BILL Holdings [BILL] — Research

Tier-2 memo · as of 2026-07-29 · framework v1.7.0 · no position verdict Entity: BILL Holdings, Inc. · CIK 0001786352 · NYSE: BILL · FYE 30 June · SIC 7372

Latest primary filing used: 10-Q for the quarter ended 2026-03-31, filed 2026-05-08 (accession 0001628280-26-032387). Q4/FY2026 results are not yet filed; the fiscal year ended 2026-06-30 and BILL has filed its 10-K in the last week of August in each of the last three years. Filing age at the as-of date: 82 days. Nothing in this memo uses a period later than 2026-03-31 except live prices and the options chain.

1. The finding, stated first

The screen credits BILL with a 31.6% demonstrated revenue CAGR against a required 15.3%, a +16.3pp margin. Both numbers are wrong, and they are wrong in opposite directions:

The screen was right that BILL clears the implied-path test. It was right for two compensating wrong reasons, and the true margin is one-sixth the size it reported.

2. The three-way revenue split — the thing this memo was commissioned to do

BILL's income statement carries two lines. The press release carries three. The three are economically unrelated to each other.

Trailing twelve months to 2026-03-31

Reconciliation: 286.2 + 1,164.0 = $1,450.2m, which ties to the 10-Q-derived subscription-and-transaction-fees TTM of $1,450,214k to the last thousand. Float $150,109k. Total $1,600,323k versus the screen's $1,600,324k — a $1k rounding difference, so the screen's TTM revenue is confirmed correct.

Only 17.9% of BILL's revenue is software. A screen that reads SIC 7372 ("prepackaged software") and treats 3.4x sales as a software multiple is mispricing a business that is 82.1% payments-and-float by revenue.

The full quarterly series (all from BILL's own 8-K earnings releases, reconciled to the 10-Qs)

¹ Q3/Q4 FY24 subscription derived by residual from the FY2024 total of $257.1m and the stated YoY growth rates; marked approximate. Every other cell is a stated figure. Each quarter's subscription + transaction sums to the 10-Q's stated subscription-and-transaction-fees line to within $0.1m.

Annual series (10-K, audited)

Subscription vs transaction is disclosed only from FY2024 onward in the MD&A revenue table; FY2021–23 give the combined line only.

3. Growth restated ex-float

The screen's 31.6% is the FY2022 → FY2025 total-revenue CAGR ($641,959k → $1,462,570k; (1,462,570/641,959)^(1/3) − 1 = 31.58%, which reproduces the scanner's 0.3158401 exactly).

Float supplied $153.2m of the $820.6m of reported revenue growth over that window — 18.7% of all growth, from a line that is a function of the federal funds rate and nothing the company did. Restating removes 4.5pp from the headline CAGR.

But the ex-float restatement is the smaller of the two corrections. The larger one is that the window itself is stale. The 31.6% ends at FY2025, thirteen months before the as-of date, and BILL has decelerated every single year inside and after it:

The number that should have entered the reverse DCF as "demonstrated" is 12.5%, and the company itself guides to 12–13%. The 31.6% is a fossil.

Did the acquisitions inflate it? No — and this is worth stating plainly

Per the FY2022 10-K: Divvy closed 1 June 2021 and Invoice2go closed 1 September 2021. FY2022 — the base year of the 31.6% window — therefore already contains Divvy for a full twelve months and Invoice2go for ten. The acquisition step-up lives in FY2021 → FY2022 (+169%), outside the measurement window. Goodwill has been unchanged at $2,396,509k at 30 Jun 2024, 30 Jun 2025 and 31 Mar 2026 — no acquisitions since.

So the answer to the brief's question is: the 31.6% is not acquisition-inflated. It is float-inflated (4.5pp) and, far more importantly, stale (19pp). The M&A hypothesis is the one that does not survive contact with the filings; report it as tested and rejected rather than assumed.

4. Accounting quality — is the growth real?

Six things were tested. Two are clean, four are not.

