurring associations across 12 categories.
​
The second level of topic modelling (above, right) examined recurring questions and concerns within each category, revealing the customer needs behind broader trends. This gave product, service, and operational teams a shared basis for distinguishing requests for guidance from process barriers, plan-design concerns, and service issues, while identifying where further investigation or coordinated action was warranted.
​
Notably, when findings warranted further investigation, the pipeline allowed authorized analysts to quickly trace patterns back to their underlying customer feedback. This preserved a clear connection between reported insights and source evidence, enabling teams to verify findings and investigate specific issues without repeating the full analysis.
Organization: Medavie Blue Cross
Role: Product Strategist
Project: Quarterly VOC analytics
Key Outcomes:
-
Reduced quarterly reporting turnaround from 4 weeks to fewer than 3 days.
-
Built a privacy-compliant Python pipeline processing 100,000+ long-form responses quarterly from call-centre transcripts, website surveys, and emails.
-
Created standardized customer intelligence for product and operations teams to independently access.
-
Freed analysts from hundreds of hours of manual data cleaning and organization, allowing them to focus on complex, high-priority cases requiring investigation and human judgment.
Deliverables:
-
Python-based customer feedback clustering and classification pipeline in collaboration with UX, Business Intelligence, and Sales.
-
Machine-learning classification framework covering internal categories of customer feedback, identifying recurring themes and associations between customer concerns.
-
Quarterly reports and standardized documentation, enabling cross-functional teams to evaluate findings and establish common priorities.
-
Self-service reporting resources in Excel, Power BI, and SharePoint, supported by standardized procedures and training materials that allowed colleagues to independently investigate findings, contribute interpretations, and coordinate improvement initiatives.
From Analysis to Organizational Capability
My pipeline reduced a four-to-five-week analysis cycle to fewer than three days, saving hundreds of hours of manual work each quarter. It also established a predictable reporting schedule, ensuring that quarterly customer intelligence reporting became routinely available.
​
I published categorized findings through Excel, Power BI, and SharePoint, enabling product and operations teams to independently investigate customer needs, emerging patterns, and recurring service problems.
​
These reports helped leadership evaluate existing processes, assess customer demand, and prioritize product and operational improvements based on behavioural evidence and customer feedback. The analysis also informed a broader enterprise-platform redesign initiative that contributed to an increase in administrator satisfaction from 6.8 to 8.1.

To sustain these improvements, I developed reusable analytical methods, documentation, and training resources that enabled colleagues to reproduce analyses and apply the reporting framework independently. By automating routine classification and reporting, I helped free analysts to concentrate on complex, high-urgency customer cases requiring investigation, contextual judgment, and individualized resolution.
Understanding the problem
At Medavie Blue Cross, customer intelligence was fragmented across contact centre transcripts, email correspondence, and website behavioural data. The contact centre alone generated more than 100,000 customer messages per quarter, yet identifying recurring needs, operational problems, and product demand required extensive manual analysis.
​
Producing consolidated quarterly insights took four to five weeks, often leaving stakeholders with findings that arrived after priorities had shifted. The challenge was not a shortage of information. It was the absence of a repeatable, privacy-compliant analytical capability to translate fragmented customer evidence into timely product and operational decisions.
My data flow diagram below provides a conceptual overview of the solution, tracing customer feedback from ingestion and preparation through classification, analysis, and quarterly reporting.

Summary
-
Customer feedback was fragmented across multiple systems.
-
Teams needed timely insights within privacy and compliance requirements.
-
I developed a Python pipeline to analyze more than 100,000 customer responses per quarter.
-
Quarterly analysis fell from four to five weeks to fewer than three days, freeing analysts for complex investigations and supporting faster cross-team decisions.






Methodology & Approach
Working with the UX, Business Intelligence, and Sales teams, I developed a Python-based analytical pipeline to process large volumes of unstructured customer feedback.
I structured and cleaned incoming text, automated classification by issue type and urgency, and applied sentiment analysis, thematic clustering, and association mining to uncover recurring concerns, behavioural patterns, and product-demand signals. I examined behavioural evidence alongside the qualitative findings to distinguish isolated complaints from recurring problems and identify opportunities requiring further investigation.

