Data sources
Last updated September 30, 2026
Every number on Angie comes from somewhere. This page names each source, what we use it for, and the terms it comes with. Angie’s valuations are models built on these figures, not advice. See the Terms for how you may use them.
1 · U.S. Securities and Exchange Commission, EDGAR
What it provides. Company financial statements: revenue, cash flow, capital spending, cash, debt, equity, interest, dividends, and the history behind every valuation.
Terms. Public filings made by the companies themselves. Angie reads the SEC's structured data under the SEC's fair-access policy and does not alter the filed figures; it selects and combines them, as each valuation page explains. The figures are the companies' own statements, not Angie's, and are not independently verified.
Freshness. As filed. A company's newest annual report can be many months old, and a valuation says so when it is.
2 · Yahoo Finance
What it provides. Figures the SEC's structured data does not carry for a particular company, such as a finance arm's loan book filed under the company's own tags. Every valuation that uses one says so on the page.
Terms. Third-party data, used for information and not licensed to you. Do not republish it.
Freshness. Delayed.
3 · Yahoo Finance
What it provides. Share prices, market capitalisation, shares outstanding, per-share earnings, company names, and sector and industry labels. Forward earnings and the forward P/E are shown for reference only and never enter a valuation.
Terms. Third-party data, used for information and not licensed to you. Do not republish it.
Freshness. Delayed, typically by 15 minutes or more.
4 · Alpaca (IEX exchange feed), with Yahoo Finance as a fallback
What it provides. Intraday quotes and chart bars on the paper desk and charts.
Terms. Third-party data for information only. IEX is a single exchange, not the consolidated market, so quotes can differ from other venues. Not for trading or execution.
Freshness. Real time where shown as live; otherwise delayed.
5 · Aswath Damodaran, NYU Stern School of Business
What it provides. Country risk premiums, added to the discount rate of a company based outside the United States.
Terms. Published by its author for public use. Credited here. Angie uses the premium over the US figure and does not alter the table.
Freshness. Updated each January. Angie uses the January 2026 table.
pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/ctryprem.html
6 · Federal Reserve Bank of St. Louis (FRED) and Yahoo Finance
What it provides. The risk-free rate and economic series on the Conditions page.
Terms. Public economic data, used with attribution.
Freshness. As published; most series are daily, monthly or quarterly.
7 · Publishers' own public news feeds
What it provides. Headlines in the feed.
Terms. A headline, a short summary and a link back to the publisher. Articles are not reproduced, and the copyright in each stays with its publisher.
Freshness. As the publisher posts.
8 · Methodology history
Every change to how Angie values a company, dated, with what it did to the board when it was measured. A value that moved on one of these dates moved because the method changed, not the market. The daily record of prices against fundamentals is on Market mood.
- 2026-10-20 · Data. Depreciation and amortisation is read from CostOfGoodsAndServicesSoldDepreciationAndAmortization when a filer tags it nowhere else (last alternative), so owner earnings exist for the maintenance-capex split. Measured: PACS owner earnings now 486m (was missing); its value does not move because the growth-capex add-back is off above 30% growth (PACS 32%), a method question for review (#202). Controls AAPL MSFT KO WMT HD CVX unchanged.
- 2026-10-19 · Data. The nightly sweep stamps the board with the last US session whose close has passed, not the ET calendar date it runs on. GitHub starts the 02:00 UTC cron near 02:30 ET, so Friday's close was labelled Saturday and every board date ran one day ahead. Measured: Labels only; no value changes. The 2026-10-17 board holds the 2026-10-16 close; the next run writes 2026-10-19 for Monday's close, with no collision.
- 2026-10-19 · Publication. London, 2026-10-19: "no verdict on floor value companies. transparently say floor value." Every floor (the Gordon-spread floor in the equity models, the margin-capped and building-cycle FCF paths) publishes verdict NO_VALUE with the label Floor value and a worth-at-least figure in the new floor_value column. Before, a floor more than 10% above the price read UNDERVALUED. Measured: First board after release (2026-10-19): 283 rows carry a floor value and no verdict. On 2026-10-17, 73 of them had a verdict (49 UNDERVALUED, 10 FAIR, 14 OVERVALUED); 210 were already NO_VALUE.
- 2026-10-18 · Financials. London, 2026-10-18: a company that beat its cost of equity in at least 12 of its last 15 filed years keeps three quarters of its median 15-year spread after the fade (cap 20% ROE unchanged); the long tier only adds to the 180b edge. Measured: Holds for JPM and AXP (15 of 15), not MS (7), GS (11), BAC (7). WFC 0.85x to 0.94x, AFL 1.57x to 1.58x; AXP unchanged at the 20% ROE cap.
