NYSE:FIX Comfort Systems USA, Inc.

Lifecycle & sector classification. Comfort Systems is an Industrials name (GICS: Engineering & Construction) scored on the Growth lifecycle profile. That label needs a caveat, and it is a change from the 25 July report, which used "established". The framework's Growth band assumes a business scaling towards profitability; Comfort Systems is the opposite — a mature contracting model running at hyper-growth revenue rates (+50.3% year on year) while already earning a 12.8% net margin. So we take the Growth classification for the growth-rate arithmetic but keep the Industrials metric set — ROIC versus cost of capital, backlog growth, operating margin, capex intensity — rather than the high-growth-software set. Using unit-economics metrics on a company that has been profitable since the 1990s would be directionally wrong.

Competitive Moat Scorecard

Moat score = 62/100 (average of the five). A good business with a moderate moat, not a fortress. What protects Comfort Systems is scarcity of capacity, and scarcity is a cycle condition, not a structural one.

Net effect on the moat: → Switching Costs trimmed to 58 (project-by-project re-bidding against a credible national rival), Cost Advantage held at 68 (real but replicable). Overall competitive_threat_level = moderate-rising, share trajectory = gaining. This propagates: the §11 Bear card carries an EMCOR/labour-driven margin-compression trigger, and the §12 thesis-invalidation rule carries an explicit competitive condition.

ROIC & Capital Allocation

Quality = 86/100 (confidence 78%). The benchmark and ROIC legs are close to maxed, the moat is merely good, and management alignment is the drag. It moved +1 from 85 — essentially unchanged, which is the correct answer when the only new information since the last report is a backlog print that confirmed what was already known.

Implied-growth read (narrative colour, not the score). Reverse the anchor: to warrant 43.63× at r = 11.13% and g_term = 3%, you need roughly 19–20% earnings growth sustained for a full decade, not five years. Consensus has 22.6% then 25.7% for the next two years and no visibility beyond that. The market is not paying for the backlog it can see — it is paying for a second backlog nobody has booked yet. That may well arrive. It is not, on this framework, something you underwrite at 43×.

Weighted: 12(0.40) + 22(0.20) + 25(0.15) + 42(0.10) + 65(0.15) = 26.9 → 27. The relative lenses can order the name within the Expensive band; they cannot lift it out of one. Score up 2 points from 25 on 25 July, purely because the forward multiples came in (37×/32× then, 36.2×/29.5× now) as estimates rose and the consensus target expanded.

Analyst Price Target Consensus

Median US$2,110. Spread high/low = 1.24× — narrow, so analysts broadly agree; no confidence penalty for dispersion. Coverage is the issue: 6 targets in the last quarter, 3 in the last month — below the 5-analyst thin-coverage line on the monthly count, so we take a 5% confidence haircut. The consensus endpoint was checked for degeneracy (high = low = median = consensus) and is not degenerate, so no Yahoo fallback was required — though Yahoo's panel (mean US$2,210.86, n=7) was pulled anyway and reads higher, not lower.

Analyst Grades Distribution

55.6% bullish, zero Sell ratings — the "Buy consensus with >30% holds" band: analysts lean positive with real hesitation, and the hesitation is about price, not the business. Grade actions in the last 30 days: two maintains (KeyBanc Overweight and UBS Buy, both 27 July) and nothing else. No upgrades, no downgrades — a neutral sentiment reading. The last rating change of any kind was KeyBanc's Sector Weight → Overweight upgrade on 24 April.

FMP Ratings Cross-Reference

An independent framework, reached from different inputs, lands in exactly the same place we did: the operating scores are maximum, the valuation scores are minimum. That convergence is the strongest single argument that Valuation 27 alongside Quality 86 is not an internal contradiction — it is the shape of the situation.

Risk-Reward Assessment

Traditional indicators average 59.

Relative Strength Benchmarking

Relative strength = 15/100. Underperforming both benchmarks on both windows is the framework's bottom band, and there is no way to dress it up: over three months Comfort Systems has lost 13% while its own sector gained 7%. Whatever is repairing on the daily chart, this is still a laggard, and the sector's Strong-Outperform macro signal is being expressed by other names.

