01 — Problem
Program Decisions Made on Gut Feel
Every time a new workforce education program was proposed, the same conversation happened: someone estimated enrollment, someone else guessed at costs, and the decision to proceed was based on institutional enthusiasm rather than arithmetic. I watched 3 programs launch in a single year that never reached break-even — not because the market wasn’t there, but because no one had modeled the sensitivity of margin to enrollment variance. A 15% shortfall in expected enrollment turned a profitable program into a loss leader, and no one saw it coming because the assumptions were never stress-tested.
I needed a tool that made the financial assumptions visible and interactive. Not a static spreadsheet that someone built once and never updated — a modeling engine where changing one input instantly reveals the downstream consequences.
02 — Architecture
Inputs, Models, Scenarios
The engine is a Streamlit application backed by a Pandas computation layer that models program financials across three dimensions:
Input Layer
Users configure tuition per student, enrollment targets, instructor costs, facility overhead, revenue-sharing percentages (for employer-sponsored programs), and per-student variable costs. Every input has a labeled range and default value derived from historical program data. The interface prevents unrealistic inputs — you can’t set a revenue share above 100% or a negative enrollment target.
Computation Engine
A Pandas-based service calculates 7 core KPIs: gross revenue, total cost, net margin, break-even enrollment, margin per student, ROI percentage, and payback period. All calculations use vectorized operations across enrollment ranges (50%–150% of target) to generate sensitivity curves rather than single-point estimates.
Scenario Visualization
Plotly renders interactive charts: break-even crossover plots, margin sensitivity curves across enrollment ranges, and revenue waterfall diagrams showing the composition of income and expenses. An auto-generated executive summary translates the numbers into plain-language recommendations — “at current assumptions, the program breaks even at 18 students and reaches target margin at 24.”
Key Design Decisions
Why sensitivity ranges instead of single-point estimates? A single break-even number creates false confidence. Showing that break-even shifts from 15 to 22 students when instructor costs increase 20% communicates risk in a way that a single number cannot. Program directors responded better to visual ranges than to precise but brittle projections.
Why Streamlit instead of a full web application? The audience is 3–5 internal stakeholders, not a public user base. Streamlit’s rapid development cycle (build to deploy in hours, not weeks) was the right tradeoff. The tool needed to exist quickly and evolve based on real usage, not survive a product launch.
03 — Outcomes
Measured Results
Financial KPIs
computed in real time from user-configurable inputs
Scenario Types
break-even, sensitivity, and revenue waterfall analysis
Assumption Visibility
every input exposed and adjustable — no hidden parameters
Recalculation Time
full model recomputes on any parameter change
04 — Reflection
Making Assumptions Visible Changes Decisions
The value of this tool was never the math — break-even calculations are arithmetic, not engineering. The value was in making hidden assumptions explicit. Before the engine, program proposals contained sentences like “we expect strong enrollment.” After, they contained charts showing what happens at 70%, 100%, and 130% of target enrollment. The conversation shifted from opinion to evidence, and two proposed programs were restructured before launch based on what the sensitivity analysis revealed.
What I’d change: the executive summary generator uses hardcoded thresholds for “healthy,” “marginal,” and “at-risk” classifications. These should be configurable per institution, since a 15% margin might be excellent for a community college program but insufficient for a private university’s cost structure.
“The most dangerous number in a financial model is the one nobody questions. The tool’s job isn’t to produce answers — it’s to make the assumptions impossible to ignore.”
Outcomes
7 financial KPIs computed in real time; 3 scenario analysis types; 100% assumption visibility; Sub-2-second recalculation on parameter change