True Cost Estimations: In-State vs Out-of-State

🔹 About the Project

College affordability is one of the most important decisions for families and students. Institutions often publish reported costs of attendance (IPEDS), but these can be quite different from realistic expenses students face (EPI-based estimates: housing, food, healthcare, transportation, etc.).

Our project, True Cost Estimations, bridges that gap. We built dashboards and metrics that directly compare reported vs realistic costs, helping students, families, and policymakers make informed decisions.

🔹 What Inspired Us

We were inspired by the fact that students often underestimate the true cost of attending college. This mismatch in expectations can lead to increased debt, financial stress, or even dropping out. By quantifying and visualizing this “hidden gap,” we wanted to empower decision-making with transparency and data.

🔹 How We Built It 1. Data Preparation • Collected and cleaned the COST_ESTIMATION_FINAL_DATA.csv dataset. • Built calculated fields for differences, affordability ratios, and cost breakdown components. 2. Semantic Model (SDM) • Created key parameters (e.g., pTopN, pAbsDiffK). • Defined must-have metrics (e.g., Avg In-State COA, Avg Out-of-State COA, Affordability Ratios). 3. Visualizations • KPI Tiles: Avg In-State COA and Avg Out-of-State COA. • Comparison Bar Chart: In-State vs Out-of-State (EPI vs IPEDS). • Donut Chart: Cost component breakdown (tuition, housing, food, etc.). 4. Agent Integration • Metrics were named in plain English to enable natural-language questions like: “What is the average in-state COA?” “Compare IPEDS vs EPI estimates.”

🔹 What We Learned • How to translate raw cost-of-attendance data into metrics that an AI agent can interpret. • The importance of parameterization (e.g., thresholds for “Best Estimating” schools). • How to simplify visualizations so that non-technical users (students, families) can quickly grasp insights.

🔹 Challenges We Faced • Tableau Next did not support traditional “Measures” like classic Tableau, so we had to redesign workflows around metrics in the semantic model. • Mapping geospatial data (latitude/longitude) was harder than expected and not as intuitive in the Next environment. • Balancing simplicity (for judges) with analytical depth (for experts).

🔹 Next Steps • Extend affordability ratios to incorporate median wages post-graduation by program. • Build a recommendation feature: “Which universities are most affordable given a family’s income?” • Package the dashboards into a student-facing tool for transparency.

Built With

• agent
• apis
• cloud-services
• csv
• databases
• dmo
• einstein
• excel
• frameworks
• machine-learning
• or-other-technologies-did-you-use?-*-built-with-languages
• platforms
• salesforce
• salesforce-data-cloud
• salesforce-data-cloud-?-for-ingesting-and-modeling-the-dataset-(dlo
• semantic-model).-?-tableau-next-?-for-creating-metrics
• tableau
• tableau-next
• visualizations