Phase Space AI

Financial Model Notes

Federal Signal [FSS]

Federal Signal [FSS] — Financial Model Notes & Defect Register

Screen record validation

Screen: reports/scan_universe/FSS_analysis.json, as-of 2026-07-30, Quality PASS, Valuation PASS at +6.8pp.

The net-debt defect, isolated — a new variant

The screen's own net_cash_components field records exactly what it did:

"assets": ["CashAndCashEquivalentsAtCarryingValue@2026-03-31"],
"debt":   ["LongTermDebt@2025-09-30"],
"net_cash_complete": true

It paired a 2026-03-31 cash figure with a 2025-09-30 debt figure — two quarters stale — and set net_cash_complete: true.

The cause is tag choice. Federal Signal stopped populating us-gaap:LongTermDebt after 2025-09-30 and reports instead through us-gaap:LongTermDebtAndCapitalLeaseObligationsIncludingCurrentMaturities, which is populated at 2026-03-31 at $553.0m.

The staleness is not incidental. Between 2025-09-30 and 2026-03-31 Federal Signal closed New Way ($424.2m, 2025-11-25), Kinloch ($14.9m, 2025-12-04) and Mega ($45.0m, 2026-01-16), and issued $400.0m of long-term debt (FY2025 cash-flow statement, ProceedsFromIssuanceOfLongTermDebt). The excluded debt is precisely the debt raised to buy the revenue the screen was scoring.

Classification. DATA_DEFECTS.md records net_cash wrong on ~20 of 22 hand-verified names, and records the "longTermDebt = 0 in the newest quarter only" pattern where 0 was correct for WDC. This is a third variant and it is more dangerous than either: the tag is neither zero nor absent — it is simply not updated, so every existence check passes. The scanner asserted recency on the cash leg (2026-03-31) and not on the debt leg. Completeness was checked; currency was not. A net_cash_complete: true flag on a two-quarter-stale debt figure is the "existence is not validity" failure class (calibration item D1) applied to a date rather than a value.

Recommended check for the scanner: assert that the debt leg's end date is within one quarter of the asset leg's, and fail the record loudly if not.

Defects found — Alpha Vantage

operatingIncome — same family as the KGS defect, much smaller magnitude

The excluded line is Acquisition and integration-related expenses, net ($16.0m in FY2025). This is the same shape as the KGS finding in this cluster — AV drops certain operating charges — and the magnitude scales with those charges (KGS: +21.9% with $67.9m of loss-on-sale and impairment; FSS: +4.4% with $16.0m of acquisition expense). Two names, same field, same direction, magnitude proportional to the excluded line. On XYL in the same cluster the field is byte-clean. So the defect is real, filer-specific, and predictable from the presence of below-SG&A operating charges.

gross_profit also differs: AV FY2025 28.1% vs a filed 28.95% (0.85pp). All opex lines in this memo were taken from the 10-K, so neither figure entered the model.

Fields checked and found CLEAN on FSS — reported per instruction

Defects found — our own tooling

normalized_fcf.py overstates Federal Signal's maintenance capex by 3.1x

The D&A proxy fails here because $18.4m of FY2025 D&A is Amortization expense on purchased customer relationships and trade names — a non-cash purchase-accounting charge, not a sustaining capital cost. The 20-year operating-cash-margin median fails because Federal Signal in 2006–2015 was a materially smaller and lower-margin company (operating margin 6.4% in 2012 against 16.0% today).

This is the same tool defect found on KGS in this cluster, from the opposite direction. On KGS the D&A proxy overstated maintenance capex 2.0x because of an acquisition asset step-up; on FSS it overstates 3.1x because of acquisition intangible amortisation. Both are acquisition artefacts, and normalized_fcf.py has no adjustment for either. Recommended fix: use disclosed capex where available and subtract acquisition intangible amortisation from the D&A proxy where it is not.

Used instead: 10.5% — trailing 12.4% haircut deliberately, because Federal Signal's growth requires cash acquisitions ($501.0m in FY2025 against $255m of operating cash flow) that never appear in capex.

steady_state_check.py — used as intended and it changed an input

At a first-tested terminal margin of 16.5% the tool returned TERMINAL_ABOVE_EVERY_OBSERVED_YEAR. The margin was moved to 16.0% — the TTM filed actual — and the flag cleared to NO_SCALING_CONTAMINATION_DETECTED. The independent reason for the move is that Q1 2026 gross margin was down 0.2pp year over year on acquired mix, so the gross-margin uptrend that would fund a higher terminal margin has stopped. Recorded because a flag that changes an input is the tool working, and because the change is documented with a reason that is checkable against the filing rather than asserted.

Model construction notes

• Organic growth is not disclosed and cannot be computed precisely. No ASC 805 supplemental pro forma revenue is given — "pro forma" appears exactly once in the FY2025 10-K and refers to a debt covenant. Model the growth rate as a bounded estimate (9.5%–11.3%), never as a point.
• Do not use FY2024 or any earlier year in a like-for-like growth series without removing Standard (2024-10-04), Hog (2025-02-12), New Way (2025-11-25), Kinloch (2025-12-04) and Mega (2026-01-16).
• Model two segments. ESG 85.2% of revenue at a 16.7% margin; SSS 14.8% at 25.4%. Corporate is −$13.0m per quarter and rising.
• Model the ESG product lines separately — vehicles/equipment, parts, rental, other. Aftermarket (parts + rental + other) is 20.05% of ESG revenue, down from 22.15%. Federal Signal does not disclose aftermarket margins; do not assume one.
• Backlog is quarterly and is the forward indicator. $1,037.5m at Q1'26 against $1,102.0m a year earlier. Note the seasonal shape: FY2025 year-end backlog was $1,042.4m against a Q1'26 of $1,037.5m, so Q1 is not a seasonal peak — the year-over-year comparison is the right one.
• Effective tax rate FY2025 = 77.9 / 324.5 = 24.0%. Used in ROIC and the exit-multiple identity.
• Capex is 1.2% of revenue. Growth is bought, not built: $501.0m of FY2025 acquisition cash against $28.7m of TTM capex. Any FCF model that treats capex as the only growth investment overstates distributable cash.
• Debt: use LongTermDebtAndCapitalLeaseObligationsIncludingCurrentMaturities, not LongTermDebt. See the screen defect above.