Methodology & Scientific Basis

Potentia's scoring engine is not a set of ad hoc heuristics — each sub-score is a direct, documented application of a concept from a specific piece of published theory. The primary theoretical source is Structural Semantics: An Engineering Framework for Measuring Meaning (internal Potentia research monograph, first edition, 2026), which develops a formal quantitative apparatus — drawing on information theory, cybernetics, statistics, and network science — for treating narrative, influence, and organizational structure as measurable fields rather than impressions. That work's own bibliography (21 sources across four disciplines) is the deeper foundation underneath it, and is reproduced in full below.

This page exists so a report reader — or a founder deciding whether to consent to being scored — can see exactly where a number comes from, not just trust that it's "AI-generated."

How each sub-score is grounded

Semantic Entropy Index (SEI) — thematic focus vs. scatter in a founder's own writing.

Computed as Shannon entropy H(N) = −Σ p·log₂(p) over the founder's text clustered into themes — the exact formulation from Shannon (1948) and Cover & Thomas (2006). Because entropy estimated from a small, clustered sample is biased upward, the clustering and normalization approach follows the correction logic in Miller (1955) and Grassberger (1988), both of which deal specifically with finite-sample entropy bias.

Narrative Consistency Index (NCI) & Founder-Market Fit (FMF) — coherence over time, and overlap between pitch and demonstrated work.

Both reduce text to vectors and compare them by cosine similarity — a measurement-under-noise problem. The vector-space approach and its error bounds follow Eckart & Young (1936) on low-rank approximation, Trefethen & Bau (1997) on numerically stable linear algebra, and Carroll et al. (2006) on measurement error in nonlinear models — since a proxy signal from public text is, formally, a noisy measurement of an unobserved true state.

Influence Authenticity Score (IAS), Track Record Retrospective (TRR) & Signal Efficiency Index (SGI) — bounding noisy proxy ratios into an honest 0–100 scale.

Raw ratios (followers/following, stars/repo, hype words per 1,000 words) don't map linearly to a meaningful score — they saturate. The bounded, logistic-shaped normalization used here follows the same family of link functions formalized in Berkson (1944), Cox (1958), and McCullagh & Nelder's Generalized Linear Models (1989), with estimator behavior checked against Lehmann & Casella's Theory of Point Estimation (1998).

Network Environment Index (NEI) — breadth of a founder's professional network, from their own data only.

The underlying idea — that a network's structure (not any one person in it) predicts diffusion and influence outcomes — comes from network and social-influence theory: Granovetter (1978) on threshold models of collective behavior, Friedkin & Johnsen (2011) on social influence networks, Coleman (1990) on the foundations of social capital, and Boyd & Richerson (1985) on cultural transmission. The graph-theoretic machinery for treating a network as a measurable field follows Chung (1997) on spectral graph theory and Aubin (1998) on the differential-geometry results used for field/flow arguments. Critically, none of these require — or justify — reading any individual colleague's account: they describe network-level properties recoverable from aggregate signals the founder's own accounts already expose (org counts, follow counts, mention counts), which is exactly how NEI is computed. See Data Protection for why no other person's account is ever accessed.

Why RVI, DSH, and ODS are marked unavailable in real-mode reports

Team Requisite Variety Index is named directly after Ashby's Law of Requisite Variety (Ashby, 1956, 1958) — a team's ability to absorb disturbance scales with the variety of backgrounds it contains, which cannot be measured from one founder's personal accounts alone. Decision Structure Health draws on Conant & Ashby (1970), "every good regulator of a system must be a model of that system": it needs a member × topic engagement matrix across the whole team, which is internal data no personal account exposes. Rather than approximate these from data that can't support the theory behind them, Potentia marks them unavailable and renormalizes the composite score over only the sub-scores actually computed — the same honesty principle Wiener's Cybernetics (1948) applies to feedback systems: a control loop that reports confidence it doesn't have is worse than one that reports a gap.

Full bibliography

Reproduced from Structural Semantics: An Engineering Framework for Measuring Meaning, §6.5 References.

Cybernetics and information theory

• Ashby, W. R. (1956). An Introduction to Cybernetics. Chapman & Hall.
• Ashby, W. R. (1958). Requisite Variety and its implications for the control of complex systems. Cybernetica, 1(2), 83–99.
• Conant, R. C., and Ashby, W. R. (1970). Every good regulator of a system must be a model of that system. International Journal of Systems Science, 1(2), 89–97.
• Cover, T. M., and Thomas, J. A. (2006). Elements of Information Theory (2nd ed.). Wiley.
• Shannon, C. E. (1948). A mathematical theory of communication. Bell System Technical Journal, 27, 379–423 and 623–656.
• Wiener, N. (1948). Cybernetics, or Control and Communication in the Animal and the Machine. MIT Press.

Statistics and decision analysis

• Berkson, J. (1944). Application of the logistic function to bio-assay. Journal of the American Statistical Association, 39(227), 357–365.
• Carroll, R. J., Ruppert, D., Stefanski, L. A., and Crainiceanu, C. M. (2006). Measurement Error in Nonlinear Models (2nd ed.). Chapman & Hall.
• Cox, D. R. (1958). The regression analysis of binary sequences. Journal of the Royal Statistical Society B, 20(2), 215–242.
• Eckart, C., and Young, G. (1936). The approximation of one matrix by another of lower rank. Psychometrika, 1(3), 211–218.
• Lehmann, E. L., and Casella, G. (1998). Theory of Point Estimation (2nd ed.). Springer.
• McCullagh, P., and Nelder, J. A. (1989). Generalized Linear Models (2nd ed.). Chapman & Hall.
• Miller, G. A. (1955). Note on the bias of information estimates. In H. Quastler (ed.), Information Theory in Psychology, 95–100. Free Press.
• Trefethen, L. N., and Bau, D. (1997). Numerical Linear Algebra. SIAM.

Differential geometry and PDE

• Aubin, T. (1998). Some Nonlinear Problems in Riemannian Geometry. Springer.
• Chung, F. R. K. (1997). Spectral Graph Theory. American Mathematical Society.
• Grassberger, P. (1988). Finite sample corrections to entropy and dimension estimates. Physics Letters A, 128(6–7), 369–373.

Adjacent literature (social & network theory)

• Boyd, R., and Richerson, P. J. (1985). Culture and the Evolutionary Process. University of Chicago Press.
• Coleman, J. S. (1990). Foundations of Social Theory. Harvard University Press.
• Friedkin, N. E., and Johnsen, E. C. (2011). Social Influence Network Theory. Cambridge University Press.
• Granovetter, M. (1978). Threshold models of collective behavior. American Journal of Sociology, 83(6), 1420–1443.

None of the above is presented as a peer-reviewed validation of Potentia's specific formulas — it is the published theoretical basis those formulas were adapted from. See Disclaimer for how to weigh a report.