How Missionomics works

1. Facts vs. scores — the distinction that matters

• Facts — specific data points drawn from public records and disclosures, each attributed to its source (for example, a company's OSHA penalty total from the Department of Labor, or its CEO-to-worker pay ratio from its own SEC proxy statement). We report these as the source states them.
• Scores — composite numbers (like a "Climate Action" score of 62/100) that Missionomics computes from those facts using the method described in Section 3. Scores are our own editorial composites: opinions derived from underlying data under a disclosed methodology, not independent factual certifications of any company.

Crucially, the platform is worldview-agnostic. We don't declare companies "good" or "bad." You set the sliders and tags that reflect your priorities — climate, faith-based screens, domestic manufacturing, governance, anything — and the same underlying data is weighed according to your values, not ours.

2. Where the data comes from

We currently track 41 registered data dimensions across roughly 1,500 U.S.-listed companies. Every dimension carries a source grade so you always know how solid the ground is:

The dimensions by category, with their primary sources:

Market data (prices, market cap, P/E, dividend yield, beta) is sourced from a licensed commercial market data provider, with prices refreshed daily after U.S. market close.

3. How theme scores are computed

• Point-based composites. Each theme starts from documented signals with fixed point values — for example, an SBTi-validated emissions target adds points to Climate Action, while a large verified emissions footprint or a walked-back climate pledge subtracts them. Enforcement outcomes (OSHA penalties, NLRB findings) carry more weight than self-reported initiatives, which are deliberately capped because nearly every large company claims some.
• AI-assisted extraction. Many inputs (workforce diversity percentages, human-rights policies, geographic revenue splits) are extracted from companies' 10-K filings and public reports using AI (Anthropic's Claude models), then structured into a facts database. Extraction can make mistakes; that's one reason the error-reporting mechanism in Section 6 exists.
• Cross-sectional normalization. Raw composite points are graded on a curve across the whole universe: normalized to a mean of 55 and standard deviation of 16, clamped to the 3–98 range. A score of 70 therefore means "well above the average company we cover on this theme," not "70% good" in any absolute sense.
• Non-disclosure imputation at the 25th percentile. When a company chooses not to disclose a material number its large-cap peers routinely report (workforce diversity, women in leadership, board independence), we don't treat silence as zero and we don't treat it as average. We impute the 25th percentile of disclosed values — the company scores like the bottom quartile of companies that did disclose, and never worse than a documented bad actor. Every imputed value is flagged in the score breakdown you see in the app.
• Confidence labels. Each theme score carries a confidence badge (high / medium / low) based on how many hard, externally-verifiable data points back it, versus estimates and self-reported claims. Low-confidence scores are visibly flagged in the interface.
• Controversy adjustments. Documented controversies (regulatory enforcement, litigation, credible investigative reporting) apply severity-weighted score penalties that decay over time, and the specific items driving a penalty are shown in the score breakdown.
• Exposure scores are simpler. Industry-exposure scores (e.g. "tobacco-free") reflect estimated revenue exposure to that industry — 0% exposure scores 100, and the score falls as exposure rises.

To state it plainly: theme scores are Missionomics' own editorial composites. They are our opinion of how the underlying documented facts add up under the method above. They are not certifications, ratings endorsed by any regulator, or statements that a company has or hasn't done any particular thing beyond the sourced facts themselves.

4. How non-disclosure is handled

• When a company doesn't report a data point, it is recorded as "unreported" — never silently treated as zero.
• For material numbers that peers routinely disclose, non-disclosure is scored conservatively via the 25th-percentile imputation described above, and clearly flagged as imputed in the app.
• Screens and filters are coverage-gated: a filter only appears in the interface if the underlying dimension has enough real data to be meaningful. We'd rather show you fewer controls than dead ones backed by empty data.
• Each company's transparency score reports the share of applicable dimensions it actually discloses, so disclosure behavior is itself visible data.

5. Our use of AI

6. Limitations, corrections, and reporting errors

• Data lags reality. Filings are annual or quarterly, enforcement databases update on their own schedules, and third-party indexes have publication cycles. A score reflects the data available when it was last computed, not necessarily today's facts.
• Coverage is uneven. Some dimensions are well-populated (SEC-mandated disclosures); others are sparse (voluntary disclosures). Sparse dimensions are coverage-gated out of the interface rather than presented as complete.
• Errors can occur — in source data, in AI extraction, or in our own pipeline. We review every error report and correct verified errors promptly. To report one, open any company's detail view and use the "Report a data error" link at the bottom, or reach us through the contact form.
• When a verified correction changes a company's underlying facts, its scores are recomputed under the same disclosed method — corrections change inputs, never the rules.

7. Not investment advice

Missionomics is an informational and educational research tool. Nothing on the platform — facts, scores, model portfolios, or algorithmic allocations — is personalized investment advice or a recommendation to buy or sell any security. Portfolios on Missionomics are simulated. See our Terms of Service for the full disclaimers.