Where does the next marketing dollar buy the most qualified starts?

Your analytics team holds the real per-brand spend, so a pure cost-per-start ranking is something they can already pull. What's structurally hard from inside the org chart is a recommendation that's allowed to conclude "shift spend out of the brand we grew fastest," and one drawn inside the regulatory box rather than around it. A pure-ROI optimizer says pour everything into the cheapest funnel; the rules say you can't, because which students you recruit carries a 90/10 and Gainful-Employment dimension, and every earnings claim in the creative is FTC-substantiated. This tool prices that gap: it shows the unconstrained optimum, then shows how many starts the constraints actually cost, so the conversation is about the real trade-off, not a fantasy one.

• Curve shape: diminishing-returns (logarithmic) response of qualified starts to spend, the standard media-saturation shape. Synthetic, calibrated so each brand's relative efficiency matches the public signal: Walden's online-grad funnel competes on auction-priced keywords (against WGU, SNHU, Capella), so its marginal cost-per-start runs higher; Chamberlain's nursing demand is pulled by a documented workforce shortage, so its early dollars convert cheaper.
• The Q1 FY26 anchor: management said Chamberlain "underperformed in local marketing effectiveness" and funnel conversion "fell below benchmarks" while Walden grew 13.6%. That's a real, public divergence in marketing physics; the curves reflect it, they don't invent it.
• 90/10 constraint: proprietary institutions may draw no more than 90% of revenue from federal Title IV funds; the constraint caps how far spend can tilt toward the most aid-dependent funnel.
• GE constraint: programs near a Gainful-Employment debt-to-earnings line can't be aggressively marketed; the GE-safe regime freezes spend on the flagged slice.

What this is not: these are not Covista's real response curves: the company discloses neither per-brand spend nor cost-per-start. The shapes are synthetic and labelled. The method (an independent, compliance-aware frontier) is the deliverable; calibrating it to your real curves is what the engagement scopes.