Where the numbers come from.

MFA calculators are deterministic models: the same inputs should produce the same output. Federal annual limits and tax parameters are tied to the active 2026 rule set. When an agency publishes a future-year value early, MFA may show it as a preview, but future-year values do not enter calculator math until their effective year and the complete activation checks pass. Agency projections are not substituted for official limits. State-tax and location models are separately versioned because a nationwide state dataset can lag the federal annual update. MFA advances the state-tax year only when the replacement 50-state + D.C. table validates as a whole; retirement-income and local-tax overlays remain separately year-stamped because they can change independently. Regional Price Parities (RPP) are versioned by BEA observation vintage, not by the current calendar year: MFA keeps the actual state + metro vintage visible, checks BEA for the current official release, and only advances after a newer complete dataset passes location QA. Results can omit credits, deductions, local rules or special circumstances; those limitations should be shown on the relevant tool.

The cadence labels below are freshness targets, not a promise that a background cron runs continuously when nobody is using MFA. Supported live datasets are checked when the relevant MFA surface opens or returns to the foreground, with server-side caching matched to the source. Slower-changing annual and regulatory datasets are reviewed against their release cycle. Clear changes from official sources can be applied; ambiguous or unverifiable changes are held rather than guessed.

A freshness target describes how old MFA wants a verified source check to be before another review is due; it does not imply 24/7 background polling. An automatic source check also does not mean MFA will overwrite uncertain data. For broader credit-card rewards, credits and lounge rules, MFA can flag that an official issuer page changed without translating that page drift into new economics. Changes that are not explicit on the primary source stay in a verification state, and calculation-affecting updates must pass the MFA regression suite before becoming the new known-good state.

Savings APYs can change at any time. When the Savings experience loads or returns after its cache is stale, MFA re-checks machine-readable official bank pages and can overlay clear numeric changes without waiting for a code release. If a bank blocks automated access, hides pricing behind a ZIP/relationship lookup, or exposes terms ambiguously, MFA keeps the last verified snapshot and labels it rather than guessing. Standard APYs and promotional APYs remain separate; tiered rates use the balance entered by the user. Automatic bank parsing is intentionally narrow: only explicit APY, promotion and supported tier fields can override the verified snapshot. Fees, minimums, access rules and eligibility language remain source-reviewed unless MFA has a dedicated parser for that field.

Automatic bank-source check completed Oct 27, 2026, 8:53 AM. “Rate as of” preserves the institution's effective/disclosure date where available.

MFA Deals starts with the official bank or issuer offer page. Third-party deal roundups and community sources can be used as discovery signals to find an offer worth checking, but they are never the source of truth for a numeric MFA Deal. Before a new or changed bonus enters the public catalog, MFA checks the original bank, issuer, retailer or program page and normalizes the verified hurdle, deadline and important restrictions. Supported official deal pages can then be re-checked when the Deals experience refreshes, with a server cache of up to about six hours. MFA verifies expected offer signatures such as the bonus amount, deadline and key qualifying hurdle rather than assuming that a URL which still loads still contains the same offer.

The Deals comparison can also model cash opportunity cost, administrative-time return, explicit qualifying-transaction counts and encoded state availability. Unknown regional availability remains a verify state rather than being assumed eligible. Those are MFA calculations, not issuer promises and not investment-return projections. Tax treatment, eligibility, fees and the user's ability to complete an offer still require individual review.

Decision Lab: snapshots, sources and reminders

Decision summaries preserve user-entered reasoning, attached result summaries and captured assumptions. The capture time records when the result was kept; it does not verify the underlying rate, offer or rule. Reopen the linked tool to check its source dates and model exclusions before acting. Comparing two saved results requires compatible inputs and time horizons.

The optional quick cash comparison is upfront cash plus monthly cost multiplied by the entered months. It assumes constant costs and does not model equity, amortization, taxes, opportunity cost or investment returns. Review deadlines are entered by the user. In-app reminders appear when the saved workspace is opened; calendar reminders activate after importing the calendar file and follow calendar delivery settings. Editing a saved date does not update an already imported calendar event. Browser drafts remain local unless copied to a signed-in account.