Position-Risk Signal

Nearest daily support is US$1,705.00, then 1,659.28 and 1,621.01. A logical stop sits just under the 1,659.28 swing low, at US$1,640 — US$134 away = 1.37 ATR, inside the 1.5-ATR "favourable" line. On stop distance alone this scores 78.

Two adjustments pull it down. First, price is within 3% of resistance. The authoritative swing-high set from get_technical_indicators is 1,824.98 / 1,879.99 / 1,998.81 / 2,039.56 / 2,072.31 — so the nearest overhead level is US$1,824.98, just +2.87% away, inside the 3% band that carries a −15 penalty. And the binding level is nearer still: the falling 50-day at US$1,794.49 is only +1.15% overhead, and §9 and §12 both treat it as the level that must be reclaimed. Buying 1.15% under a falling 50-day and 2.9% under the nearest swing high is a poor entry location, and the score has to say so. Second, no swing-low bonus. The last major swing low was Friday 7 August at 1,659.28 — five sessions ago, at the outer edge of the framework's 5-day window — and the rebound has run on below-average volume throughout. That is not the fresh, volume-confirmed bounce the bonus is designed to reward, so no bonus is applied.

Position-risk = 78 − 15 = 63.

Risk-reward composite = (59 + 15 + 63) / 3 = 45.7 → 46. A tight stop is available; the relative weakness and the overhead supply are what hold the leg down.

Macro Regime Overlay (weight 15%)

Sentiment Layer (weight 18%)

• Analyst grades & revisions (primary): two maintains in the last 30 days (KeyBanc Overweight, UBS Buy, both 27 July), zero upgrades, zero downgrades → the "all maintain" band, 55. The last actual change was KeyBanc's upgrade on 24 April.
• Estimate revisions: rising. FY26E EPS US$48.97, FY27E US$60.04, FY28E US$75.46 — and the quarterly beats have been enormous (Q1 actual US$10.51 against a US$6.78 consensus; Q2 US$12.53 against US$10.42). Bullish, 75.
• News tone (recency-weighted): heavily positive. Seven of nine FIX-specific Polygon sentiment tags in the last 90 days read positive, most recently a 14 August piece framing it as the under-the-radar AI data-centre play. Two dissents matter: a 13 May piece flagging stretched valuation against PwC's late-decade decline forecast, and a negative tag on Third Point cutting its position 47%. Positive but crowded — and a crowded consensus on a 43× stock is itself a risk. 62.
• Options skew / retail: not available through the tools used; no score.

Sentiment = 56 (grades weighted heaviest as the framework's primary signal; the news tone discounted for crowding).

Catalyst Layer (weight 17%)

Catalyst clustering = 62. No company-specific catalyst inside 30 days and no earnings blackout, which is calm; but a steady run of macro releases that a discount-rate- sensitive multiple reacts to, and the two events that actually decide this name — Q3 earnings and hyperscaler 2027 capex guidance — both land in late October, outside the window. That is the "one clear catalyst" band rather than the fully calm band. No position-size reduction required on catalyst density.

Upcoming events (next 30 days)

Recent surprises (last 7 days)

Industrials carries MEDIUM macro sensitivity, so the 3-trading-day WAIT-for-event override that applies to Materials, Financials and rate-linked REITs does not apply here — and nothing in the next fourteen days is a high-impact release for this specific name in any case. The one genuinely relevant pair is FOMC minutes on 19 August and core PCE on 26 August, and the transmission is entirely through the discount rate: at 43.63× trailing earnings, this stock is a long-duration asset, and the 10-Y at 4.63% is exactly what forces the 11.13% discount rate that makes the warranted multiple 16.9×. A hot PCE print does not touch Comfort Systems' backlog; it makes the backlog worth less today.

The last week's data was a genuine split: the real economy weakened badly (payrolls −23k, retail sales −0.6%, Michigan sentiment 51.0) while inflation came in exactly on forecast and PPI at zero. For most cyclicals that combination is a problem. For this one it is close to neutral — its four biggest customers are executing capital plans set in boardrooms, not responding to consumer confidence — and the soft PPI is a mild positive on input costs. No economic release in this window changes the signal. The events that would are hyperscaler Q3 capex guidance and the company's own Q3 backlog print, both in late October.

