Verdict: KEEP (60/40)

QQQ weight 0.47% (rank 37 of 50 selectable). Not a filtration candidate. Structurally excellent business (V-Score 4.13, Bustamante 3/3) suffering a sector-wide derating that has already compressed the multiple 40%. Operational execution is solid — ES margins expanding, bookings growing, client satisfaction at record highs. The employment deceleration narrative is priced for worse than what's happening.

Factor Regression

Trailing 250 days:

SPY   β = +1.47   (58.9% of variance)
XLK   β = -0.63   (-26.7% of variance)
MTUM  β = -0.06   (-2.3% of variance)

α     = -40.7% annualized
σ_idio = 18.9%
R²    = 29.9%
Idio variance = 70.1%

ADP is a payroll infrastructure company trapped in a tech index. The negative XLK beta (-0.63) is the structural tell: when tech rallies on AI excitement, money rotates out of defensive names like ADP. When tech sells off, ADP relatively outperforms.

For the basket, this creates a useful property. Keeping ADP adds diversification against the tech concentration in QQQ. Removing it and redistributing 0.47% into tech-correlated survivors makes the basket marginally more concentrated in the direction it's already overweight.

The 70.1% idiosyncratic variance falls below the 75% target — returns are partly regime-dependent. But R² is only 29.9%, meaning 70% of the story is still company-specific. The trailing alpha of -40.7% is massive and backward-looking — it measures the damage already done, not what comes next.

What's Actually Happening

Two Different Businesses, One Stock Price

ADP reports two segments that are experiencing opposite dynamics. The market is pricing the whole company off the weaker one.

Employer Services (67% of revenue): Healthy and improving.

ES revenue $7,099M for H1 FY2026 (+7% reported, +5% organic constant currency). Q2 alone $3,607M (+6%, +5% organic CC). EBIT margin 35.4% in Q2, expanding +50 bps YoY. Full H1 margin 35.3%, flat to prior year despite absorbing $117M of incremental selling and marketing expense — meaning the underlying margin expansion is stronger than the headline.

Bookings are "solid" with "broad-based strength." The fastest growth came from international, US enterprise, and compliance businesses. Small business and mid-market also contributed. Client satisfaction hit its "single best quarter" in company history (Q2 FY2026 transcript). Retention "matched expectations" — still running at ≈92.1%, implying ≈13-year average client life.

R&D spending up 8% ($509M H1) — ADP is investing aggressively in Lyric HCM (enterprise platform, two 20K+ employee wins in Q2), Workforce Now NextGen (first 1,000+ employee sale), ADP Assist AI agents, and the WorkForce Suite integration from the Workforce Software acquisition. This is a company investing for growth, not retrenching.

PEO Services (33% of revenue): The problem child.

PEO revenue $3,443M (+6%), but EBIT $472M — down 1% YoY. Margin 13.7% for H1, compressing 100 bps. Q2 margin 14.4%, compressing 70 bps. The margin compression is real and the direction is wrong.

Benefits pass-through costs rose $153M (+7.3%) — healthcare cost inflation flowing directly through the P&L. Workers' compensation and state unemployment taxes added another $13M. Average worksite employees grew only 2% and decelerating. PEO bookings came in "slightly below expectations" (Q1 transcript).

This is confirmed across the sector: TNET worksite employees -8%, NSP missed four consecutive quarters, IBISWorld estimates the PEO industry declining at -2.5% CAGR over five years. PEO-lite competitors (Justworks at $99-149/PEPM vs ADP at $150-250) and EOR alternatives (Deel, Remote) are structural drags. This isn't a one-quarter blip.

Pays-Per-Control: The Critical Trajectory

Pays-per-control is ADP's same-store employment metric — the number of employees on existing clients' payrolls. It's the single most important organic growth indicator because ADP prices per employee.

Q1 FY2025 (Oct 2024): +2%
Q2 FY2025 (Jan 2025): +1%
Q3 FY2025 (Apr 2025): +1%
Q4 FY2025 (Jul 2025): +1%
Q1 FY2026 (Oct 2025): ≈0%  ← TROUGH
H1 FY2026 (10-Q):     +1%  → Q2 recovered to +1-2%
FY2026 guidance:       flat (conservative)

The deceleration from +2% to 0% drove the narrative that employment growth is dying and ADP's per-employee model is structurally impaired. But the inflection in Q2 — recovery from ≈0% back to +1-2% — is not in the consensus story. The 10-Q (L1568-1571) confirms H1 averaged +1%. If Q1 was ≈0% and H1 averaged +1%, Q2 was at or above +1%.

