Vendor feeds zero out deployment and inflate debt for sales-type-lease REITs

Claim

Two independent vendor errors on the same issuer class, both material, both pointing at the wrong conclusion.

Error 1 — the acquisition line reads zero while billions are deployed

VICI, per fin.py (Yahoo-sourced) and roic.ai:

Acquisitions	FY22	FY23	FY24	FY25
Vendor "Purchase of Business"	$4,574.5M	$1,266.9M	$0	$0
Actual deployment	$9,304.0M	$2,899.1M	$922.8M	$904.8M

The zero is not a fact about the business. VICI books virtually every property acquisition as an "Investment in leases — sales-type" or "— financing receivables", not as PP&E and not as a business combination. Those flow through "Net Other Investing Changes" and "Purchase of Investment", leaving the mapped "Acquisitions" line empty.

What it cost: two separate desks opened their VICI work on the premise that "the acquisition engine has stalled since FY24" and built a thesis on it — a cost-of-capital paralysis story. 1H 2026 deployment was ~$2.1B gross, more than FY24 and FY25 combined. The engine was running the whole time. The correct criticism was never "they cannot do deals"; it was "they are doing deals at a ~zero spread" — a different diagnosis with a different verdict.

Where to read it instead: total investing outflow, less the loan-book line, from the company's own cash flow statement. Or the supplemental's transaction page, which lists each deal with its cap rate.

Error 2 — total debt overstated by lease liabilities

FY25 total debt	Source
$17,689.7M	Yahoo, `fin.py`, roic.ai
$16,773.2M	VICI balance sheet
$916.5M	the gap — operating/finance lease liabilities, carried inside "Other liabilities"

Consequences if taken at face value: Debt/Assets 37.86% instead of the true 35.7%, and an inflated invested-capital denominator that depresses computed ROIC (worth ~16bp on VICI — small, but it compounds with the WACC error in [[principle-reit-wacc-has-no-tax-shield]]).

A related instance of the same family: roic.ai reported net debt/EBITDA of 4.44x against the company's 4.9x, because it divides by GAAP EBITDA rather than the LQA Adjusted EBITDA the covenant and the company use. The company's more conservative figure was the correct one — the unusual case where the vendor flatters rather than penalises.

Error 3 — the GAAP payout ratio reaches a decision artifact and survives there

Second instance, different ticker, and the most expensive of the three — because this one was already in the guard table below and got past us anyway.

NLCP, a cannabis net-lease REIT held in both accounts:

	Value	Source
Yahoo `payoutRatio`	1.38 (138%)	dividends ÷ GAAP EPS
What `Watchlist.md` carried	"payout 134%. 🔒 Hold only — do not add until payout normalizes."	copied forward, unverified, for months
True FY25 AFFO payout	82%	company-reported AFFO
True Q1'26 AFFO payout	90%	company-reported AFFO

The error is 44 points wide, and analysis_notes.md §4 already says outright that "a GAAP-EPS payout ratio on a REIT or BDC is meaningless — say so rather than reporting it."

What makes this instance different from the VICI one is what it teaches, and it is not "check the payout ratio."

The wrong number produced the right conclusion. "Hold, do not add" was correct at $15.75 — the scenario-weighted fair value came in at $15.80. A wrong premise pointing the right way is not a saved call; it is an unmonitored one, and it is harder to catch than a wrong premise pointing the wrong way, because nothing about the recommendation ever looks off.
The gate it set was unfalsifiable. "Do not add until payout normalizes" named a metric that was never going to move, because it was never measuring anything. The position could not be re-rated in either direction by any real event. A gate written on a vendor field is a gate that never opens.
Vendor errors do not stay in the analysis layer. The VICI instance was caught inside a Phase 1 brief. This one had already been promoted into Watchlist.md — the file that drives what gets bought — and was propagating into the bilingual site's Chinese sidecar as well.

Companion finding at NLCP: four independent screens all read cheap, and all four were artifacts — P/B 0.84 (marks to ~0.95x once three unimpaired vacant buildings are written down), Graham IV $22.94 (overstates ~48%), DYT on a 2021-regime band ($18.20, disqualified), and a Yahoo mean target of $17.50 built from 2 opinions including an unattributable $18. When every screen on a REIT agrees it is cheap, that convergence is evidence about the screens, not about the price.

Guard

For any net-lease REIT, and especially any REIT whose leases are sales-type or financing receivables (gaming and other single-tenant trophy-asset REITs are the common case):

Metric	Do not trust	Read instead
Acquisitions / capital deployed	Vendor "Purchase of Business"	Total investing outflow from the cash flow statement, or the supplemental's transaction page
Total debt	Vendor "Total Debt"	The balance-sheet debt schedule — check whether lease liabilities were folded in
Payout ratio	Vendor `payoutRatio` (GAAP EPS-based)	AFFO payout
Free cash flow	Vendor FCF	AFFO — and see [[pitfall-roicai-fcf-field-returns-ocf]]
Net debt / EBITDA	Vendor ratio	The company's covenant definition (LQA Adjusted EBITDA)

Tell for error 1: an acquisition line of exactly zero for consecutive years at a REIT whose asset base is visibly growing. Growth without a purchase is an accounting-mapping artifact, not a discovery.

Tell for error 2: vendor debt exceeding the balance-sheet debt schedule by roughly the size of "Other liabilities."

Why this family keeps recurring

Vendor feeds normalise everything into an industrial-company template. Every pitfall in this Playbook involving Yahoo or roic.ai is the same failure: a chart of accounts built for manufacturers applied to an issuer that does not use it. For REITs the mismatch is structural rather than occasional, because the entire earnings model runs through lease accounting that the template has no slot for.

What would falsify this

Nothing about the accounting — it follows from ASC 842 classification. A vendor could fix the mapping, which would be visible as the acquisition line becoming non-zero; re-check before relying on the guard rather than assuming it is permanent.

[[principle-primary-source-beats-vendor]] — the general case; this is its most expensive REIT-specific instance. [[pitfall-roicai-fcf-field-returns-ocf]] · [[pitfall-entg-margins-corrupted-in-vendor-feeds]] · [[pitfall-adjusted-close-breaks-multiple-bands]] — the fourth vendor trap found in the same session.

History

2026-08-03 — both errors found during the VICI /analyze. The zeroed acquisition line had already propagated into the Phase 1 briefs of two desks before Fundamentals caught it against the primary filing.
2026-08-04 — Error 3 added from the NLCP /analyze. The GAAP-payout trap was already in the guard table when it reached Watchlist.md and stood there for months as "payout 134%," gating a held position in both accounts behind a metric that could never move. Upgraded from a data-quality note to a process note: the guard has to run at the point a vendor field is written into a decision artifact, not only when a report is drafted. Also confirmed the mirror case — NLCP's acquisition stall is real, not an Error-1 artifact ($1.49M FY25, $0 Q1'26, verified against the primary cash flow statement), so the guard is a prompt to check, never a conclusion.