4.1 Net dollar-based retention is BELOW 100%, and has been for two years

From the FY2025 10-K MD&A, verbatim: "Our net dollar-based retention rate was 94%, 92%, and 111% during fiscal 2025, 2024, and 2023, respectively."

BILL's own definition excludes interest on funds held for customers and is measured on BILL AP/AR customers. An existing cohort of BILL AP/AR customers generates less revenue this year than last year. Every dollar of AP/AR growth is coming from new logos, and logo retention is "approximately 86% of BILL AP/AR customers as of June 30, 2024 were still customers as of June 30, 2025."

This is the single most important quality fact in the name and it is not in the screen. A software business with 94% NDR is not compounding; it is replacing. The industry-wide SMB-fintech NDR deceleration the brief asked about is present here in an acute form: 111% → 92% → 94%, a 17pp collapse in one year (FY23→FY24) that has only partially stabilised.

The word "retention" appears zero times in each of BILL's last six quarterly earnings releases (it appeared once per release through Nov-2024). The metric is still disclosed annually in the 10-K, so this is de-emphasis, not withdrawal — but it is de-emphasis of the worst number in the file.

4.2 Customer count turned negative in the March-2026 quarter

"Businesses using our solutions," from the quarterly releases:

498,500 → 493,800 is a sequential loss of 4,700 businesses (−0.9%), the first sequential decline since the March-2024 quarter. Year-over-year growth is +1.1% (488,600 → 493,800), down from +4% in FY2025 and +3% in FY2024. BILL's customer base has stopped growing.

4.3 The disaggregation that was retired

The FY2025 10-K revenue note discloses revenue by solution (BILL AP/AR $667.8m, BILL Spend and Expense $555.0m, Embedded and Other $78.0m) — not by subscription versus transaction. The subscription/transaction split survives only in the MD&A table and the press release, and only from FY2024. Prior years give the combined line. This is not a Twist-style withdrawal, but it means the composition of the 82% of revenue that is not software must be assembled from three different documents.

4.4 What is clean

• Accruals: strongly negative, i.e. high quality. 9M FY2026 net income $7.2m against operating cash flow of $304.8m. Cash conversion is not the problem here.
• Revenue is not concentrated. ~493,800 businesses; no customer, distributor or partner approaches materiality. There is no AAOI-style hidden channel.
• No non-recurring items propping the top line. No settlements, milestones or one-off licences in the revenue lines. The FY2025 net income of $23.8m is, however, entirely a function of $111.0m of "other income, net" — corporate interest income and a $37.9m gain on debt extinguishment — against an operating loss of $80.6m. BILL's GAAP profitability is an interest-rate artifact, twice over.

4.5 Non-GAAP distance

FY2026 guidance: non-GAAP operating income $303.6–308.6m on total revenue of $1,642–1,652m, an 18.6% non-GAAP operating margin. TTM GAAP operating margin is −3.8%. The bridge is stock-based compensation of $177.0m in nine months (~$236m annualised, 14.7% of revenue) plus D&A. This is a 22-point gap between the margin management presents and the margin that exists. The terminal-margin assumption in the valuation is exactly the question of how much of that gap closes.

5. The mechanism — named, evidenced, and singular

BILL's growth mechanism is take-rate expansion on a flat customer base. Not customer acquisition, not seat expansion, not price. Monetisation per dollar of payment volume.

Transaction fees ÷ total payment volume (TPV is reported rounded to the whole $bn, so each figure carries roughly ±0.4bp of precision):

Annual cross-check from the 10-K: FY2025 $1,028.7m on $329.8bn = 31.2bp; FY2024 $865.6m on $292.4bn = 29.6bp. +5.4% take-rate expansion in FY2025.

Take-rate direction is UP — the brief asked, and the answer contradicts the usual SMB fintech bear point. The arithmetic of the latest quarter: transaction fees +18% = TPV +12% × take rate +5%.