Management is guiding flat for the full year, which they do every time they're being cautious. From the Q2 transcript: clients are "cautious" on headcount in the "current environment" but "continue to hire."

Client Fund Interest: The Structural Floor

Interest on client funds: $595.5M for H1 FY2026 (+13.2% YoY). Average client fund balance $36.2B (+6.5%). Average interest rate 3.3% (up from 3.1%). Guided $1.31-1.33B for full FY2026 (raised $10M at Q2).

This matters because it's structural income from the OCC-chartered ADP Trust Bank — a national trust charter holding $67.5B in client funds. The portfolio is laddered; it reprices slowly on rate cuts. Even with 200 bps of Fed cuts (the current market expectation), the ladder provides 2-3 years of visibility. No competitor has this infrastructure. Rippling is still seeking state money transmitter licenses. Deel has MSB/money transmitter only. The regulatory moat around float income is probably ADP's single most durable competitive advantage.

Corporate fund interest income added another $198.3M for H1 ($96.8M in Q2 alone, up from $83.9M). Total interest income across both pools: ≈$794M for the half, annualizing to ≈$1.6B. At the current market cap of $82.7B, interest income alone represents a 1.9% earnings yield before any operating profit from actually running payroll.

Valuation

The market has derated ADP from ≈28-30x forward P/E to 17.1x — a 40% compression. This is severe for a business that's still growing revenue 6-7%, expanding ES margins, and delivering 9-11% EPS growth.

Forward P/E:    17.1x  (vs ≈28x 5-year avg)
Trailing P/E:   19.7x
Dividend Yield: 3.36%  (vs ≈2.0% historical)
FCF Yield:      ≈5-6%  (estimated from ≈$4.5B annual FCF)

Back-of-envelope segment valuation: ES generates ≈$5B annualized EBIT. At a 20x multiple (cheap for a 35% margin, 92% retention, 13-year client life business), ES alone is worth $100B. Current enterprise value is ≈$86B ($82.7B market cap + ≈$4B net debt). The market is assigning roughly negative value to PEO — or, more precisely, the derating reflects a belief that the entire business model is structurally impaired by AI headcount displacement and employment deceleration.

Is that belief correct? Over a decade, partially yes — AI will gradually reduce headcount at existing companies, shrinking ADP's same-store TAM. Over 15 weeks, the rate of AI displacement is effectively zero. The damage to the stock price is a narrative repricing, not a fundamental one.

Share buybacks provide a floor: $6B authorized, 3.2M shares repurchased in H1 at higher prices. At $205, every dollar buys substantially more shares. ADP returned $2.1B in H1 ($1.3B dividends + $0.9B buybacks). The dividend yield of 3.36% at this price is the highest in at least a decade.

Comparable

PAYX is the closest peer: -35.5% 1Y (worse than ADP), forward P/E 15.87, dividend 4.63%, RSI 33.3. Directors Doody and Bonadio made open-market purchases at ≈$99 in February 2026. These are Tier 1 signals — insiders who see the business daily putting personal capital at risk. Analyst consensus is neutral (14 Hold, 3 Sell, 1 Buy). The sector is universally hated.

PCTY (Paylocity) articulated a specific AI moat defense on Q2 FY2026: service moat (implementation handholding), data moat (payroll data can't be scraped), compliance moat (regulations change faster than AI adapts). The payroll/HCM sector is collectively making the case that AI is a tailwind (automates their operations) not a headwind (displaces their clients' workers). The market isn't buying it yet.

Counterparty Analysis

ADP is $83B market cap with deep analyst coverage. No informational edge exists on the basic story. Everyone knows PEO is weak, pays-per-control is decelerating, AI is a theoretical headwind.

The counterparty on a KEEP decision is the market participant who sold ADP at 17x because they believe AI headcount displacement will sustainably compress organic growth. They may be right on a 5-year view. On a 15-week view, they're front-running a secular trend with a massive positional bet against a company growing EPS 11%.

