Author: Ekaterina Suvalova — Junior Data Analyst

Date: August 2025

Tools: Excel, Power BI (Power Query, DAX), GitHub, Notion

📌Contents

🔎 Executive Summary

🎯 Project Objective

📂 Dataset Overview

📐 Methodology

📊 Dashboard Insights

📝 Project Conclusions

⚠️ Limitations of the Analysis

🗂️ Project Files

🔗 References

⚡️Quick Access

🔗 Open on GitHub

⬇ Download all assets (.zip)

🎥 Dashboard GIFs

▶️ How to Reproduce

<aside> 💡

Key Insights

📚 AI has become a common study tool ⏱️ Short sessions are more efficient ⚙️ Practical tasks are completed more successfully than research ones

</aside>

🔎 Executive Summary

• This report covers 10,000 sessions of AI Assistant (hereafter — AI) usage by students over the course of one year. The main audience consisted of undergraduates — 60%, high school students and graduate students — 20% each.
• In 48% of sessions, students successfully completed the assignment; in 29%, they drafted an idea; in 16%, they got confused; and in only 8%, they completely gave up.
• 71% of students reused the AI tool after their first session.
• The highest efficiency was observed in sessions lasting 5–10 minutes; low-efficiency sessions involved 2–3× more prompts.
• Peaks of activity occurred on Fridays and Sundays, with the busiest months being August, June, and January.
• The most common Task Types were Writing (3.1K) and Studying/Homework Help (2.0K). Coding showed the highest share of successfully completed sessions — 60%.

🎯 Project Objective

Objective: To demonstrate a complete analytical process using an educational dataset: from data preparation and metric calculation to visualization and interpretation of results. The project showcases proficiency in Excel, Power BI (Power Query, DAX), as well as the ability to draw meaningful conclusions and articulate the limitations of analysis when working with synthetic data.

Key Questions:

1. Which groups of students use AI more actively?
2. How are sessions distributed across disciplines and task types?
3. How are sessions distributed across efficiency levels, and how do duration and number of prompts influence efficiency?