🏦 Housing Finance · Power BI · Python · SQL
Mortgage Program Performance Dashboard & Reporting Automation
Built an automated data pipeline and live dashboard system for Connecticut Housing Finance Authority — replacing a 3–4 day manual reporting process with overnight automation and real-time leadership visibility.
The Problem
CHFA's finance team was producing monthly reports by manually copying data from multiple systems into Excel. This took 3–4 days every single month. By the time the report reached leadership, the numbers were already outdated.
On top of that, there was no way to catch errors automatically. If something looked off in the data, nobody would know until someone spotted it manually — usually after the report had already gone out.
Leadership was making decisions about mortgage programs, housing market performance, and budget allocation based on information that was days old and occasionally wrong.
The Goal
Build a system where leadership could open a dashboard and see live, accurate numbers without waiting for anyone to manually pull data. Automate the overnight data refresh, catch errors before they reached any report, and cut the reporting cycle from days to hours.
My Role
I was the data analyst assigned to this project. I owned the full pipeline — connecting the data sources, writing the SQL and Python to clean and transform the data, building the Power BI dashboards, and setting up the anomaly detection alerts. I also worked directly with the finance team and program managers to make sure the dashboards showed exactly what leadership needed to see, not just what was technically available.
The Team
👔
Finance Manager
Defined which KPIs and numbers leadership needed — owned the reporting requirements
💻
IT Administrator
Provided database access and managed the technical infrastructure
🏠
Mortgage Program Manager
Explained what the mortgage data meant and validated that the numbers made sense in context
Tools & What I Used Them For
Outcomes
⚡
40% reduction in reporting time. What used to take 3–4 days of manual work now ran automatically overnight. Reports were ready before the team started their morning.
🎯
30% less manual review time. The anomaly detection caught data issues before they reached any report, which meant the team spent far less time hunting for errors after the fact.
📊
Leadership had live dashboards. Instead of waiting a week for stale numbers, they could open Power BI and see current mortgage performance, program spend, and housing market trends at any time.
🔄
Reporting process became repeatable. The pipeline was documented and automated so the process didn't depend on one person's manual effort every month.
What a Typical Day Looked Like
Morning
Check overnight data refresh — did it run? Any anomaly flags? If yes, investigate before the finance team's day starts.
Mid-Morning
SQL queries or Python scripts for upcoming reports. Or working through a backlog of dashboard requests from the finance team.
Afternoon
Meetings with finance team or program managers. Understanding what they need next, validating numbers, getting context on anything that looked unusual in the data.
End of Day
Update dashboards, fix any data issues that surfaced during the day, document changes so the next refresh picks them up correctly.