Weighting note, stated up front. The hourly and 15-minute frames came back with volume_ratio of 0.00 and 0.01 — degenerate volume data. Rather than take 86 and 82 off unverifiable bars at a combined 20% weight, this report uses the framework's own no-intraday fallback: Monthly 35% / Weekly 30% / Daily 35%. That gives 78(0.35) + 71(0.30) + 52(0.35) = 67. Including the intraday frames at the standard weights would have produced 71 — a 4-point difference that changes nothing, but the choice is disclosed rather than buried.

What the table says. The monthly and weekly frames are unambiguously bullish: price is far above every long-term average (SMA200 US$1,441.69, some 23% below spot) and both frames show a resistance breakout. The monthly RSI at 76.3 is deeply overbought, which is what a 2,000% five-year move looks like on a long chart — informative, not immediately actionable. The weekly MACD histogram has turned negative (−43.6), the first sign of the intermediate trend losing thrust.

The decision timeframe is the daily, and it is honestly mixed. Price at US$1,774.06 is above the 20-day (1,722.06) but below a falling 50-day (1,794.49, down from 1,816.37 on 3 August) — the textbook 'weakening' configuration. The multi-timeframe tool still carried a support breakdown flag from its 13 August bar (close 1,727.64); the 14 August close of 1,774.06 reclaimed that level, but one session does not repair a structure.

Cross-timeframe pattern: higher-timeframe bullish with a lower-timeframe pullback in progress. On the framework's own reading that is normally a buy setup — but the qualifier matters here: the pullback has already retraced 19% off its low, so price is no longer at support, it is mid-range and 1.1% under a falling 50-day. The levels that matter: reclaim and hold US$1,794.49 (the 50-day) turns this constructive; two closes below US$1,659.28 resumes the breakdown.

Six months of daily closes (Polygon, RAW unadjusted) to the 14 August 2026 close, sampled every second session — 63 plotted points drawn from 125 trading days — with the 50-day simple moving average overlaid (the 50-day line begins once 50 bars of history are available). The shape of the last four months is the whole argument: a vertical run to an intraday high of US$2,073.99 on 14 May, a lower high closing at US$2,066.51 on 22 June, then a descending sequence into the US$1,491.24 intraday low on 29 July, and a +16.7% close-to-close recovery that has so far stalled just under the falling 50-day at US$1,794.49. Marked: the hard stop (US$1,640), the 7 August swing-low support that defines it (US$1,659), the 50-day that must be reclaimed, and the US$2,057.86 analyst consensus.

Forecast:

ENTRY — Fundamental group: price below US$828 fair value
→ FORECAST: Unlikely on any horizon short of a cohort de-rating.
→ BASIS: US$828 is 53% below spot. The only path that reaches it is the §11 bear (multiple compression towards 20×), and even that lands at US$1,080. Note what would move the target rather than the price: rolling the anchor onto FY27E EPS of US$60.04 lifts fair value to about US$1,015, and each year of ~23% earnings growth lifts it a further ~20%. On current estimates the anchor's fair value crosses US$1,700 around 2029.
→ CONFIDENCE: Unlikely — the trajectory moves the threshold up towards the price rather than the price down to the threshold.

ENTRY — Technical group: a close above the 50-day SMA on >1.5× volume
→ FORECAST: 1–3 weeks at the current trajectory.
→ BASIS: The gap is small and closing from both directions. Price is US$20.43 (1.1%) below the 50-day, and rising — up US$99.22 over the four sessions from 10 August. The 50-day itself is falling at roughly US$4/session (1,816.37 on 3 Aug → 1,794.49 on 14 Aug), so the two converge even on flat price. A single average day closes the price gap. The binding condition is the volume qualifier: 1.5× the 20-day average means about 700k shares, and the last four sessions have run 272k–395k. That is catalyst-dependent, and there is no company catalyst before late October.
→ CONFIDENCE: Moderate on the price condition, Low on the volume condition — which is why the group forecast is 'reclaim likely, confirmation uncertain'.
→ RISK: a failed reclaim at the 50-day is the classic lower high. Two closes below US$1,659.28 invalidate it.

