V-Score Card

TICKER:             PD (PagerDuty, Inc.)
V-SCORE:            1.72
VERDICT:            COLLAPSED
kappa:              0.00

DIMENSIONS          SCORE   WEIGHT   CONTRIBUTION
C  Compound Cognition    2    0.25     0.50
E  Irreducible Infra     2    0.22     0.44
U  Ecosystem Breadth     2    0.18     0.36
A  Distribution          2    0.12     0.24
M  Ecosystem Gravity     2    0.15     0.30
F  Friction (penalty)    2   -0.06    -0.12
                                      ─────
RAW                                    1.72

GATES
G1 = 1[E > 1] = 1[2 > 1] = 1         PASS
G2 = 1[A > 1 | C+E+U >= 12]
   = 1[2 > 1 | 6 >= 12] = 1           PASS

V = 1.72 x 1 x 1 = 1.72              DEAD ZONE

FAST SCREEN (Bustamante)
  Proprietary data unsynthesizable locally?    NO
  Regulatory mandate requiring routing?        NO
  Transaction-embedded (IS the rail)?          NO
  b(s) = 0/3 --> strong prior E <= 3

Regime Context

WINDOW:   2025-12-17 to 2026-04-07 (T = 15 weeks, 75 trading days)
MODEL:    r_PD = alpha + beta1 * r_SPY + beta2 * r_IGV_orth + epsilon

alpha (ann)    = -211.1%     (t = -3.20, p = 0.002)
sigma_idio     =   35.4%     (two-factor residual)
IR             =  -5.96      (statistically significant negative alpha)
beta_SPY       =   1.032
beta_IGV_orth  =   1.434
R^2            =   0.555
%Idio Var      =  44.5%      (sector-dominated regime)
rho_intra      =   0.574     (elevated, NOT indiscriminate)

WINDOW RETURNS
  SPY:  -2.3%
  IGV: -24.3%    (SaaS sector excess: -22%)
  PD:  -50.2%    (PD excess vs IGV: -25.9%)

The IR measures the regime, not the stock. But in this regime, the IR is doing double duty: the SaaS sector sold off hard (IGV -24.3%), and within that selloff the market differentiated. Intra-sector correlation at 0.574 is elevated from normal (≈0.2-0.3) but well below the indiscriminate threshold (≈0.8+). Idiosyncratic signal is present.

Peer IR confirms the market is sorting:

Ticker    IR     Assessment
CRWD    -0.57    Defensible -- sector drag only
DDOG    -0.54    Defensible -- sector drag only
HUBS    -2.00    Moderate structural repricing
ESTC    -2.13    Moderate structural repricing
NOW     -2.93    Repricing (even Fortress names hit)
MNDY    -4.02    Severe structural repricing
PD      -5.96    Severe -- worst in peer group
TEAM    -4.58    Severe structural repricing

PD's -25.9% excess return vs IGV is idiosyncratic. The market isn't applying a uniform discount -- it's pricing structural vulnerability name by name. CRWD and DDOG, with platform-grade infrastructure and embedded distribution, absorbed the sector selloff with minimal idio damage. PD, with computable tasks and declining NRR, got structurally repriced.

delta = V_scored - V_market ~ 0. At EV/Rev 0.47x (vs SaaS median 8-12x), the market already prices Dead Zone. No mispricing identified.

Dimension Analysis

C -- Compound Cognition: 2/5

PagerDuty's domain is narrow and fully computable. The task set -- alert routing, escalation logic, on-call scheduling, event correlation, noise reduction -- consists of well-defined rules and pattern matching. A competent team re-derives core incident management in weeks, not months.

The company's own 10-K (line 1142-1143) admits it: "Foundation models and agentic frameworks via open-source releases may lower barriers to entry." When management tells you their cognition is replicable, believe them.

Four platforms have already proven the re-derivation cost is low: Datadog built On-Call, ServiceNow built Incident Management, Atlassian has Opsgenie, Freshworks acquired FireHydrant. If your compound cognition can be absorbed as a feature inside someone else's platform, it wasn't compound.

