An Up-C structure corrupts six separate vendor fields at once — and two that look broken are correct

The structure that causes it

An Up-C puts a public C-corp (Class A, listed) over an operating LLC. Pre-IPO owners hold membership units in the LLC plus a Class B common share that carries votes but no economic interest. The unit and the stub travel together and exchange into one Class A share.

Almost every listed alternative asset manager is an Up-C. So is a large share of the 2015–2021 IPO cohort outside financials.

The six failures, from HLNE on 2026-08-12

Field	Vendor	Truth	Cause
impliedSharesOutstanding	67.0M	55.63M	Double-counts the exchange. 55.6M already contains the units; adding the 11.4M Class B common stub counts the pair twice
marketCap	$6.92B	$5.75B	Cascades. nonDilutedMarketCap ($5.78B) was the correct field
enterpriseValue	$5.27B	$5.68B	Reproducible from no combination of Yahoo's own fields. Discard
"Acquisitions"	−$96.7M	$0 M&A	Maps line-for-line to "Contributions to Funds" — GP commitments into the manager's own funds. PaymentsToAcquireBusinesses was $0 in FY24, FY25 and FY26
"Dividends"	−$141.8M	$87.7M	Merges Class A dividends ($87.7M) with Class B/C member distributions ($54.1M)
totalDebt	$364.8M	$274.2M	Corporate debt only; the consolidated funds carried none

⚠️ Two fields that look broken are correct — check before discarding

trailingEps 6.57 and priceToBook 4.93 both verified. P/B survived by offsetting errors (parent-only equity ÷ Class-A-only count ≈ economic equity ÷ economic base, within 3 cents). The lesson is not "the feed is garbage" but "audit it field by field" — a wholesale discard would have thrown away the two numbers that were usable.

🔴 The diluted share count is NOT a vendor bug — and this is the trap inside the trap

HLNE's Shares(dil) reads 53.90M (FY24) → 40.31M (FY25) → 54.47M (FY26). That looks impossible and invites a data-error conclusion. It is exactly what the company filed. The FY25 10-K excluded 14,016,324 Class B/C units as antidilutive under the if-converted test, dropping them from numerator and denominator. The units were excluded in FY18–21 and FY25, included in FY22–24 and FY26.

Never CAGR an Up-C diluted share count. It is a test outcome, not a trajectory.

This is the mirror image of pitfall-yahoo-share-count-dual-class-fpi: there the vendor understated shares by omitting a class; here the vendor faithfully reports a figure that legitimately oscillates. Both produce a nonsense series; only one is the vendor's fault.

The share-issuance line is also misread by default

An Up-C's financing section frequently shows a large equity issuance that is not a capital raise. HLNE FY26:

Proceeds from offering 55,484 Purchase of membership interests (55,484)

100% of the proceeds bought LLC units from insiders. Class A count rises; the economic base does not; the Class A holder is not diluted. Read as a raise, it looks like serial dilution — HLNE's Class A grew from 19.0M to 41.6M over eight years while the economic base went 49.6M → 54.5M, i.e. +1.2%/yr, not +119%.

Detection — four checks, all cheap

1. marketCap ÷ price vs sharesOutstanding. A gap means one field is on a different base. Then determine which is wrong — do not assume it is the share count.
2. Reconstruct the economic base from the 10-Q cover page: Class A outstanding + LLC units from the EPS note. Do not add Class B common — it is a voting stub with no economics.
3. Read the income statement's three-way split: consolidated net income → NCI (funds) → NCI (the LLC) → attributable to the parent. The LLC NCI belongs in the economic base; the fund NCI does not.
4. Check for a TRA (tax receivable agreement). Up-Cs typically owe ~85% of the exchange's tax benefit back to the selling insiders. It appears as a deferred tax asset and a TRA liability — two halves of one transaction. Any book-value model (Graham especially) credits the gross asset to Class A holders who receive ~15% of it.

Rule

On any Up-C, rebuild market cap, EV and every per-share figure from the 10-Q before using a single vendor ratio — then re-test each vendor field individually rather than discarding the feed. State which share base every figure uses.

What would falsify this

Yahoo populating impliedSharesOutstanding from the LLC unit count rather than summing share classes. Re-test the reconciliation on any Up-C before trusting the field again.

Related

[[pitfall-yahoo-share-count-dual-class-fpi]] — the sibling failure, opposite direction · [[pitfall-vendor-ev-inverts-net-cash]] · [[principle-primary-source-beats-vendor]] · [[pattern-non-gaap-definition-change-leaves-a-restatement-fingerprint]] — the other half of the HLNE analysis.

History

• 2026-08-12 — found during /analyze HLNE. The manager's opening hypothesis was that sharesOutstanding (43.35M) was wrong per the existing dual-class note. It was correct; impliedSharesOutstanding was the broken field. Filed slow rather than static because it is vendor behaviour, though the underlying structure is permanent.