Methodology
How Clarity values stocks — from SEC filing to fair value estimate.
Overview
Clarity runs six independent valuation models on live SEC filing data and triangulates them into a single fair value estimate. Rather than relying on any single model — each of which has known blind spots — we blend them using confidence-weighted Bayesian triangulation, then wrap the output in Monte Carlo simulation to produce confidence intervals.
The philosophy is simple: no model is right, but a well-weighted ensemble is less wrong than any individual approach. Earnings-anchored models like DCF provide a floor, growth-oriented models like EV/Revenue capture optionality, and comparables ground the estimate in market reality.
The Six Models
Bayesian Triangulation
Each model's estimate receives a confidence weight based on two factors: the model's general reliability for this type of company (detected via sector classification and financial profile), and the quality of the input data available.
Company profile detection uses a continuous intensity scale. For example, a high-growth SaaS company with negative earnings will see EV/Revenue receive a higher confidence multiplier, while DCF and residual income are still included — they provide a bearish floor that guards against valuing growth as if it never stalls. The profile intensity is hard-capped to prevent any single model from dominating.
The final blended fair value is the confidence-weighted average across all models that produced valid outputs. We keep all models in the blend rather than excluding them — triangulation over model selection is a core principle.
Monte Carlo Simulation
After triangulation produces a point estimate, we run 5,000+ Monte Carlo simulations to generate confidence intervals. The critical detail: Monte Carlo inputs use normalized free cash flow (post-SBC haircut, post-growth-capex conversion), not raw reported FCF. This ensures the confidence intervals are centered around and coherent with the blended fair value.
Each simulation randomly perturbs key inputs — growth rate, discount rate, terminal multiple, and margin assumptions — drawn from distributions calibrated to the company's historical volatility and sector norms. The output is a probability distribution of fair values, reported as P10, P25, median, P75, and P90 percentiles.
Data Sources
All financial data is pulled directly from primary sources, not scraped from aggregators:
Limitations
Clarity is a quantitative tool. It does not assess qualitative factors like management quality, competitive moat durability, regulatory risk, or macroeconomic regime changes. SEC filing parsing is inherently fragile — companies format their filings differently, and edge cases exist. Sector detection uses keyword matching with hard overrides for well-known tickers; misclassification is possible for unusual businesses.
The models assume mean-reverting economics and relatively stable capital structures. They are less reliable for pre-revenue companies, SPACs, companies undergoing restructuring, or those with non-standard accounting. Fair value estimates should be treated as one input in your investment process, not as buy/sell signals.