Prototype · grand-canyon-education · public + illustrative data · noindex

The next marketing dollar: cheap inquiry, or enrolled Start?

GCE spends $212.4M a year on marketing (FY2024 10-K) to fill an enrollment funnel against five mega-competitors. Split that budget across four channels under per-vendor ceilings (the "boast") and a hard cap (the "bound") — then flip the objective from cost-per-Inquiry (what a channel manager is paid against) to cost-per-Start (what the institution is paid against). Watch the plan, and the money, move.

$212.4M

FY2024 "Marketing and communication" expense — LOPE 10-K (the budget split below scales to a quarter of it)

79%

GCU students enrolled exclusively online — IPEDS Fall 2024 (UnitID 104717)

51,203 → 7,281

Applicants → first-time enrolled, IPEDS Fall 2024 — the funnel shape the start-rates are tuned to

Try a scenario — one click sets the plan

Fine-tune (optional)

Objective

CPI = cost-per-Inquiry, the number the channel manager is paid against. CPS = cost-per-Start, the number GCE is paid against (it earns 60% of GCU tuition only when a student actually enrolls). The finding: a CPI-optimal plan is CPS-pessimal.

Spend cap — the bound

Hard upper bound on the quarter's spend. Default ≈ $53.1M — one quarter of the FY2024 $212.4M marketing line. "No increase in total cost."

Boast headroom — the ceilings

Flat ceiling each channel may grow over last quarter. A flat % is a hack — the honest version is a per-vendor number (some vendors +50%, some −20%) negotiated in a week of calls. That negotiation is the real engagement.

Recent-quarter weight

Multi-period weighted scorecard; the rest splits 5 / 10 / 15 across the older three quarters. A politics tool dressed as a statistics tool — you need four bars to point at when the CFO asks why.

Channel	CPI $	CPS $	Plan $	vs base

Sources & method

$212.4M marketing line — "Marketing and communication" expense, FY2024, from the GCE (LOPE) Form 10-K consolidated income statement (SEC EDGAR, LOPE 10-K). The spend cap defaults to one quarter of it (~$53.1M). REAL.
79% exclusively online · funnel 51,203 → 40,450 → 7,281 — Grand Canyon University, IPEDS Fall 2024 via NCES College Navigator, UnitID 104717. REAL. The IPEDS funnel is first-time degree-seekers only; it's used here for the inquiry→start ratio shape, not as a volume claim.
Per-channel CPI and start-rates — ILLUSTRATIVE, calibrated to the structure of a 2012 GCU paid-media pilot (Blue Camel / Sparkroom): TV buys the cheapest inquiries and converts to Starts at the worst rate — that's where the CPI-vs-CPS trap springs. GCE does not disclose cost-per-start, so no per-channel figure here is a reported GCE number. Tuned to the method's published bands — a 4–10% cost-per-Start reduction and a 5–8% Starts lift at no increase in spend: the default (+10% boast) lands at the conservative edge (≈−4.5% CPS, with the CPI-optimal plan ≈9–10% worse on cost-per-Start, matching the published 2012 finding); widen the boast headroom and both move into the upper bands.
Method — boast-and-bound, published at jeffpinto.com/notes/gcu-media-planning. Solver: greedy fill of the highest-yield channel to its ceiling, then the next — optimal for a single linear objective under one budget bound + box ceilings; agrees to the dollar with the scipy LP in the published repo. What it deliberately can't do: multi-touch attribution (the brand-halo where TV lifts branded paid-search). Don't read a single channel's bar as a real spend recommendation — read the gap between the two objectives.