Research Methodology and Processing Workflow

This page describes how StrongBuyAnalytics retrieves third-party data, applies deterministic transformations, and presents research outputs. The platform contains feature-specific methods; it does not use one universal five-pillar model or one weighting scheme across every page.

By StrongBuyAnalytics Editorial Team · Publisher: StrongBuyAnalytics · Published September 15, 2026 · Updated October 8, 2026

1

Acquire and identify source data

The requested feature determines the provider path. Chart history may combine Polygon historical bars with recent Yahoo bars; intraday charts use Yahoo first and Polygon as fallback. Other implemented surfaces use Alpaca, SEC EDGAR, Financial Modeling Prep (FMP), CoinGecko, social/search feeds, news providers, or OpenAI. Provider availability, entitlement, symbol mapping, and asset type can change which fields are returned.

A retrieval timestamp is evidence that the application ran, not proof that every underlying observation was created at that moment. Filing period dates, market timestamps, and provider timestamps are the relevant freshness evidence where available.

2

Normalize and calculate

Application code standardizes symbols, dates, intervals, numeric fields, and table shapes before calculating ratios, trends, chart indicators, valuation scenarios, or feature-specific classifications. Calculations depend on available inputs. For example, a report DCF is unavailable without positive free cash flow and a valid share count, and the standard corporate DCF is intentionally skipped for financial-style companies.

No universal weighting claim. The stock-page score uses five fixed input adjustments, while the fundamental report sums deterministic section points and normalizes them. Charts, scanners, filing pages, and calculators have their own logic. See How Scores Work for the exact distinction.
3

Score, classify, and gate incomplete outputs

The fundamental report assigns points through defined section rules. Its overall rule score excludes the ten_year section and converts earned points over available maximum points to a 0–100 percentage. Separate confidence and freshness logic can suppress or qualify classifications. Statements no more than 120 days old are labeled fresh; 121–240 days stale; older statements very stale; unavailable dates are unknown.

Report labels describe fundamental quality and, when available, valuation attractiveness. They are research classifications—not Buy, Sell, or Hold recommendations. The simpler stock-page labels belong to a different calculation.

The five-input stock-page screen retains a reported finite zero and applies its own threshold: zero revenue growth is in the “0% or lower” band; zero gross margin is in the “35% or lower” band; zero net margin and zero return on equity are in the “0% or lower” band; and zero debt/equity is in the “below 0.30” band. Missing, invalid, and non-finite values are not scored as zero. A ratio with a zero denominator is undefined, not a measured zero. The separate report DCF requires positive free cash flow; a measured zero or negative TTM cash flow does not qualify for DCF.

4

Generate and present narrative

OpenAI is used on implemented features to turn supplied data and computed results into readable narrative. It is not the source of exchange prices or SEC filings, and deterministic score arithmetic does not depend on the model inventing points. If AI access is unavailable, some paths use deterministic text while others may omit narrative or report an error.

Generated prose can still misstate, omit, or overemphasize facts. Readers should inspect the displayed tables, dates, assumptions, and primary filings rather than treating narrative as an authoritative record.

Worked processing example

What this methodology does not establish

No performance calibration claim
Published scores are not represented here as calibrated probabilities or validated forecasts.
No predictive-accuracy claim
A high score does not imply a stated chance of gain, price target attainment, or market outperformance.
No track-record validation
A founder’s trading history does not validate these formulas or establish their future efficacy.
No complete-data guarantee
Provider delays, restatements, mappings, outages, caches, and missing fields can affect results.

References

Educational use only: This methodology organizes research inputs and does not provide individualized investment, financial, legal, or tax advice. Outputs may be wrong or stale; verify material facts with primary sources and qualified professionals.