ENTRY — Technical group, pullback branch: a tested bounce at support
→ FORECAST: catalyst-dependent, not time-projectable.
→ BASIS: it needs a retest of US$1,659–1,705 with a higher low. From US$1,774.06 that is a 3.9–6.5% pullback, or 0.7–1.2 ATR — an ordinary week's range for this stock. The likeliest trigger inside the window is a hot core PCE print on 26 August.
→ CONFIDENCE: Moderate.

ENTRY — Catalyst group: a >+5% post-earnings move on >2× volume
→ FORECAST: ~22 October 2026 (estimated Q3 date; not calendar-confirmed).
→ BASIS: purely event-driven. The company has beaten consensus heavily for four straight quarters (Q1 US$10.51 against US$6.78; Q2 US$12.53 against US$10.42), so the probability of a beat is high. The probability of a beat producing a +5% move is not: the last two beats were sold. The condition that would actually change the tape is the backlog print, not the EPS line.
→ CONFIDENCE: Low — high beat rate, poor reaction record.

Ladder outcome: 0 of 3 groups met → WAIT. Unchanged from 25 July. The framework's own language for this state applies exactly: good business, no entry edge now. That is not a contradiction of Timing 59 — Timing says the tape is repairing; the ladder says none of the three defined paths is open at US$1,774.06.

Forecast:

EXIT — Stop-Loss: two closes below US$1,659.28
→ FORECAST: unlikely in the next 4–6 weeks at the current trajectory.
→ BASIS: the trigger is 6.9% below spot, or 1.17 ATR, and momentum is currently pointing the other way (MACD histogram positive ten sessions, higher swing lows since 29 July). Reaching it requires a fresh catalyst, not drift.
→ RISK TRIGGER: a hot core PCE print on 26 August, or a hyperscaler capex signal at any point. The 27–29 July precedent shows this stock can travel −12.3% in three sessions on sector-wide capex scrutiny rather than on any news of its own.

EXIT — Thesis Invalidation
→ FORECAST: not before the Q3 print (~22 Oct). Every leg is a quarterly disclosure; none can change intra-quarter. The one to watch is the backlog line, and the specific number is US$14.0bn.
→ CONFIDENCE: High that it stays clear through the window; the informational content arrives all at once in late October.

EXIT — Profit-Target: US$2,110 with RSI > 70
→ FORECAST: unlikely inside 3 months. It needs +19.0% and an overbought RSI simultaneously; the stock has not seen RSI above 70 on the daily since May.
→ NOTE: this is the one exit condition a holder should actively want to hit.

Exit outcome: 0 of 3 groups live → HOLD. Nothing mechanical says sell today. Read that precisely: the Valuation-Ceiling gate blocks buying; it does not generate a sell.

Position sizing not computed — no risk budget or portfolio role was specified for this watchlist name, and the framework does not invent one. The §12 Conviction Ladder reads Wait (0 of 3 entry paths met), which sets the ladder factor to 0×: there is no entry edge at US$1,774.06, so the honest guidance is the levels to watch rather than a percentage.

Levels instead of a percentage. A reclaim of the 50-day at US$1,794.49 on volume above roughly 700k shares opens the Technical path; a retest of US$1,659–1,705 with a higher low opens it at a better price. The hard stop is US$1,640. The Fundamental path needs US$828 and is not reachable without a cohort de-rating.

Volatility context — this is the part to take seriously. ATR(14) is US$98.20 = 5.54% of price: a normal day moves this stock more than a normal month moves a utility. Beta is 1.701, so a 5% position behaves like an 8.5% position in risk terms against the market. The maximum drawdown over the last twelve months was −28.1% (an intraday US$2,073.99 on 14 May to an intraday US$1,491.24 on 29 July), and it took eleven weeks. The S&P's own realised volatility over the same window, with VIX at 14.63, is a fraction of that. Anyone sizing this name off its quality score rather than its volatility will be surprised by the P&L swings.

Catalyst clustering is 62 — above the 50 line, so no additional size reduction on event density would be warranted even if a path were open.

