Financial Model Notes

Xylem [XYL]

Xylem [XYL] — Financial Model Notes & Defect Register

Screen record validation

Screen: reports/scan_universe/XYL_analysis.json, as-of 2026-07-30, Quality PASS, Valuation PASS at +8.2pp.

screen field	screen value	verified value	source	result
`shares`	233,486,812	233,486,812	10-Q cover page 2026-07-24	exact
`revenue_ttm`	8,812,000,000	9,126,000,000	four filed quarters: 2,268 + 2,397 + 2,125 + 2,336, each EDGAR-verified	WRONG by −$314m, −3.44%
`net_cash`	−1,671,000,000	−1,650,000,000	cash $1,276m − total debt $2,926m (10-Q debt note)	$21m / 1.3% off — the screen took `LongTermDebt` = $2,947m from XBRL; the 10-Q's own debt table totals $2,926m
`ev`	28,902,566,884	28,881,600,000		0.07% off
`ev_sales`	3.2799	3.164		3.7% overstated by the revenue error
`op_margin_pct`	14.0 (FY2025)	13.54 (FY2025 filed) / 14.48 (TTM)	10-K	close; TTM is the better base
`gross_margin_pct`	39.9	38.46 (FY2025 filed)	10-K	1.4pp high
`revenue_cagr_demonstrated`	18.1	1.3 organic	10-Q organic decomposition	invalid — Evoqua acquisition
`exit_multiple`	23.5, `GROWTH_MATCHED`, `peer_n: 286`	9.59x	ROIC identity	invalid — 286-name "peer" set
`required_cagr_pct`	9.9	19.3	reverse_dcf	wrong by 9.4pp
`valuation_margin_pp`	+8.2	−18.0		sign flip, 26.2pp swing

The revenue defect, isolated

$9,126m is the sum of four consecutive, individually EDGAR-verified quarters:

quarter	revenue $m	EDGAR source
2025-07-01 → 2025-09-30	2,268	10-Q filed 2025-10-28
Q4 2025 (FY 9,035 − 9M 6,638)	2,397	10-K filed 2026-02-25
2026-01-01 → 2026-03-31	2,125	10-Q filed 2026-04-28
2026-04-01 → 2026-06-30	2,336	10-Q filed 2026-07-28
TTM	9,126

The screen's $8,812m is not any four-consecutive-quarter sum, and it is not FY2025 ($9,035m) either. The direction is conservative — understating revenue overstates EV/Sales — but it is wrong, and on a name where the memo's conclusion is a re-rating from the 6th percentile, a 3.4% revenue error moves the percentile.

Note the shape: the screen got the hardest input right (a cover-page share count updated six days before the run) and the easiest input wrong (a sum of four filed quarters). Difficulty is not a predictor of which fields are wrong.

Defects found — Alpha Vantage

Fields checked and found CLEAN on XYL — reported per instruction

The DATA_DEFECTS.md discipline is explicit that an untested defect list becomes a prior and a prior causes an agent to reject correct numbers. Every listed field was checked against the filing:

field	AV value	filed value	result
`totalRevenue`, all quarters	2,336 / 2,125 / 2,397 / 2,268	identical	CLEAN, byte-identical
`grossProfit` FY2025	38.5% of revenue	3,475 / 9,035 = 38.46%	CLEAN
`operatingIncome` FY2025	13.5% of revenue	1,223 / 9,035 = 13.54%	CLEAN — contrast KGS, where the same field is +21.9%
`operatingIncome`, quarterly	390 / 244 / 353 / 334	ties to segment tables	CLEAN
`SPLITS`	`{"symbol":"XYL","data":[]}`	—	nil, queried explicitly and stated
`cashAndShortTermInvestments`	not used	`shortTermInvestments` and `longTermInvestments` are both absent from the balance sheet	not applicable — nothing for the composite to exclude. The INSM defect does not fire
`EARNINGS_ESTIMATES`	41 records, full 7/30/60/90-day history	—	CLEAN — not the empty-array-at-HTTP-200 failure
`EARNINGS_CALL_TRANSCRIPT`	10 of 12 quarters, `speaker`/`title`/`sentiment` populated	—	partial coverage, stated
`av_vs_edgar.py`	5 periods compared, 0 disagreements	—	VERIFIED

AV's normalized statements are clean on XYL across every field this project has recorded a defect in. That is the finding, and it is reported because a defect list without negative results is a prior, not evidence.

Fields deliberately not used regardless: ebit, ebitda, sellingGeneralAndAdministrative, OVERVIEW.OperatingMarginTTM, OVERVIEW.MarketCapitalization. All opex lines came from the FY2025 10-K income statement.

Our own tooling

tool	result
`normalized_fcf.py --ticker XYL`	INDETERMINATE — "no growth capex identified above the maintenance rate." This is the correct output, and it was checked rather than worked around: Xylem's capex is 3.7% of revenue and it is genuinely not building. Trailing FCF (10.5%) used directly, which is the tool's own stated condition
`steady_state_check.py`	two flags, both accepted: `TERMINAL_ABOVE_EVERY_OBSERVED_YEAR` (16.0% vs a 13.5% annual peak — accepted with a named mechanism: Q2'26 printed 16.7%, restructuring ends 2027 by disclosure, 350bp of quantified productivity) and `EXIT_MULTIPLE_UNWARRANTED_AT_STEADY_STATE` (9.59x vs 6.1x warranted at its 6.5% ROIC — not overridden; the identity multiple is used, not the market's)
`reverse_dcf.py`	ran with `--fcf-margin 0.105`. Confirmed to be a DCF, not the undiscounted required-revenue test that CORZ and IREN substituted
`av_vs_edgar.py`	VERIFIED, 5 periods. Note the KGS finding: this tool compares revenue, not operating income

Model construction notes

Use the company's own organic decomposition. Xylem publishes organic / acquisitions / divestitures / FX by segment every quarter. Do not compute growth from consolidated revenue.
Do not use any period before Q3 2023 in a growth series. Evoqua consolidates from 2023-05-24 and stepped quarterly revenue $1,722m → $2,076m (+20.6%).
FX is material and must be separated. +2.2pp of the six-month +2.1% revenue change was currency translation. On an FX-neutral basis six-month revenue growth is +0.5%.
Model four segments, not one. Q2'26 operating margins: Water Infrastructure 23.0%, Applied Water 20.2%, Water Solutions and Services 14.1%, Measurement & Control 12.4%. The spread is 10.6pp and the mix is moving (WI +720bp YoY, M&C −20bp).
Segment operating income sums to $412m against a consolidated $390m; the −$22m is corporate.
Tax rate: FY2025 effective 231/1,181 = 19.56%; Q2 2026 was 28.3% against 25.0% a year earlier. Used 19.56% in the ROIC and exit-multiple identity, which is the conservative choice for ROIC.
ROIC has two defensible answers and both are reported: 8.80% on reported invested capital ($12,071m, including $8,256m of goodwill and $2,150m of intangibles) and 63.8% ex-goodwill ($1,665m). The verdict is a FAIL on either.
Backlog is annual-only. $4,615m (FY2025), $5,070m (FY2024). Do not attempt a quarterly backlog series; it does not exist.
Supply-chain financing: $269m at 2026-06-30, $244m at 2025-12-31, inside accounts payable. Add back neither — the balance is disclosed, small (2.9% of revenue) and DPO shortened.

Generated from each name's analysis.json. Figures are research output, not investment advice. Gaps are shown as “—” rather than filled: a missing input is recorded as indeterminate, never estimated.