Portfolio
The optimizer picks the funded set under the budget cap to maximize expected NPV − λ·downside. Move the budget or the risk-aversion λ and watch the funding change — at higher λ the high-NPV trap gets cut.
Total capital committed 1200 of 1200 · objective (E[NPV] − λ·downside) -303 · λ = 1.0
Funded 4
| Project | Capital | E[NPV] | CVaR | Score |
|---|---|---|---|---|
| emissions-scrubber Emissions scrubber (compliance) |
200 | -195 | -250 | -445 |
| paint-booth-east Paint capacity — east line |
350 | 57 | -46 | 11 |
| plating-line New plating line |
500 | 103 | -40 | 63 |
| tank-monitoring Tank-monitoring retrofit |
150 | 67 | 27 | 67 |
Cut 4
| Project | Capital | E[NPV] | CVaR | Score |
|---|---|---|---|---|
| paint-booth-west Paint capacity — west line |
350 | 68 | -70 | -2 |
| power-upgrade Electrical service upgrade |
100 | -46 | -93 | -139 |
| robot-cell Robotic work cell |
400 | 121 | -21 | 99 |
| specialty-coating-line Specialty-coating line |
650 | 281 | -887 | -605 |
The project we cut: specialty-coating-line (Specialty-coating line)
had the highest expected NPV (281)
of any candidate — but a fat downside tail (CVaR
-887,
P(NPV<0) 0.30). We traded expected return for a
portfolio that survives a bad scenario.
The cut, on one screen
Highest-NPV project we cut (left) vs. the steadiest project we funded (right). The picture is the argument.
Investment memo
The optimizer decided; the model only explains the decision, grounded in the numbers above. You edit and sign.
Generating…