The forced seller is more interesting: ADP has fallen 30%+ from highs, likely triggering momentum-based exits, growth-mandate sales (no longer meets growth criteria at decelerated PPC), and passive rebalancing (falling market cap = lower index weight = selling). Forced selling creates temporary dislocations. At RSI 24.8, the forced selling may be exhausted.

The AI Headcount Displacement Question

This deserves direct engagement because it's the central narrative driving the derating.

The bear thesis: AI agents replace workers → fewer employees per company → ADP's per-employee TAM shrinks → organic same-store growth goes permanently negative → ADP is a structurally declining business.

The math: US nonfarm employment is ≈160M. If AI displaces 1% of jobs per year (aggressive), that's 1.6M fewer workers. ADP processes payroll for ≈42M workers (26% of US employment). At proportional exposure, ADP loses ≈416K workers/year from existing clients — roughly a -1% hit to pays-per-control, net of normal organic growth.

But: (1) displaced workers don't disappear — they get redeployed, often into new roles that still need payroll processing, (2) ADP's TAM is not just headcount — it's payroll complexity per worker, which increases with regulatory change, multi-state employment, benefits administration, compliance requirements, (3) ADP prices on "per employee per month" but also on product attach, and the trend toward more modules per client offsets headcount decline, (4) new business formation creates new ADP clients continuously.

The honest assessment: AI headcount displacement is a real but slow-moving headwind that reduces same-store growth by perhaps 0.5-1.0% per year, partially offset by complexity growth and product attach. It does NOT make ADP a structurally declining business in the next 1-5 years. A 40% multiple compression is pricing in a far worse outcome than the math supports.

Factor Scenarios

Idiosyncratic: adp-employment-model (DEMAND, 365d half-life)

Thematic: payroll-infrastructure-derating (sector factor, 180d half-life)

Composite expected return: idio component +1.6% + sector component +0.5% = +2.1% over 15 weeks. Modest positive, insufficient to drive a high-conviction call. The conviction comes from asymmetry at the tails: the 30% upside scenario (earnings beat) generates +12-18%, while the 45% inline scenario is roughly flat. You're getting paid for the optionality of a rerating without a large downside if execution merely continues.

What the 10-Q Doesn't Say

The 10-Q (filed Jan 29, 2026) contains zero tariff-specific language. Risk factors section (L2500-2502): "There have been no material changes in our risk factors." The forward-looking statement at L1504 mentions "overall market, political and economic conditions" generically but doesn't flag any emerging stress.

This matters because the 10-Q covers the period through December 31, 2025 — before peak tariff escalation in early 2026. Q3 FY2026 (January-March 2026) is the first period that captures the tariff uncertainty impact on hiring decisions. ADP has real-time visibility through the ADP National Employment Report; if tariffs are causing a hiring freeze, they'll know before anyone else. April 29 earnings will be the reveal.

Open Gaps

1. Q3 FY2026 pays-per-control direction. The most important number on April 29. If +1% or better → narrative inflection confirmed, rerating likely. If 0% or negative → employment deceleration resuming, re-evaluate KEEP.

2. PEO margin floor. Where does PEO margin compression stop? Healthcare cost inflation is running 7%+. If PEO margins drop below 12%, the segment becomes a material earnings headwind. No visibility until April 29.

3. Tariff impact on hiring. ADP sees this in real-time. Management commentary on April 29 about Q3 hiring trends is the canary for the broader economy.

4. AI displacement quantification. No public data on the actual rate of AI-driven headcount reduction at ADP clients. This is the central bear thesis and it's entirely unquantifiable from current filings. ADP could start disclosing this (they have the data). Until they do, the market will price worst-case.

Filtration Call

KEEP. ADP is not one of the weakest names in the selectable universe. The business is executing well — ES margins expanding, bookings growing, satisfaction at records, PPC inflecting positively. The severe derating (17x forward vs 28x historical) prices in substantial structural impairment that hasn't materialized in the numbers.

The risks are real: PEO is structurally declining, AI headcount displacement is a legitimate long-term concern, and tariff-driven recession could cause PPC to go negative. But these risks argue for monitoring, not removal. At 0.47% weight, the P&L impact is ≈5 bps in either direction. The natural hedge property (negative XLK beta) adds diversification value to a tech-heavy basket.

The trigger to flip: April 29 earnings showing PPC ≤ 0% with softening bookings and management flagging tariff-driven hiring weakness. Until then, the math says keep.