Phase Space AI

Financial Model Notes

Intuit [INTU]

Intuit [INTU] — Financial Model Notes & Defect Log

As of 2026-07-29. Sources: Alpha Vantage normalized statements (81 quarterly periods, 20 annual), the Q3 FY2026 10-Q (0000896878-26-000025, filed 2026-05-20, period ended 2026-04-30), ten EARNINGS_CALL_TRANSCRIPT pulls, and Alpaca for prices and options.

---

1. Fiscal-period discipline — the rule this file obeys

Intuit's fiscal year ends 31 July. Q2 (ending 31 January) and Q3 (ending 30 April) carry the tax season.

Every comparison in this model is year-over-year on matched fiscal periods. No sequential comparison is made. The TTM window spans exactly one complete seasonal cycle and is compared only to the same four fiscal quarters a year earlier.

This project has already shipped a period-mismatch defect class. Nothing here adds to it.

---

2. The quarterly series actually used

All $m. S&M is derived as operatingExpenses − R&D − G&A, because AV's SG&A field carries G&A only — see defect AV-1.

Operating margin on matched TTM windows: 21.67% → 24.38% → 27.47%. +5.8pp in three years.

Verified against the filing for Q3 FY2026: revenue $8,558m, cost of revenue $1,331m, gross profit $7,227m, R&D $840m, SG&A(as filed by AV mapping) $409m, total operating expenses $3,221m, operating income $4,020m, interest income $42m, interest expense $70m, income before tax $4,047m, tax $983m, net income $3,064m. Segment revenue ties: GBS $3,285m + Consumer $5,273m = $8,558m ✓. Product detail ties: QBO Accounting $1,278m + Online Services $1,219m + Desktop Accounting $507m + Desktop Services $281m = $3,285m ✓; TurboTax $4,364m + Credit Karma $631m + ProTax $278m = $5,273m ✓.

Share count verified against filed EPS, not against a vendor: $3,064m ÷ $11.09 GAAP diluted EPS = 276.3m. Not dual-class, no split in the series — neither the WDAY dei failure mode nor the CRWD split-basis failure mode can apply.

---

3. Defects found — Alpha Vantage

DEFECT AV-1 (HIGH, new) — sellingGeneralAndAdministrative carries G&A only

Reproduced on both names in this pair, with near-identical magnitude. AV's operatingExpenses total is correct on both, so grossProfit − operatingExpenses = operatingIncome reconciles and every internal consistency check passes.

Why this is severe. The mandated terminal-margin bridge is m_EBIT,T = m_gross,T − R&D − S&M − G&A − other. Built from AV fields alone Intuit's bridge reads 81.02% − 15.87% − 7.91% = 57.2% against the true 81.02% − 29.85% − 15.87% − 7.91% = 27.47% — a 29.8pp error on the parameter this project has broken five times, and a terminal margin that would have been physically impossible for a business with a 27.5% actual. This is the brief's defect #2 exactly: a vendor that is wrong consistently defeats every consistency test. Not in known-silent-failures.md. Recommend adding, with the workaround: derive S&M as operatingExpenses − researchAndDevelopment − sellingGeneralAndAdministrative.

DEFECT AV-2 (MEDIUM, new) — ebit is not operating income

AV computes ebit = incomeBeforeTax + interestExpense, which absorbs non-operating items. On Intuit the error is modest in the tax quarter and 22% in the January quarter. On NOW it is 3.11x. The brief flags ebitda; ebit is equally unsafe on both names and is not currently flagged. Recommend adding.

DEFECT AV-3 (LOW) — shortTermDebt

AV reports $833m against the filed balance-sheet figure of $750m — an $83m overstatement that propagates directly into net cash. The filed figure is used. Reported, not silently adopted.

DEFECT AV-4 (confirmed) — ebitda unusable

INTU 2026-04-30: AV ebitda $4,327m = its own (wrong) ebit $4,117m + D&A $210m. Never used; EBITDA is computed from operating income plus |D&A| from the cash-flow statement where needed.

DEFECT AV-5 (HIGH, new) — EARNINGS_ESTIMATES returns an empty array

EARNINGS_ESTIMATES for INTU returns {"symbol":"INTU","estimates":[]}. Identical for NOW. The premium tier's headline feature — built-in 7/30/60/90-day revision history — is unavailable on two names above $90bn of market capitalisation. Consensus Criteria is INDETERMINATE, which blocks nothing. No consensus figure was substituted from elsewhere and none was invented; the NTM base is built from filed actuals plus the company's own guidance instead.