{
  "ticker": "FIX",
  "date": "2026-08-15",
  "version": "v6",
  "brand": "",
  "company": "Comfort Systems USA, Inc.",
  "exchange": "NYSE",
  "exchange_ticker": "NYSE:FIX",
  "isin": "US1999081045",
  "api_ticker": "FIX",
  "currency": "USD",
  "finder_ticker": "FIX",
  "finder_exchange": "\ud83c\uddfa\ud83c\uddf8 NYSE",
  "sector": "Industrials",
  "gics_sector": "Industrials",
  "analysis_status": "on-going",
  "lifecycle_stage": "growth",
  "price_at_rating": 1774.06,
  "prior_price_at_rating": 1733.6,
  "eps_trailing": 40.66,
  "trailing_pe": 43.63,
  "signal_short": "HOLD",
  "signal_medium": "HOLD",
  "signal_long": "HOLD",
  "primary_signal": "HOLD",
  "user_horizon": null,
  "short_entry_confirmed": false,
  "short_hold_reason": "expensive",
  "short_cap_reason": "moot \u2014 no BUY at any horizon. The base matrix returns HOLD on the High-Quality/Expensive row before any cap, and the Valuation-Ceiling gate independently caps all three horizons. Technical and Catalyst entry groups are both UNMET, so short_entry_confirmed=false is recorded, but no short-horizon cap was applied. The quality-starter override is barred: it requires Valuation >= 40 (not Expensive).",
  "quality_score": 86,
  "valuation_score": 27,
  "timing_score": 58,
  "driver_score": 83,
  "driver_label": "Strong Tailwind (amplification-eligible but blocked \u2014 base signal HOLD, and Expensive band bars STRONG BUY)",
  "driver_commodity_trend": "n/a \u2014 non-commodity driver (AI / hyperscale data-centre capex cycle)",
  "quality_detail": {
    "industry_benchmark_name": "ROIC vs WACC + Backlog Growth (Industrials)",
    "industry_benchmark_value": "ROIC ~44% vs WACC 11.13%; backlog +73.2% YoY",
    "industry_benchmark_score": 95,
    "moat_score": 62,
    "roic_pct": 44.0,
    "roic_basis": "NOPAT US$1,444m / invested capital US$3,274m (equity US$3,220m + REPORTED total debt US$54m, not FMP's lease-inclusive US$303.9m)",
    "net_cash_usd_bn": 1.8,
    "total_debt_reported_usd_m": 54,
    "total_debt_fmp_lease_inclusive_usd_m": 303.9,
    "capital_allocation": 82,
    "management_skin_in_game": 42,
    "backlog_usd_bn": 14.06,
    "backlog_yoy_pct": 73.2,
    "backlog_qoq_pct": 12.9,
    "backlog_to_ttm_revenue": 1.25
  },
  "valuation_detail": {
    "fcf_yield": 3.55,
    "fcf_yield_yahoo_basis": 2.84,
    "implied_growth_rate": 19.5,
    "consensus_growth_rate": 22.6,
    "historical_valuation_decile": 8,
    "fwd_pe_fy26": 36.2,
    "fwd_pe_fy27": 29.5,
    "peer_pe_eme": 26.13,
    "peer_pe_iesc": 33.59,
    "ev_ebitda": 31.1
  },
  "timing_detail": {
    "mtf_confluence": 67,
    "risk_reward_score": 46,
    "position_risk_score": 63,
    "composite_short": 56,
    "composite_medium": 57,
    "composite_long": 64,
    "relative_strength_vs_spy": -3.3,
    "relative_strength_vs_sector": -3.4,
    "relative_strength_vs_spy_3m": -16.9,
    "relative_strength_vs_sector_3m": -20.0,
    "catalyst_clustering_score": 62,
    "dynamic_macro_weight": 0.15,
    "rsi_14": 52.89,
    "sma50": 1794.49,
    "sma200": 1441.69,
    "atr_14": 98.2
  },
  "relative_strength_vs_spy": -3.3,
  "relative_strength_vs_sector": -3.4,
  "economic_alignment_stance": "Trend-Following",
  "economic_alignment_conviction": 78,
  "economic_alignment_pressure": "Tailwind",
  "economic_alignment_source": "sector-map",
  "macro_report_date": "2026-08-12",
  "warranted_multiple": 16.9,
  "actual_multiple": 43.63,
  "val_multiple_basis": "clean trailing P/E on US$40.66 diluted TTM EPS (fwd 36.2x FY26E / 29.5x FY27E \u2014 all three above the 23x Industrials guardrail)",
  "warranted_ratio": 2.58,
  "val_band": "expensive",
  "discount_rate_r": 0.1113,
  "risk_free_10y": 0.0463,
  "g_near": 0.1,
  "g_term": 0.03,
  "nonop_pct_of_net_income": -0.8,
  "clean_pe": 43.63,
  "clean_peg": 1.93,
  "competitive_share_trajectory": "gaining",
  "competitive_threat_level": "moderate",
  "hard_gate_state": "caution",
  "gates_triggered": [
    "Valuation Ceiling (43.63x clean trailing P/E vs 16.9x warranted = 2.58x; also >= the 23x Industrials guardrail on trailing, FY26E and FY27E)"
  ],
  "gates_caution": [
    "Insider selling (US$152m / 12m, zero purchases) \u2014 below the DNB Trigger-4 bar"
  ],
  "do_not_buy_triggers": [],
