The quarterly trajectory tells a messier story than the 65% CAGR headline. Revenue went $55M → $95M → $135M → $223M → $167M — that Q1 2026 sequential drop of 25% off a Q4 print is the single most important data point in this file, and nobody upstream flagged it. Net income is even more erratic: -$0.8M, $30M, $90M, $15M, $45M. A 66% net margin in Q3 2025 followed by 6.8% in Q4 is not a business — that's gain-on-sale accounting, mark-to-market on loan portfolios, or one-time blockchain token/equity gains being run through the P&L. Figure originates HELOCs and securitizes them via Provenance; revenue recognition on whole-loan sales and retained interests is inherently lumpy and pro-cyclical to rate expectations. The $62.6M operating cash flow against $134M reported net income confirms it: less than half the earnings are converting to cash. That's the "Poor Cash Flow Quality" signal doing real work, and the synthesis underweights it.

The classification engine calling this "high_growth_profitable" at 0.67 confidence and the pre-flight calling it "pre-profit-platform" are directly contradictory, and the pre-flight is closer to right. A company with one full year of GAAP profitability, where that profitability swings from -1.5% to 66% margin quarter-to-quarter, is not a profitable compounder — it's a mortgage originator with a blockchain wrapper whose earnings quality has not been tested through a full rate cycle. The narrative layer correctly identifies this as "mission-driven-bet" with fragile durability; that's the most honest read in the stack. The synthesis verdict of "High Conviction Required" is a non-verdict — it's punting. At 11x sales, 20x EV/revenue, and 34x TTM P/E for a lender (because that's what this fundamentally is, regardless of the DLT plumbing), the comp set isn't Coinbase or Block — it's Rocket, LendingTree, SoFi, all trading at 2-4x sales. The blockchain premium is doing 60-70% of the valuation work.

The contrarian case the models miss: insiders. "Net Insider Buying" is misleading — the transactions list is dominated by F-InKind (tax withholding on vesting) and M-Exempt (option exercise), with actual S-Sale activity. There are zero open-market purchases here. Post-IPO lockup expiry dynamics are likely driving the print, and the secondary signal mislabels mechanical vesting as conviction. Second, the Q1 2026 revenue decline coincides with a rate environment where HELOC origination should be a tailwind (homeowners tapping equity rather than refinancing) — if Figure can't grow sequentially in that setup, the TAM thesis is in trouble. Third, the missing balance sheet data (no total debt, no total equity disclosed) on a financial-services company is a significant gap; warehouse lines, securitization residuals, and tangible book value are the actual valuation anchors here, and we're flying blind on them.

I dissent from the synthesis "High Conviction Required" framing — that's a hedge, not a call. My read: Figure is overvalued at $27.93 / $5.1B market cap by 35-50%. Strip out the blockchain narrative premium and value this as a high-growth specialty lender at 4-5x sales on $500M run-rate revenue → $2.5-3.0B fair value, or $14-17 per share. The bull case requires sustained 40%+ growth AND margin stabilization in the 20-25% range AND Provenance becoming a real third-party platform (not just Figure's captive rails) — three conditional bets, not one. The market cap fell from $78 to $28 because the first round of post-IPO holders figured this out; the next leg requires either a clean Q2 2026 print proving Q1 was a blip, or a multiple compression to $18-20 where the lending business alone supports the valuation. The narrative is fragile, the cash flow is poor, the margins are unstable, and the insider "buying" is an accounting artifact. Pre-flight got it right calling this pre-profit-platform; the synthesis should have followed through.