T0

Most GTM practitioners treat conversion rates, win rates, and pipeline velocity as properties of their sales team — as if coaching harder would move them proportionally. The claim that these coefficients are CONFOUNDED by an unobserved latent variable (PMF, Phi) and a macro regime scalar (Psi) is uncomfortable because it limits how much credit any individual rep or campaign can claim. The 'correlational membrane' framing is also contested because it implies that standard A/B testing on funnel steps is structurally misleading: you can't hold Phi constant across test cells if Phi is segment-level. Marketing attribution vendors, who profit from treating every link as causal, have institutional incentives to resist this framing. Finally, the word 'identity' is strong: critics correctly note that the MRR walk is the only true accounting identity; everything else in GTM is an estimator built on assumptions.

The GTM system has three epistemic layers. The innermost is an accounting identity: MRR(t+1) = MRR(t) + New + Expansion - Contraction - Churn. This is true by definition. The middle layer is a set of estimators — LTV, CAC payback, Magic Number, Rule of 40 — that summarize the identity under explicit assumptions about segmentation, horizon, and retention model. These are NOT identities; they are regime-local. The outer layer is the behavioral membrane: conversion coefficients r_i that govern how leads flow through funnel stages. These coefficients LOOK like fixed properties of a firm's GTM machine, but they are in fact mixtures over an unobserved buyer-state B and modulated by Phi (product-market fit, a latent multiplier) and Psi (macro regime: growth-priced vs. FCF-priced). The practical implication: R = Phi * f(acquisition, retention, expansion) * Psi_weight is the structural form, where Phi enters multiplicatively in low-switching-cost segments and additively (R = Phi*f1 + S*f2) where switching costs are high. Treating the membrane as causal when it is conditional on Phi and Psi is the root cause of most GTM over-attribution errors.

• Fader-Hardie sBG/BdW customer-base audit: firms using homogeneous LTV = ARPA * margin / churn overstate LTV 2-3x versus the heterogeneous survival model, confirming that estimator assumptions matter enormously
• Bessemer Venture Partners Rule-of-X (growth rate + FCF margin) changed its recommended weighting from equal to 2x growth in ZIRP (2012-2021) and reverted toward FCF post-rate-rise, confirming that Psi re-weights the objective function measurably
• 6sense 2023 B2B buyer research: 67% of the purchase decision is made before a buyer contacts a vendor, implying that funnel conversion coefficients are conditioned on a pre-contact buyer state that most CRMs do not observe
• LinkedIn B2B Institute 95-5 rule: only ~5% of any B2B category is in-market at any given time, structurally capping what any funnel intervention can move
• Companies with high PMF (NPS > 40, organic growth > 30% of new logo ARR) show 2-4x higher funnel conversion at equivalent CAC spend, consistent with a multiplicative Phi term

• The exact functional form of Phi's interaction with conversion coefficients is unobserved and unfitted — the model specifies the structure but not the magnitude
• Psi is not directly observable; it must be proxied by interest-rate spreads, public SaaS multiple compression, or VC funding velocity, all of which are lagged and noisy
• Some high-cadence funnel experiments (e.g., Outreach's send-time optimization) do show causal lift that is not fully explained by buyer-state confounders, suggesting the membrane has genuine causal content too

Stop attributing funnel conversion changes to rep behavior or campaign execution alone. Before declaring a GTM intervention successful, check whether Phi changed (did product improve, did ICP tighten, did organic word-of-mouth accelerate?) and whether Psi shifted (did capital costs move, did category sentiment change?). Practically: instrument your pipeline to segment by buyer-state signal (intent data tier, prior engagement score, shortlist position proxy) before measuring conversion lift. Accept that roughly 25-40% of period-over-period GTM variance is irreducible noise under any model — this is the refusal zone. The identity layer (MRR walk) is your reliable anchor; build dashboards around it, not around estimators presented as laws.