Superlinearity test: adding Process Automation to Incident Management does not create emergent cross-domain complexity. The products are serial extensions (incidents trigger runbooks trigger status pages), not a mesh where removing one module breaks five others. Compare SAP, where removing payroll breaks tax reporting breaks compliance breaks HR workflows.

Calibration: DocuSign (C=2) and UiPath (C=2-3) are comparable. PD's domain is simpler than RPA workflow encoding (UiPath) but more developed than pure e-signature (DocuSign).

E -- Irreducible Infrastructure: 2/5

The dominant signal is empirical: NRR declined from 107% to 98% over five consecutive quarters at -2pts/quarter (Q3 FY2025 transcripts through Q4 FY2026). If infrastructure were irreducible, customers could not leave. They are leaving.

No task in PD's domain has infinite local re-derivation cost. No regulatory mandate forces routing through PagerDuty. FedRAMP authorization is Low -- the weakest tier. The 10-K contains zero mentions of "switching cost," "lock-in," "portability," or "migration barriers" across its entire 150+ pages. The company does not even claim stickiness as a competitive advantage.

Contracts are typically one year (10-K line 1326). Only 5% of revenue is multiyear (Q2 transcript). Customers have "no obligation to renew" (10-K line 1324). RPO beyond 12 months is just $163M. Management "underestimated current headwinds [to] retention" (Q3 transcript) -- they can't predict their own churn, which tells you the infrastructure is not durable.

Stress test: the case for E=1 (all tasks computable, G1=0, V=0) is legitimate. Every task in the domain is computable and four competitors replicated the infrastructure in months. E=2 is held because real-time multi-cloud alerting infrastructure creates non-zero switching cost, even if that cost is measured in days rather than months. This is the ceiling, not the floor.

U -- Ecosystem Breadth: 2/5

PagerDuty serves one domain (incident management) across 1-2 departments (Engineering/IT Ops, with Customer Support as secondary). 70% of ARR comes from core incident management (Q2 transcript line 125). The remaining products -- AIOps, Process Automation, CSOps, Status Pages, PagerDuty Advance -- are serial extensions of the same workflow, not independent capabilities serving different departments.

Switching cost is linear, not superlinear. Removing AIOps doesn't break Incident Management. Removing Automation doesn't break AIOps. There is no mesh of interdependencies that would make phi_switch scale as |W(s)|^alpha with alpha > 1.

Compare ServiceNow (U=4-5): ITSM, HR Service Delivery, Customer Service, Security Operations, Legal, Risk -- each independently valuable, each serving different buyers. PagerDuty's "Operations Cloud" narrative describes one cloud serving one persona.

A -- Distribution: 2/5

PagerDuty has 700+ integrations and MCP server support. Agents can find it. But the compounding distribution loop (usage -> training data -> agent preference -> more usage) runs for platforms, not for PD.

Datadog built its own On-Call product rather than integrating PD -- the clearest possible signal that a platform sees the function as absorbable, not essential. ServiceNow targets incident management natively. Freshworks acquired FireHydrant rather than partnering with PD. When three adjacent platforms independently choose BUILD or ACQUIRE-ALTERNATIVE over INTEGRATE, the market has spoken on distribution destiny.

An agent asked to "handle this incident" routes to the platform already in the stack. PD's MCP endpoint is discoverable but not default. Native AI revenue is 2% of ARR (Q2 transcript) -- PD is not where agents live.

M -- Ecosystem Gravity: 2/5

Migration cost decomposition: redirect webhooks (hours), rebuild escalation policies (days), migrate on-call schedules (hours), retrain team (1-2 days), rebuild runbooks (days-weeks). Total: 1-3 weeks. Compare SAP (2-5 years), ServiceNow (6-18 months), Workday (6-18 months).

