Yield Durability Score methodology

SwitchWize Research Desk | Version 0.1

A savings account's advertised APY tells you what a bank pays today. It does not tell you what that bank will still be paying in three months, or whether a "welcome rate" quietly drops once its promotional window ends. The Yield Durability Score rates how well an institution tends to hold the rate it advertises. Expected Realized Yield (ERY) turns that into a single dollarized number: what you'd expect to earn over the next 12 months on a $25,000 balance, not just on the day you opened the account.

What feeds the score

Each tracked institution is measured on five components, each scored 0-100:

• Competitive proximity (40% weight). The weighted share of observed days this institution's APY sat within 10, 25, or 50 basis points of that day's category leader.
• Cut resilience (20% weight). Built on the Rate-Cut Speed measurement below: how little of a Fed rate cut this institution actually passed through, relative to holding steady. Null, not zero, until this panel has observed at least one completed Fed rate-cut cycle.
• Rate stability (15% weight). Penalizes "silent drift" — APY decreases with no Fed decision anywhere nearby (within 3 days). A cut that lands near a real Fed decision is not penalized here; that behavior is captured by cut resilience instead.
• Terms integrity (15% weight). Penalizes the size of the APY drop when a promotional rate expires. An institution with no observed promotional activity is not penalized, but is flagged as such rather than assumed clean.
• Data continuity (10% weight). Coverage: the share of the trailing 365-day window (or the institution's full tracked history, if shorter) with an actual recorded observation.

The composite score is a weighted average of whatever components are available, with weights renormalized over the components present — a missing component is excluded, never treated as zero. At least 3 of the 5 components must be available for a composite score to be shown at all.

Confidence tiers

An institution needs a minimum amount of observed history before it is scored at all. Below that floor, the institution is shown as "Not enough history", never estimated.

• Tier A: at least 365 days of tracked history, at least 95% observation coverage.
• Tier B: at least 270 days, at least 90% coverage.
• Tier C: at least 180 days, at least 80% coverage.

These thresholds gate on observation depth and coverage only. They do not yet account for product-identity conflicts (a renamed or replaced product) or balance-tiered rate structures — both real limitations, disclosed here rather than silently assumed away, and both candidates for methodology v0.2.

Expected Realized Yield

ERY projects an institution's current APY forward 12 months and expresses the result in both percent and dollars on a $25,000 reference balance (the same reference balance used throughout SwitchWize's Bank Gap Index work). Version 0.1's base case is a hold scenario — no further Fed moves assumed — because a market-implied Fed path (Fed funds futures) is not yet integrated. Under that scenario, ERY reduces the current APY by:

• This institution's own historical silent-drift rate, annualized, and
• An assumed decay if a promotional rate is active and known to expire within the next 12 months.

ERY is null, not zero, when neither signal is available yet. A future version will add alternate Fed-path scenarios (faster cuts, an extended hold) once a market-implied path is integrated.

Rate-Cut Speed (cut resilience's input)

FOMC decision dates and target-range moves are derived directly from the Federal Reserve's own published upper and lower target-rate series (FRED DFEDTARU and DFEDTARL), never hand-entered. For each institution and each Fed rate cut, we find the institution's tracked APY immediately before the cut, then look forward for the first date its APY is recorded lower than that pre-cut level — the gap between those two dates is days-to-cut. Pass-through compares how much the APY moved to how much the Fed moved, over fixed 30- and 90-day windows:

passThrough = |change in APY over the window| / |change in Fed funds target|

An institution needs at least two observed Fed rate-cut cycles with real post-decision data to be ranked on cut speed alone. Cut resilience within the durability score uses the same underlying pass-through figures, but scores a lower pass-through (holding the rate) as more resilient — the inverse framing of the neutral, purely descriptive cut-speed table.

Published track record

Before each FOMC decision, predictions for each tracked institution's expected reaction lag and pass-through are logged and locked ahead of the announcement, based mechanically on that institution's own prior observed cycles — never a discretionary guess, and explicitly "no prior cycle" where none exists. Results are reconciled at 30, 60, and 90 days after each decision, and the running hit rate is published rather than restated. Predictions are never edited after the decision is announced.

What this does not tell you

Past repricing behavior is a track record, not a guarantee — an institution's funding position can change, and a bank that has historically held rates could still cut sharply if its own costs change. This score does not account for FDIC/NCUA insurance status, minimum balances, or account fees, which still matter and are shown alongside it, not replaced by it. It is not a recommendation to open or close any specific account.

Changelog