.png)
Integrating data from multiple departments required navigating differences in systems, operational practices, and reporting expectations. I worked within these constraints and Medavie's privacy and compliance requirements to establish a consistent analytical framework, ensuring that findings were traceable, defensible, and useful to stakeholders with different responsibilities.
​
To accommodate both strategic and operational decision-making, I organized my standardized reporting into two levels. The quarterly overview (above, left) provided leadership with a consolidated view of feedback volume, comment types, urgency, and recurring associations across 12 categories.
​
The second level of topic modelling (above, right) examined recurring questions and concerns within each category, revealing the customer needs behind broader trends. This gave product, service, and operational teams a shared basis for distinguishing requests for guidance from process barriers, plan-design concerns, and service issues, while identifying where further investigation or coordinated action was warranted.
​
Notably, when findings warranted further investigation, the pipeline allowed authorized analysts to quickly trace patterns back to their underlying customer feedback. This preserved a clear connection between reported insights and source evidence, enabling teams to verify findings and investigate specific issues without repeating the full analysis.
Organization: Medavie Blue Cross
Role: Product Strategist
Project: Quarterly VOC analytics
Key Outcomes:
-
Reduced quarterly reporting turnaround from 4 weeks to fewer than 3 days.
-
Built a privacy-compliant Python pipeline processing 100,000+ long-form responses quarterly from call-centre transcripts, website surveys, and emails.
-
Created standardized customer intelligence for product and operations teams to independently access.
-
Freed analysts from hundreds of hours of manual data cleaning and organization, allowing them to focus on complex, high-priority cases requiring investigation and human judgment.
Deliverables:
-
Python-based customer feedback clustering and classification pipeline in collaboration with UX, Business Intelligence, and Sales.
-
Machine-learning classification framework covering internal categories of customer feedback, identifying recurring themes and associations between customer concerns.
-
Quarterly reports and standardized documentation, enabling cross-functional teams to evaluate findings and establish common priorities.
-
Self-service reporting resources in Excel, Power BI, and SharePoint, supported by standardized procedures and training materials that allowed colleagues to independently investigate findings, contribute interpretations, and coordinate improvement initiatives.
Understanding the problem
At Medavie Blue Cross, customer intelligence was fragmented across contact centre transcripts, email correspondence, and website behavioural data. The contact centre alone generated more than 100,000 customer messages per quarter, yet identifying recurring needs, operational problems, and product demand required extensive manual analysis.
​
Producing consolidated quarterly insights took four to five weeks, often leaving stakeholders with findings that arrived after priorities had shifted. The challenge was not a shortage of information. It was the absence of a repeatable, privacy-compliant analytical capability to translate fragmented customer evidence into timely product and operational decisions.
My data flow diagram to the left provides a conceptual overview of the solution, tracing customer feedback from ingestion and preparation through classification, analysis, and quarterly reporting.
From Analysis to Organizational Capability
My pipeline reduced a four-to-five-week analysis cycle to fewer than three days, saving hundreds of hours of manual work each quarter. It also established a predictable reporting schedule, ensuring that quarterly customer intelligence reporting became routinely available.
​
I published categorized findings through Excel, Power BI, and SharePoint, enabling product and operations teams to independently investigate customer needs, emerging patterns, and recurring service problems.
​
These reports helped leadership evaluate existing processes, assess customer demand, and prioritize product and operational improvements based on behavioural evidence and customer feedback. The analysis also informed a broader enterprise-platform redesign initiative that contributed to an increase in administrator satisfaction from 6.8 to 8.1.

To sustain these improvements, I developed reusable analytical methods, documentation, and training resources that enabled colleagues to reproduce analyses and apply the reporting framework independently. By automating routine classification and reporting, I helped free analysts to concentrate on complex, high-urgency customer cases requiring investigation, contextual judgment, and individualized resolution.