Project Overview

Business Problem

The institution faced significant "Data Friction." Student records were inconsistent, with academic years stored in formats (e.g., "2023/2024") that made filtering and indexing nearly impossible. This led to:
• Identification Lag: Faculty could not easily identify students falling below attendance thresholds.
• Non-Unique Identifiers: Lack of distinct Student IDs made it difficult to link scores to individual demographics.
• Manual Overload: Staff spent hours every week cleaning data instead of analyzing it.

Business Goal

• Standardize Data Architecture: Split and normalize academic years and create unique Student IDs for every record.
• Visualize Critical KPIs: Provide instant visibility into Total Enrollment, Pass Rates, and Average Scores.
• Identify Behavioral Trends: Correlate study habits and attendance with final academic outcomes.
• Automate Reporting: Build a "set-and-forget" pipeline where new data (2025+) integrates seamlessly without manual intervention.

Data Workflow

• Automated ETL: Used Power Query to clean 3,000+ records, transforming inconsistent year strings into numeric formats and implementing a "Late-Typing" strategy to prevent data-type errors.
• Unique Indexing: Engineered a custom Primary Key system to generate unique student_id identifiers, ensuring data integrity across multiple academic years.
• Performance Logic: Authored DAX measures to segment student data, specifically flagging "At-Risk" students (Attendance below 70\%) and "High Performers" based on study-hour correlations.
• Executive Delivery: Built three high-resolution, interactive dashboard utilizing Custom Format Strings to display professional KPIs without breaking the model's underlying numeric logic.

Technologies Used

• Power BI: Data Modeling, Advanced DAX, and Interactive Visualization.
• Power Query (M): Advanced ETL, Column Splitting, and Data Normalization.
• SQL : Used for initial data profiling and validation of the raw 3,000+ records

Dashboard 1: Performance Overview

Key Insights

• Overall Academic Health Is Strong, But Not Excellent:
• Pass rate: 94.2% → institution is doing well at getting students through.
• Average score: 64.1% → this is mid-range, not high-performing.
• Translation: students are passing, but many are barely clearing the bar.
• Grade Distribution Shows a Performance Ceiling:
• A & B grades ≈ 60% combined, which is solid.
• C & D grades ≈ 34% — that’s a big chunk sitting in the “vulnerable” zone.
• F at 5.8% may look small, but at 3,000 students, that’s ~174 failures.
• Programs Perform Similarly — No Clear Standout Except One:
• Most programs cluster tightly around 63–65% average score.
• Data Science leads slightly (64.7%) and is flagged as Elite Performance.
• This suggests: Standardized grading rigor Or lack of program-specific academic innovation
• Coursework Beats Exams Across All Courses:
• In every course, coursework scores more than exam scores.
• Indicates: Exams are more difficult Or students rely too much on continuous assessment
• Risk: exam-heavy semesters could spike failures.
• Year-on-Year Performance Is Flat:
• 2023 vs 2024 average scores are almost identical.
• No visible academic growth over time → stagnation risk.

Recommendations

• Shift Focus From Passing to Excellence:
• Set a strategic KPI: Increase average score from 64% → 70%.
• Reward departments that push B → A conversions.
• Target the “C-Grade Majority”:
• Introduce tutoring or revision bootcamps for students scoring 55–65%.
• This group is the easiest to uplift with minimal resources.
• Audit Exam Difficulty & Alignment:
• Review whether exams truly reflect coursework objectives.
• Consider:Better exam prep, Mock exams, Exam skills workshops, Exam skills workshops
• Replicate Data Science Best Practices:
• Identify what Data Science is doing differently: Teaching methods?, Assessment mix?,Exam skills workshops,Student engagement?
• Roll that model into other programs.

Dashboard 2: Retention Dashboard

Key Insights

• Retention Is the Weakest Link:
• Retention rate: 63.3% → this is low.
• 1,100 at-risk students (36.7%) — that’s not a warning, that’s a fire alarm.
• Attendance Is the #1 Risk Driver:
• 1,512 students below 70% attendance.
• Attendance has a clear, direct relationship with performance.
• Even students with decent grades are flagged high-risk if attendance drops.
• Risk Is Consistent Across Programs:
• No program is immune.
• Risk rates hover around 35–40% everywhere.
• This confirms the problem is institutional, not departmental.
• High Grades ≠ Low Risk:
• Some A and B students are still marked high risk due to: Low attendance, Inconsistent engagement
• This is an early-warning goldmine.
• Year of Study Matters:
• Performance and attendance peak around Year 2–3, then dip.
• Likely causes:Burnout, External responsibilities, Reduced academic monitoring

Recommendations

• Attendance-Based Early Warning System:
• Auto-flag students when attendance drops below 75%, not 70%.
• Trigger advisor outreach immediately.
• Separate “Academic Risk” from “Engagement Risk”:
• Not all risks are about grades.
• Build dual risk scores: Academic (scores), Behavioral (attendance, consistency)
• Prioritize High-Risk High-Potential Students:
• Students with A/B grades but low attendance are salvageable.
• Personalized interventions here give the highest ROI.
• Year 3–4 Retention Programs:
• Mentorship
• Career-aligned projects
• Flexible attendance options
• These students don’t need basics — they need relevance.

Dashboard 3: Inisghts Dashboard

Key Insights

• Optimal Study Time Exists — And More Isn’t Better:
• Best performance occurs at b15–25 hours/week.
• Hustle culture ≠ smart studying.
• Attendance Is a Stronger Predictor Than Study Hours:
• Students with 80%+ attendance score ~15% higher.
• Study hours show weak correlation alone.
• Being present beats studying longer.
• High vs Low Performers Aren’t That Different in Time:
• High performers: 20.2 hrs/week
• Low performers: 18.8 hrs/week.
• Difference is quality, not quantity.
• Gender and Age Have Minimal Impact:
• Performance is consistent across: Gender, Age (18–26)
• This confirms: Fair assessment systems, No demographic bias
• Senior Students Perform Better:
• Year 3–4 students outperform juniors.
• Experience, adaptation, and exam familiarity matter.

Recommendations

• Teach Study Strategy, Not Study Hours:
• Promote: Active recall, Spaced repetition, Exam-focused revision.
• Stop encouraging “study longer” messaging.
• Make Attendance Non-Negotiable:
• Attendance policies are not administrative — they are academic tools.
• Incentivize 80%+ attendance with: Bonus marks, Priority access to resources
• Design for Sustainable Performance:
• Encourage 20-hour study models.
• Actively discourage burnout behaviors.
• Leverage Senior Students:
• Peer tutoring
• Mentorship programs
• “How to pass this course” sessions led by Year 3–4 students

Final Summary