  "dnb_checked": "Trigger 1 (leverage + rising rates): NOT fired \u2014 net cash US$1.80bn (cash US$1.85bn less REPORTED total debt US$54m; FMP's lease-inclusive US$303.9m still leaves net cash US$1.55bn), interest coverage 196.7x. Trigger 2 relative arm: NOT fired \u2014 43.63x sits ~8th decile of the 5-yr range, not the top decile (47-48x printed Apr-May 2026). Trigger 2 absolute arm (a) deep-expensive-alone: BOTH numeric legs are satisfied (2.58x >= 2.0x warranted; 43.63x >= 1.5 x 23x guardrail = 34.5x) and the trigger is held off SOLELY by the exceptional/proven/durable-growth carve-out. Evidence: revenue +50.3% YoY (US$3,265.7m vs US$2,173.3m), diluted EPS +91.9% (US$12.53 vs US$6.53), backlog US$14.06bn = +73.2% YoY and +12.9% QoQ, of which US$13.70bn (97%) is SAME-STORE i.e. organic not acquired, backlog = 1.25x TTM revenue (~15 months of signed work), gross margin 25.87% in Q2 confirming the Q1 expansion was structural rather than a one-off close-out gain, and forward consensus of +22.6% (FY27, n=7) then +25.7% (FY28). DURABILITY, argued head-on because the carve-out requires it and S5 separately scores the forward driver leg only 68: the durability evidence is CONTRACTED, not forecast \u2014 US$14.06bn of signed backlog equal to 1.25x TTM revenue, i.e. ~15 months of work already booked, carrying to roughly end-2027, and 97% of it same-store rather than acquired. That is what separates 'proven and durable' from 'extrapolated'. The 68 on the driver's forward leg is about the NEXT backlog (2028+), which is genuinely unknowable and is precisely why this is a HOLD and why the thesis-invalidation floor is a bookings number (sequential backlog at or below US$14.0bn). Trigger 2 absolute arm (b) expensive + live de-rating catalyst: NOT fired, on three grounds. (i) FIX IS a genuine AI-capex-cohort member (~75% of revenue technology+industrial) and therefore inherits the cohort de-rating leg in the S11 Bear, which it carries in full (43.63x -> 20x, -39.1%). (ii) But the macro report's armed tail is specifically an EARNINGS-QUALITY / index-concentration unwind, and FIX has no earnings-quality gap for it to expose: nonop_pct_of_net_income is NEGATIVE at -0.8%, i.e. reported earnings are understated versus operating, and FIX is not a top-weight index constituent. (iii) The tail's trigger is RECEDING per the 2026-08-12 macro report (breadth broadening: RSP +3.6% vs SPY +2.8% vs QQQ +0.6% over 1mo). An armed-but-not-triggering systemic tail is not a LIVE de-rating catalyst; it caps via the Valuation-Ceiling gate, not via a hard prohibition. Trigger 3 (persistent negative revisions): NOT fired \u2014 estimates are RISING (FY26E 48.97, FY27E 60.04, FY28E 75.46). Trigger 4 (insider selling spike): NOT fired \u2014 the bar is 3+ C-suite selling >25% of holdings inside 60 days. The three-insider cluster (CEO Lane, CFO George, director Myers) was late FEBRUARY, ~6 months ago; the only sales inside the 60-day window are one DIRECTOR's (Myers, 7,500 shares, 10-11 Aug). Recorded as a Quality drag (management skin-in-game 42) and a caution gate, not a DNB. Trigger 5 (structural business-model threat): NOT fired \u2014 competitive threat is moderate and FIX is GAINING share (backlog +73% vs EMCOR RPO +44%; revenue +50% vs +20%).",
  "entry_groups_met": 0,
  "entry_conviction": "Wait",
  "exit_groups_live": 0,
  "exit_action": "Hold",
  "fair_value_est": 828,
  "fair_value_basis": "warranted 16.9x x FY26E EPS US$48.97 (TTM basis gives US$687; FY27E basis gives US$1,015)",
  "stop_loss": 1640,
  "target_price": 2058,
  "scenario_base_target": 1950,
  "scenario_bull_target": 2400,
  "scenario_bear_target": 1080,
  "scenario_probabilities": {
    "bull": 25,
    "base": 55,
    "bear": 20
  },
  "scenario_weighted_value": 1888.5,
  "analyst_consensus_target": 2057.86,
  "analyst_target_high": 2225,
  "analyst_target_low": 1800,
  "analyst_target_median": 2110,
  "analyst_target_upside_pct": 16.0,
  "analyst_grades_consensus": "Buy",
  "analyst_bullish_pct": 55.6,
  "analyst_coverage_count": 9,
  "fmp_rating": "B+",
  "fmp_overall_score": 3,
  "recent_upgrades_30d": 0,
  "recent_downgrades_30d": 0,
  "quality_confidence": 78,
  "valuation_confidence": 74,
  "timing_confidence": 62,
  "driver_confidence": 66,
  "overall_confidence": 62,
  "next_update_date": "2026-08-31",
  "next_update_basis": "default +14d from 2026-08-15 (2026-08-29 is a Saturday, rolled to the next trading day). Q3 earnings est. ~2026-10-22, inferred from filing cadence \u2014 far beyond the 14-day cap; no dated company catalyst inside the window",
  "next_check_date": "2026-08-31",
  "price_as_of": "2026-08-14 close (2026-08-15 is a Saturday; US markets shut)"
}