- 2026-10-18 · Data. Latest twelve months: on the SEC path the newest statement column is the trailing twelve months to the newest 10-Q (flows: fiscal year + year-to-date - prior year-to-date; balances: newest quarter end), replacing the newest fiscal-year column. Both engines. 20-F filers unchanged. Measured: 27 names, one run: 23 moved to 2026 quarter-end statements; median value/price 0.659x to 0.757x. MS 0.53x to 0.95x, GS 0.59x to 0.91x, WFC 0.83x to 0.85x, AAPL 0.33x to 0.42x, LLY 0.25x to 0.38x, KO 0.35x to 0.56x, LUV 0.91x to 1.29x, XOM 0.37x to 0.29x. Verdicts: MS, GS, WFC OVERVALUED to FAIR; ALL, LUV FAIR to UNDERVALUED.
- 2026-10-17 · FCF. Equity value is scaled to the parent's share of consolidated earnings (median of the last three years of net income to common over net income including noncontrolling interests, 0.3 to 1.0). Consolidated cash flow and net debt include minority owners' part of subsidiaries. Skipped when the share count spans several classes. Measured: TLK 4.0x to 3.07x (now under the 4x hold: UNDERVALUED); LNG 0.67x to 0.53x (owns about half of Cheniere Partners); MPC 0.32x to 0.22x (owns about two thirds of MPLX); SHAK 0.55x to 0.52x; AAPL unchanged.
- 2026-10-17 · Financials. Asset managers add back after-tax amortisation of acquired intangibles (AmortizationOfIntangibleAssets), capped at half of earnings. Measured: BEN (protected) 0.475x to 0.759x, AMP 0.944x to 1.136x; BLK unchanged (no annual amortisation tag since its 2024 reorganisation).
- 2026-10-17 · Financials. Loss years before a company turned profitable and stayed so for 2 years leave its ROE window. Data fixes with it: common equity falls back to total equity less noncontrolling interest when no parent-equity tag is filed (Robinhood 2025), and one-off adjustment rows (sale gains, impairments) keep their history when the tag goes dark (Welltower). Measured: HOOD 0.029x to 0.143x (2025 restored; start ROE -11% to about 20%). WELL own growth -7.3% to +7.5%, value unchanged at 0.242x: the REIT model funds growth only from retained cash, Welltower grows by issuing shares.
- 2026-10-17 · Financials. Median-ROE routes (brokers, insurers, life) start halfway between the latest year and the 5-year median, the rule the earnings model already uses. Measured: MS 0.433x to 0.451x, GS 0.485x to 0.505x, ALL 0.584x to 0.862x, AFL 0.804x to 0.711x.
- 2026-10-17 · Financials. London, 2026-10-17: a company that beat its cost of equity in 4 of its last 5 years keeps half its median past spread after the 10-year fade, terminal ROE at most 20%, in every residual-income route. Replaces the flat 2-point bank edge. Measured: AXP 0.349x to 0.515x, MS 0.451x to 0.527x, GS 0.505x to 0.593x, AFL 0.711x to 1.481x, ACGL (protected) 1.049x to 1.969x, AIZ 0.595x to 0.952x; banks with narrow spreads fall slightly (BAC 0.881x to 0.868x).
- 2026-10-17 · Financials. Durable financials (net income to common positive in 8 of the last 10 filed years) take the long-run risk-free blend of fix 156a, residual-income and earnings routes only; the blend never moves a company onto the Gordon floor it was not on. Measured: MS 0.415x to 0.433x, GS 0.466x to 0.485x, AXP 0.330x to 0.349x, C 0.804x to 0.840x; BLK unchanged (4 filed years since its 2024 reorganisation).
- 2026-10-17 · Data. SEC statements more than 15 months old give way to the market-data provider's newer annual year when it carries the lines the valuation needs. SEC companyfacts holds no 2025 IFRS figures for several 20-F filers although the reports are filed. Also: IFRS revenue read from goods-and-services and telecommunication-services tags when no total tag is filed. Measured: TLK: revenue now read (USD 8.3bn), growth 10% (FCF history) to 3% (revenue), 5.6x to 4.0x, still held. TLK and CCU stay on 2024 SEC figures: the provider has no operating cash flow line either. BBAR moves to the provider's 2025 year, 0.48x to 0.55x.
- 2026-10-17 · WACC. Country risk premium: the HQ country's premium over the US (Damodaran, January 2026 table) joins the equity risk premium, so cost of equity = risk-free + beta x (ERP + premium). Tax domiciles (Bermuda, Cayman, Ireland, Jersey, Netherlands, Luxembourg) take 0; Puerto Rico counts as the US. Measured: Sample of 82 non-US companies (2026-10-17, one run): premium applied to 62; median value/price 1.03x to 0.99x, median value -4.9%; no verdict changed in the sample. TEO 2.5x UNDERVALUED to 1.02x FAIR (cost of equity 9.1% to 17.5%); GGAL 0.59x to 0.39x; CX 0.62x to 0.56x; Israel median 1.19x to 0.94x. US companies unchanged.
- 2026-10-14 · Financials. Mortgage REITs split: securities books stay at marked book; credit books (net loans over equity) are valued at book x ROE / cost of equity, capped at book, on the lower of median and latest ROE. Lender loan tags added to the EDGAR loan line. Both released. Measured: All 27: securities 13, median 1.12x price; credit 14, median 0.99x (was book, up to 7.7x); RC, ACRE, BRSP refused.
