The one thing your in-house team can't build: a record they didn't grade themselves.
An internal team that built the cost-per-Inquiry buy can't credibly audit whether it's cost-per-Start-pessimal ; that's grading your own homework. And in a shop in active FTC discovery over how it markets, an independent, every-input-cited record of why each dollar moved is an evidentiary asset, not overhead. This is that record: a quarter of spend decisions, each one rebuilt from a citation, scored on whether it would survive an examiner asking "why did you move budget there?"
Dec 2023
FTC sued GCE over deceptive marketing; partial dismissal DENIED Aug 2024 ; discovery ongoing (10-K Item 1A)
$37M
ED fine on GCU, Oct 2023, over doctoral-program cost representations , under appeal (10-K)
$212.4M
FY2024 marketing line whose every move this record reconstructs and cites (LOPE 10-K)
Pick a record-keeping posture : one click rebuilds the ledger
The spend-decision ledger (one quarter)r
Decision
Moved
Stated reason
Provenance
The examiner's walk : the four questions a cited row has to survive
The artifact that leaves the building
Sources & method
FTC suit · Dec 2023, dismissal denied Aug 2024 · $37M ED fine Oct 2023 · ~67.2% 90/10 Title IV (FY2024) : all REAL, from the GCE (LOPE) Form 10-K Item 1A risk factors and GCU's audited financials (SEC EDGAR, LOPE 10-K). Cited here to ground why a defensible record has value to this buyer ; not as commentary on the merits.
$212.4M marketing line ; "Marketing and communication" expense, FY2024, LOPE 10-K consolidated income statement. REAL.
The spend rows, dollar moves, and "stated reasons" : ILLUSTRATIVE. GCE does not disclose per-vendor spend decisions; these are a synthetic quarter built to the structure of an enrollment-marketing buy (a TV move, a third-party-lead expansion, a brand-search defense), never GCE's real decisions. The point is the record-keeping discipline, not the rows.
The score : counts how many rows are reconstructible from a citation an outsider could check (10-K line, IPEDS funnel, a dated export) vs. how many rest on "the team's model" or "we felt." Each posture toggles which rows are which; nothing is hardcoded ; the score recomputes from the rows.
Why independence is the moat : the optimizer in prototype 01 is a `linprog` call an in-house team can write by Friday. What they structurally can't write is an audit of their own buy: the team that chose cost-per-Inquiry can't certify it isn't cost-per-Start-pessimal. Independence isn't a capability gap; it's a conflict-of-interest gap. Method note: the boast-and-bound note.