The patent signal is definitive. 31 patents in 17 years, and the 10-K explicitly states: "We do not believe that we are materially dependent on any one or more of our patents or other intellectual property rights" (line 787). The company tells you its IP doesn't matter. Thirty-one patents for a 17-year company, vs PANW at 1,000+ and NOW at 800+.

Network effects are zero -- each customer's incident data is independent. Top 10 customers represent just 2% of revenue, meaning a long tail of small customers with individually negligible switching cost. Institutional holders have already exited (ARK -97%, RGM Capital -100%).

F -- Friction: 2/5 (penalty dimension)

Low friction: self-serve onboarding in minutes, free tier for 5 users, 700+ pre-built integrations, no consultant dependency. This is good product design. It also means the barrier to leaving is as low as the barrier to arriving.

F carries weight -0.06 (penalty). PD's low friction reduces the penalty slightly (-0.12 vs -0.30 for F=5 companies like SAP). But this is immaterial to the verdict -- the difference between F=1 and F=2 is 0.06 points.

Thermodynamic Summary

PagerDuty is a textbook application of the Tool Death Theorem. The task domain D(PD) = {alert routing, escalation logic, on-call scheduling, event correlation, noise reduction, status page generation, postmortem creation} consists entirely of computable tasks. Local re-derivation cost c_l(tau, t) = g(tau)/M(t) approaches zero as model capability M(t) grows at mu = 1.1-2.3x/yr. Meanwhile, PagerDuty's cost floor c_s >= $252/user/year (pricing) remains fixed.

Four platform competitors have already demonstrated c_l < c_s by absorbing the function at zero marginal cost. The NRR trajectory -- five quarters of monotonic decline crossing below 100% -- is empirical proof that intelligence is flowing around PagerDuty in real time.

The dimensional uniformity (all 2s) reflects a coherent structural position, not coincidence. Narrow domain (C=2) prevents breadth expansion (U=2). No lock-in (E=2) prevents gravity accumulation (M=2). Easy adoption (F=2) means easy replacement. Platform absorption removes distribution (A=2). The reinforcing loops that protect Fortress companies (deep cognition -> infrastructure lock-in -> ecosystem gravity -> distribution default) are entirely absent.

Management's "central nervous system" narrative (Q3 transcript) collides with the numbers: ARR peaked at $499M and is now declining, revenue is guided flat-to-down for FY2027, and the 10-K added five new AI-specific risk factors not present in FY2025. The 10-K is a confession document.

Conviction Weight

kappa = (V - 3.0)+ = (1.72 - 3.0)+ = 0.00

w_PD = W_S * kappa_PD / sum(kappa_j) = 0

Basket weight: ZERO

V-Score conviction is regime-invariant. Whether the SaaS sector is up 30% or down 30%, PD's structural position does not change: the domain is computable, the infrastructure is replicable, the customers are leaving, and the platforms are absorbing the function.

The IR = -5.96 is context, not gate. It confirms what V = 1.72 already told us -- the market is discovering the structural position through price. But even in a raging bull market where IR turns positive (sector rotation lifts all SaaS), the V-Score does not change. Cyclical price recovery does not create irreducible infrastructure or compound cognition.

Basket Verdict

EXCLUDE. kappa = 0. No structural conviction. No weight in the basket.

The only non-zero optionality is PE takeout (Scalar Gauge: 28% probability), which operates on financial engineering timescale, not thermodynamic survival. Cash + investments ($548M) minus convertible notes ($402.5M) plus NOL shield (≈$83M) gives a floor of approximately $229M, well below the current $600M market cap. Without M&A, the operating business decays toward that floor as NRR continues its -2pts/quarter trajectory.

Evidence Table

Combined LR (structural): 0.3 x 0.4 x 0.3 x 0.2 x 0.5 x 0.4 x 0.3 x 0.4 x 0.5 = 0.000043 (bear signals dominate; partially correlated, effective LR ~ 0.01-0.05). No bull evidence is strong enough to shift the structural verdict.