- 2026-10-13 · WACC. Durable and high-performing companies use a 50/50 blend of today's 10-year Treasury and its 10-year average as the risk-free rate (5.22% and 2.91% gave 4.06%). Longevity: positive free cash flow in at least 8 of the last 10 filed years. Performer: revenue growth of 15% a year or more with positive free cash flow. All others keep today's rate. Measured: S&P 500 A/B, 503 names: 342 qualify (327 longevity, 15 performer); median value of qualifying names +15.2% (quartiles +5.1% and +20.0%); board median value to price 0.820x to 0.865x; within 20% of the analyst target 31 to 34 of 150; undervalued 138 to 162.
- 2026-10-13 · Board. A board verdict changes only when the price moves 5 points past the band edge. Names on an edge no longer flip on small price moves. The engine's own call is kept in the result as verdict_raw. Measured: Replay of 68 stored nights, 3,376 companies: verdict changes 3,343 to 2,062 (38% fewer); median nights a verdict lasts 4 to 7.
- 2026-10-11 · FCF. A measured hypergrowth company holds its growth 3 years before the fade. Measured: With 155a and 155b: PLTR $19.2 to $31.6, NVDA $192 to $362; whole set 0.79x to 0.84x.
- 2026-10-11 · All engines. Beta cap 1.4 instead of 1.6. Measured: Capped-beta names median 0.40x to 0.44x alone.
- 2026-10-11 · All engines. The equity risk premium is a setting (app_config.equity_risk_premium), 4.5% instead of a fixed 5%. Measured: Whole set median 0.79x to 0.82x alone.
- 2026-10-11 · FCF. No growth-capex add-back for mature names with net debt above 4x owner earnings; revenue growth leaves out acquisition years. Measured: 18 of 80 floored names changed; CHTR 9.55x to 5.05x.
- 2026-10-11 · FCF. A value above 4x the price is held for review, not published. The low tail stays published (2026-09-06 ruling). Measured: 199 board names above 4x on 2026-10-10.
- 2026-10-11 · FCF (measured, not adopted). Blume-adjusted beta. Measured and left off. Measured: Moved no extreme value; pushed the sample median 0.72x to 0.69x.
- 2026-10-11 · FCF. The WACC floor follows the risk-free rate (risk-free + 1.8 points, never under 6%). Measured: Floored names median 2.60x to 2.51x price.
- 2026-10-11 · FCF. Stock pay: a growing diluted share count spreads value over future shares (stock pay at least 5% of revenue), charged for 5 years at most. Measured: PLTR 27.62 to 19.2 after the 5-year cap; names under the gate unchanged.
- 2026-10-11 · Utilities. Acquisition years out of rate-base growth; share-basis mismatch refuses; payout above 1.5x EPS noted. Measured: 55 valued, median 0.91x.
- 2026-10-11 · Alternative managers. Published as a labelled floor on the declared dividend. Measured: BX $52.71, TPG $20.63, CG $19.32, OWL $9.59; ARES, HLNE, BAM refuse.
- 2026-10-10 · P&C insurers. Income restated to the median loss ratio from SEC premiums and losses. Measured: Reaches 35 of 49; median 0.94x to 0.93x; rank check 0.81 to 0.64.
- 2026-10-09 · REITs and earnings model. The payout follows the funding identity every year, terminal included. Measured: REIT median 0.78x to 0.94x price.
- 2026-10-08 · Financials. Equity and income paired from the latest year, then (2026-10-10) by period date. Measured: Removed inflated returns in BDC, P&C, life and capital markets classes.
- 2026-10-08 · BDCs. BDCs valued on a constant total return on NAV (NAV x r / ke, no growth). Measured: Class median 1.27x to 1.02x price; released.
- 2026-10-06 · Banks. Half of unrealised held-to-maturity losses, after tax, charged against book. Measured: Citigroup -$2.43 a share, BAC -$4.53; explains about $2 of Citi's $12 gap to its blind range.
- 2026-10-05 · FCF. Finance-lease obligations count as debt unless the debt line already includes them. Measured: Replay, 903 names.
- 2026-10-05 · FCF. The verdict band is symmetric in upside: +10% undervalued, -10% overvalued. Measured: Replay, 903 names.
- 2026-10-05 · FCF. WACC weighted on gross debt, so a cash-balanced company keeps a debt weight. Measured: Replay, 903 names.
- 2026-10-05 · FCF. Cash includes short-term investments whenever they are filed. Measured: Replay, 903 names.
- 2026-10-05 · FCF. Leverage charged once: free cash flow to the firm discounted at WACC, then net debt subtracted. Measured: Median upside -27.4% to -11.6% across the replay; the largest single fix.
- 2026-10-05 · FCF. The terminal cap scales with how far the measured rate fell below the floor instead of dropping from 25x to 18x at one point. Measured: Part of the six-defect replay over 903 names (docs/FCF-DEFECTS-REPLAY-2026-10-05.md).
9 · What Angie does not do
Angie does not place real trades or connect to a brokerage account. The paper desk is a simulation with pretend money. Nothing on Angie is an offer, a recommendation or investment advice.
