Turning data into profit
Turning data into profit
Retention and Outcome Analysis
I analyzed 10+ years of student retention data for a university program to understand what drives drop-off.
The tools: R, Tableau, and logistic regression.
The takeaway? ๐ฌ๐ผ๐ ๐ฐ๐ฎ๐ปโ๐ ๐ถ๐บ๐ฝ๐ฟ๐ผ๐๐ฒ ๐ฟ๐ฒ๐๐ฒ๐ป๐๐ถ๐ผ๐ป ๐ถ๐ณ ๐๐ผ๐ ๐ฑ๐ผ๐ปโ๐ ๐๐ฒ๐ด๐บ๐ฒ๐ป๐ ๐๐ผ๐๐ฟ ๐ฎ๐๐ฑ๐ถ๐ฒ๐ป๐ฐ๐ฒ.
Project Snapshot
Client: A university Digital Humanities program
Goal: Understand which student groups were most and least likely to complete a competitive minor, and identify opportunities to improve outcomes
Full Report
Code Repository
The Challenge
The client suspected some students were struggling to complete the program but lacked the data-driven insights to pinpoint which groups were most affected or why. With limited resources, they needed to identify high-impact opportunities for improvement.
Our Approach
We conducted an end-to-end analytics project, including:
- Data cleaning and integration across multiple sources
- Exploratory analysis of demographic trends
- Statistical modeling (logistic regression) to evaluate predictors of program completion
- Visualization of both aggregate outcomes and predictive insights to support decision-making
Importantly, we analyzed not just individual demographic factors but also how they interactโsurfacing intersectional risks and successes often missed in surface-level analysis.
The dashboard below is interactive so the data can be sliced in many different ways and viewed for different periods of time using the slider in the lower right corner.
Key Findings
Once I grouped by demographics and time constraints, hidden risks appeared โ just like in customer journeys.
- Overall completion rates were strong (77%), but certain subgroups lagged behindโparticularly students who were URM, female, and not first-gen.
- First-generation students completed at higher rates than their continuing-gen peers.
- A statistically significant interaction revealed that URM male students, despite initial assumptions, had some of the highest predicted completion probabilitiesโsuggesting effective, if informal, support mechanisms.
The same techniques apply to business:
โก๏ธ ๐ช๐ฎ๐ป๐ ๐๐ผ ๐ฟ๐ฒ๐ฑ๐๐ฐ๐ฒ ๐ฐ๐๐๐๐ผ๐บ๐ฒ๐ฟ ๐ฐ๐ต๐๐ฟ๐ป? โ ๐ ๐ผ๐ฑ๐ฒ๐น ๐ฑ๐ฟ๐ผ๐ฝ-๐ผ๐ณ๐ณ ๐ฟ๐ถ๐๐ธ.
โก๏ธ ๐ช๐ฎ๐ป๐ ๐๐ผ ๐ด๐ฟ๐ผ๐ ๐๐๐๐๐ผ๐บ๐ฒ๐ฟ ๐๐ถ๐ณ๐ฒ๐๐ถ๐บ๐ฒ ๐ฉ๐ฎ๐น๐๐ฒ (๐๐๐ฉ)? โ ๐ฆ๐ฒ๐ด๐บ๐ฒ๐ป๐ ๐ฏ๐ ๐ฏ๐ฒ๐ต๐ฎ๐๐ถ๐ผ๐ฟ, ๐ป๐ผ๐ ๐ท๐๐๐ ๐ฝ๐ฟ๐ผ๐ณ๐ถ๐น๐ฒ.
Impact & Next Steps
The program now has clear, data-driven direction for improving student success. Based on our recommendations, they are:
- Prioritizing early outreach to at-risk groups
- Investigating whatโs working for URM male students to scale support more broadly
- Monitoring program equity with newly developed KPIs
Why It Matters
This project demonstrates how looking beyond averages and digging into overlapping identities can unlock powerful insights. It also proves that predictive analytics can reveal unexpected strengths and help programs (or businesses) focus resources where they matter most.
Services Provided
- Data strategy and wrangling
- Predictive analytics (logistic regression, interaction effects)
- Custom visualization and reporting
- Strategic recommendations based on quantitative findings
Predictive analytics isnโt just for academia.
Weโll help you find whatโs working, fix whatโs not, and move forward with confidence.