What drives it, from the 10-K's own description of transaction fees: "checks, ACH, card payments, real-time payments, pay by card, invoice financing, cross-border payments… Transaction fees also include interchange fees paid by suppliers accepting card payments." The mix is shifting out of near-zero-monetisation ACH and cheques into card/interchange, invoice financing and cross-border — all ad-valorem. Divvy (BILL Spend and Expense) is pure interchange: $555.0m of FY2025 revenue, growing +21% in FY2025 and +21% in 9M FY2026 (404.4 → 491.0), versus BILL AP/AR's +11% (492.9 → 547.7).

The mechanism's ceiling is the honest risk. Take-rate expansion is bounded by supplier card acceptance and by what an SMB will pay per payment. It has run 28.2 → 33.3bp in eight quarters. It cannot run to 60bp. When it stops, transaction-fee growth converges to TPV growth, and TPV growth converges to customer growth, which is currently +1.1%.

Product-cycle intelligence

Named products with evidence of adoption: - BILL Spend and Expense (Divvy) — $555.0m FY2025, +21%; $491.0m in 9M FY2026, +21%. ~41,100 spending businesses at 30 Jun 2025. The fastest-growing solution and the most competitively exposed (Ramp, Brex, Navan are private and better capitalised in this category; none of their numbers are in a filing, so no comparison is made here). - Embedded Solutions and Other — $78.0m FY2025, +11%; $65.6m 9M FY2026, +13.9%. Small, and the slowest-growing line despite being the strategic story. - AI agents — first appears in the Aug-2025 release ("AI agent" mention count 0,0,0,0,0,0,1,3,1,1 across the last ten releases). The word "AI" itself runs 45–55 mentions per release with no trend over two and a half years. There is no revenue attribution to AI in any filing. Treat as absent from the model. - Invoice financing, procurement, supplier payments plus — named in the 10-K and in 1 mention per recent release. No disclosed revenue.

6. Transcript / disclosure mention-frequency over time

Built from BILL's own quarterly earnings press releases (EX-99.1 to each results 8-K) — primary filings, so the series is reproducible. Counts are per release, oldest to newest.

What it says. Three step-changes, all in the last three releases. (1) "retention" goes to zero from Feb-2025 and stays there — exactly when NDR was printing 92–94%. (2) "activism" appears seven times in each of the last three releases and zero times before Nov-2025, dating the campaign's entry into company disclosure. (3) The May-2026 release adds nine mentions each of "Spend and Expense" and "embedded" — the solution-level disaggregation table arriving in the release — and is the longest release in the series (5,825 words), with the "repurchase" count spiking to 16 on the $1.0bn authorisation.

Caveat, stated rather than hidden: this is a press-release frequency series, not a call-transcript series. Call transcripts are not SEC filings and no verified transcript source was available inside the time box. The series is therefore a measure of what management chooses to put in writing under Reg FD, which is a narrower and more conservative instrument than the Q&A. It should not be read as equivalent to a transcript count.

7. Governance, management and the activist

Verified from filings:

Four senior departures — the accounting officer, a director, the CTO and the GM of Payments — inside six months, alongside a proxy contest. The Payments GM departure is the notable one: payments is the business, at 72.7% of revenue.

Explicitly unverified. Press reports describe an 8.5% Starboard Value stake, board seats including Peter Feld, a strategic-alternatives review, private-equity interest and a 30% workforce reduction. No SC 13D appears in BILL's recent EDGAR filing index (the 13G/A filings there are index-fund filings through 2024), and I did not locate filings substantiating the stake size, the board composition or the strategic review inside the time box. None of it is used in the valuation. A DFAN14A exists and "activism" is in the company's own releases, so a campaign is real; its terms are not established here.

Also press-reported and unverified: a new Chief Revenue Officer effective July 2026, and a TD Cowen initiation. Not used.

8. Criteria scoring

The memo scores. It issues no position verdict.

9. Verification of the screen's inputs

Per the brief, every scanner input was checked against primary filings.

