T17

This is a meta-epistemic claim about the design of the GTM World Model itself. It is contested because there is a genuine tradeoff between completeness and precision: a maximally comprehensive model includes constructs that are unmeasurable, making it harder to use as an operational guide. The minimalist alternative (only include what is empirically calibrated) is cleaner but excludes hypotheses that practitioners need to reason about. The 'honest legend' prescription is also contested: in practice, tagging every construct with its epistemic status creates cognitive overhead that reduces adoption. The claim is that this overhead is worth paying.

The v3.0 design choice was to include every substantive structural insight (from 12 adversarial critiques) as first-class model content, even when the underlying mechanism was 'proposed, untested.' The safeguard against this producing an unfalsifiable narrative engine is the epistemic tagging system: every construct is labeled with its epistemic_type (identity / causal_regime / correlational / latent_multiplier / estimator / conditional_coefficient) and its measurement_status (measured_today / proxy_available / indirect_inference / unmeasurable_at_present). A model without these tags can rationalize any observation post-hoc by selectively invoking whatever construct fits. A model with these tags forces the user to confront, at every inference, whether the link they are using is a hard identity or a contested hypothesis. The practical function of the tagging system is to preserve the model's falsifiability despite its breadth — to distinguish regions of the map that are 'surveyed and walkable' from regions that are 'sketched and speculative.'

• Tetlock's Superforecasting research: well-calibrated forecasters who know the confidence level of each component of their reasoning outperform confident-but-uncalibrated forecasters by 2-3x on Brier score
• Scientific replication crisis: fields that did not explicitly tag theoretical vs. empirical vs. speculative claims experienced systematic over-confidence and failed replication at 50%+ rates
• GTM-specific case: companies that treated LTV = ARPA*margin/churn as an 'identity' (rather than an estimator under exponential churn assumptions) systematically over-valued customer cohorts and made incorrect retention investment decisions

• Comprehensive tagged models are harder to communicate to non-technical executives than simpler unlabeled heuristics — the honest complexity may reduce organizational adoption
• The epistemic tagging itself requires judgment about what is 'established' vs. 'directional' vs. 'contested' — these meta-level judgments are themselves uncertain
• Some practitioner frameworks (e.g., Jobs-to-Be-Done, Crossing the Chasm) achieve wide adoption precisely because they are simple and unqualified, even when technically imprecise

When building your GTM measurement framework, explicitly classify each metric and model by epistemic status. Distinguish: (a) accounting identities you will use as hard constraints (the MRR walk); (b) estimators you will use with stated assumptions (LTV under your chosen retention model); (c) causal hypotheses you are testing with experiments; (d) correlational signals you are using for pattern detection. This classification prevents the common error of treating a correlated metric as a causal lever. Present this classification to your board: a model that tells you where it knows versus where it guesses is more useful than one that presents all outputs with equal confidence.