In plain English. Comfort Systems is one of the best-run businesses on the watchlist — Quality 86, roughly 44% return on invested capital against an 11.13% cost of it, US$14.06bn of signed backlog up 73% in a year, US$1.80bn of net cash, and a share count that is shrinking. It is also priced at 43.63× trailing earnings against a warranted 16.9× and an industry rich line of 23×. Both of those things are true at once, and the framework's answer when they are is HOLD on all three horizons: a great business at the wrong price.

What changed since 25 July. Almost nothing fundamental — Q2 was already in the last report. Quality 85 → 86, Valuation 25 → 27 (forward multiples came in as estimates rose), Driver 80 → 83, Economic Alignment 73 → 78. The one big move is Timing 33 → 58, and it is a tape story, not a re-score: the post-earnings breakdown the last report was scoring into completed at a US$1,519.96 close on 29 July (intraday low US$1,491.24), and price has since recovered 16.7% close-to-close on higher swing lows. The signal, the gate, the ladder and the exit action are all unchanged.

The one judgement call. Do-Not-Buy Trigger 2 is numerically satisfied on both arms — 2.58× warranted is past the 2.0× line, and 43.63× is past 1.5× the 23× guardrail. It does not fire because the trigger carves out exceptional, proven, durable growth, and this is what that looks like: revenue +50.3%, diluted EPS +91.9%, backlog +73.2% year on year and +12.9% sequentially, 97% of it same-store rather than acquired, and 1.25× a full year's revenue already signed. Arm (b) — Expensive plus a live de-rating catalyst — does not fire either: the AI-concentration tail is armed but its trigger is receding, and Comfort Systems has no earnings-quality gap for an earnings-quality unwind to expose (non-operating income is −0.8% of net income). An armed-but-not-triggering tail caps through the Valuation-Ceiling gate; it does not manufacture a hard prohibition. Reasonable people could disagree, and the §11 bear carries the cohort de-rating leg in full so the risk is visible either way.