The net-cash error, exactly

The scanner computed 994,672 − 1,504,854 − 330,000 = −840,182 — cash and cash equivalents, less convertible notes, less credit-facility borrowings. It omitted the $1,178,673k "Short-term investments" line that sits directly beneath cash on the same balance sheet. Corporate short-term investments are separate from the $3,999,136k of "Funds held for customers," which is exactly offset by $3,999,136k of "Customer fund deposits" and correctly excluded by both the scanner and this memo.

Correct: 994,672 + 1,178,673 − 1,504,854 − 330,000 = **+338,491**

Three alternative treatments, all disclosed rather than chosen silently:

The verified row is used throughout. It is the most conservative of the three defensible treatments that include short-term investments.

This is a D1-class defect of the opposite sign to the usual one: the scanner did not fail on a missing tag, it silently used an incomplete one, and the resulting confident wrong number made BILL look 28% more expensive than it is.

10. Broken growth story, or overcorrection? The honest answer

The stock is 86.1% below its 2021-11-09 closing high of $342.26. That invites the overcorrection framing. The evidence supports a narrower and less exciting conclusion.

It is not a broken growth story. Revenue is growing 12.5%, core 15.4%, take rate is rising, gross margin is 81%, operating margin has expanded 8pp in a year, operating cash flow is $305m in nine months, net cash is positive, and the company is buying back up to $1.0bn of a $4.6bn market cap. Nothing here is breaking.

Nor is it a clean overcorrection. At 2.67x EV/Sales the market is not pricing distress — it is pricing roughly what the business is doing. The reverse DCF requires a 9.8% CAGR against a demonstrated 12.5%: a +2.7pp margin, not a +16.3pp one. And the exit multiple that produces that answer implies an exit EV/Sales of 24.1 × 0.112 = 2.70x, versus today's 2.67x — meaning the price embeds essentially zero multiple expansion and essentially zero multiple compression. The market is asking BILL to grow low-double-digits and eventually earn an 11% GAAP EBIT margin. That is not a broken story and it is not a mispricing; it is a reasonable price for a decelerating payments business.

Where a real opinion lives: BILL trades at the 30th percentile of its own EV/Sales range since January 2025 (median 3.15x). If the multiple simply returns to its own recent median, that is worth roughly +30%, and the $1.0bn buyback adds to it. That is a 12-month multiple trade, not a five-year valuation argument — and it is exactly what the 12-month target in BILL_Valuation.md is built on, kept separate from the implied-path test on purpose.

11. What is unsupported

Stated so nothing here is mistaken for evidence:

1. The exit multiple of 24.1x EBIT. Taken from the screen (GROWTH_MATCHED, peer n=254). I did not rebuild the comparator set. BILL is classified SIC 7372 (prepackaged software) and the growth-matched set is therefore drawn from a software universe, while 82.1% of BILL's revenue is payments and float. The whole Valuation Criteria result turns on this number: the break-even is 21.3x, and named payments comparables plausibly trade below it. Flagged as the dominant unresolved input, not resolved.
2. Peer Spread Criteria — INDETERMINATE. No peer multiples pulled.
3. Consensus — INDETERMINATE. Alpha Vantage quota exhausted.
4. Street price target ~$53.91. Press-sourced, not verified against any broker document. Used only as a sanity band, never as a calibration input.
5. Starboard's 8.5% stake, board seats, strategic-alternatives review, PE interest, 30% workforce reduction, new CRO, TD Cowen initiation. All press-reported; none substantiated by a filing I located. None used in the valuation.
6. Q3/Q4 FY2024 subscription fees ($65.6m, $65.8m) are derived by residual, marked approximate.
7. Take rates are computed against TPV rounded to the whole $bn: ±0.4bp precision. The direction is robust; individual quarters are not.
8. Q4 FY2026 earnings date. Not announced. The catalogue gives the prior three years' actual filing dates and nothing more.
9. Ramp / Brex / Navan competitive position. Named as the competitive set for BILL Spend and Expense on general knowledge. All private; no filings; no numbers asserted.
10. The call-transcript mention series does not exist here. §6 is a press-release series and is labelled as such.