CREDIT to Alpha Vantage

• 81 quarterly periods of normalized statements on INTU, back to roughly 2006. This is the deepest series any name in this project has had and it is what made the three-year and five-year matched-TTM comparisons possible without hand-parsing XBRL.
• Every revenue, gross-profit, R&D, operating-expense, operating-income, tax and net-income figure checked against the Q3 FY2026 10-Q tied exactly. The defects above are all field-mapping problems, not value errors.
• Transcripts with speaker/title/sentiment made the prepared-remarks-versus-Q&A analysis possible for the first time, and it produced two findings that would otherwise have been invisible: the 17% workforce reduction figure appears only in Q&A, and the ~2 million lost DIY units appear only in Q&A.

---

4. Defects found — tooling

DEFECT TOOL-1 (MEDIUM) — av_vs_edgar.py reports a rate limit as missing data

sym      cmp  dis   worst   inv  trust
INTU      --   --      --     --   AV_NO_QUARTERLY: Minute-level rate
NOW       --   --      --     --   AV_NO_QUARTERLY: Minute-level rate
  clean: 0/2   failed audit: 0   no EDGAR overlap: 0

The script hit AV's per-minute burst limiter and reported AV_NO_QUARTERLY — "no quarterly data." AV had 81 quarterly periods for INTU and 64 for NOW. It then reported clean: 0/2 · failed audit: 0, which reads as "nothing wrong found" when nothing was checked.

This is calibration item D1 — "existence is not validity" — occurring inside the tool built to catch D1. The mandated cross-check was performed by hand against the filed 10-Q instead, which is how defects AV-1 through AV-4 were found. Recommend the script distinguish a rate-limit response from an empty response and retry with backoff.

---

5. Screen-input validation

---

6. Terminal-margin provenance (required by valuation.md rule 1)

The comparison with NOW, stated because it is the point. NOW's 23.0% terminal margin needs SBC to fall from 16.4% to below 10% — a 6.9pp swing doing most of the work, backed by a company commitment but still a large single assumption. Intuit's 33.0% needs SBC to move 0.3pp. Intuit's terminal margin carries materially less parameter risk, and its Valuation Criteria consequently PASSes across a far wider region of the sensitivity surface.

---

7. Base-effect trap set for FY2027 — flagged in advance

The $300–340m restructuring charge lands in Q4 FY2026, a quarter that already runs an 8–9% operating margin on ~$4.3bn of revenue. FY2027 will therefore show flattered year-over-year margin expansion off a depressed base.

This is precisely the WDAY defect: a headline +11.5pp of margin expansion that was +4.1pp clean, with two-thirds of it a prior-year restructuring base effect, and the screen labelled WDAY INFLECTION on the unadjusted figure. Any FY2027 Intuit margin comparison must be made ex-restructuring on both sides. Recorded now, before the number exists, so it cannot be adopted later by accident.

---

8. Items not completed inside the time box, stated rather than estimated

• Credit Karma segment operating income — retired by the company effective 2026-08-01 and therefore not obtainable from primary sources. I decline to estimate it. Revenue is disclosed ($1,898m 9M FY26, +21.5%); profitability is not.
• Mailchimp revenue — never separately disclosed. Only quantifiable through management's ex-Mailchimp growth deltas (Online Services +18% reported vs +28% ex-Mailchimp, Q2 FY2026, implying a −10pp drag and an outright revenue decline). Not estimated as a dollar figure.
• Piotroski F-score — not computed. INDETERMINATE, not FAIL.
• External professional price target — none obtained, so the mandated sanity-band comparison is not performed.
• Prior-year TTM S&M/G&A split — not reconstructed, so the four-line bridge is shown for the current TTM only.
• The cause of the +17.1% five-session price move to 2026-07-29 — not established from any source read. Recorded as a fact, not explained.
• Q4 FY2026 revenue — not yet reported. My ~$4,330m is an estimate labelled as such and used only to build the FY2027 NTM base for the 12-month target; it does not enter the reverse DCF, which uses filed TTM actuals.