Data Policy
Arsenal combines third-party financial data with browser-local selections and observations. This policy explains what is saved and how Daily Brief is assembled. See also the Privacy policy.
1. What your browser saves
My Arsenal saves asset symbols, followed economic and real-estate widgets, fixed category assignments, saved learning widgets and calculator, chart ranges and expanded or collapsed states. Daily Brief saves visit-start and last-seen timestamps, the latest successful comparison observations, technical ratings, risk and regime model outputs, and identifiers of disclosures already seen. Older saved watchlists are migrated automatically. We keep a bounded latest snapshot plus up to 600 recent chart observations per followed indicator, rather than an unlimited browsing history.
Local storage persists between visits; session storage keeps one comparison baseline while you move between pages. A site page load records a visit even when feeds fail. After 30 minutes away or a new tab, the next visit compares available observations with the prior saved snapshot. A first visit or missing snapshot is explicitly labeled: we do not manufacture a “since last visit” change.
2. Where data goes
Your widget selections, saved visit history and comparison records stay in this browser and are not sent to analytics or synced to an account. Looking up an item sends its symbol or series identifier to the relevant data provider, usually through an Arsenal proxy. Those requests carry normal network metadata. Clearing browser storage or using “Clear My Arsenal & visit history” removes the saved feature data; later visits begin saving again. There is no cross-device recovery.
3. Source dates and refresh timing
Market quotes and price histories come from Yahoo Finance. Economic indicators come primarily from FRED; earnings schedules from Finnhub; disclosures from the site’s congressional filing feed; headlines from the site’s news aggregation feed. Market observations may be delayed. Economic observation dates refer to the measured period, not necessarily the publication date; historical values may be revised.
Active briefing and watchlist pages check for updates approximately every five minutes and pause their feature requests while hidden. Research inputs and technical scans are cached for approximately 30 minutes. Provider and proxy caches can add delay. Sources that fail remain unavailable; their absence is not interpreted as “unchanged.” Last successful comparison observations are preserved.
Cap-rate surveys, the published FOMC projection reference, private-capital references and curated delinquency data retain their stated publication periods. They are not intraday feeds. Economic calendar dates may be recurring estimates and are marked accordingly. Earnings schedules can change; being on the calendar does not prove results have been published. Congressional disclosures may be filed weeks after a transaction.
4. How Daily Brief is written
The briefing is assembled from deterministic templates and available observations, with a main read capped at 200 words. It does not invent analysis when a feed is missing. The main briefing is the same for all readers. It considers the US Economy and US Real Estate indicator catalog, ranks supported changes against measure-specific thresholds, and omits small moves and sources without comparable prior observations. A separate IPO calendar lists expected listings over the coming seven days; dates and pricing may change. Prices and Technical Monitor ratings compare against the prior trading session; macro indicators compare consecutive source observations; Crash Risk compares with a reconstructed prior monthly model snapshot; regime compares classifications across monthly macro observations. These historical reconstructions use currently available data and can include revisions. Earnings appear when reported EPS differs from the available estimate. Personal watchlists do not select or rank the main briefing. The edition changes automatically with Eastern time; a separate “Since your last visit” view shows saved comparisons.
The crash-risk score uses the same model as the Crash Risk page. It is a model score, not a calibrated forecast probability. Regime classification follows the GDP/inflation model used by the regime page. The public technical scan uses the complete, fixed Technical Monitor universe and its calculations. Successful coverage and comparison dates are disclosed; failed symbols are excluded rather than counted as unchanged. Server snapshots cache for approximately 15 minutes; preview fallback scans use the same fixed universe. Price changes exclude dividends and other distributions.
Headlines are filtered for recency, duplicates and promotional language, then prioritized by broad-market subjects such as central banks, inflation, employment, policy and geopolitical developments. Publisher links provide the source; selection is not a guarantee that a story is complete or the day’s most important event. A disclosure appearing for the first time means newly observed by this browser, not necessarily newly traded.
5. Controls and limitations
You can add and remove items within seven fixed categories: Stocks, ETFs, US Economy, US Real Estate, Crypto, Commodities and Learning. Historical charts offer 1-, 5- and 10-year views where data exists; younger assets and shorter series only show available observations. Learning offers compact Dictionary, Asset Returns and CAGR calculator widgets. Calculator inputs and dictionary queries are held in the open page; only the widget selection and display preferences are saved. Asset name searches send the search text to the quote provider for suggestions. The feature supports up to 60 saved widgets. Removing a widget stops following it; use the clear-data control to remove all saved comparison records. Private browsing, storage restrictions or clearing site data may reset your selections and history.
All data and research are informational. Confirm consequential figures at the original source. The Terms describe the site’s general limitations.