NerdWallet Mortgage Rates Data

It is the retail rate board for American home lending, published by NerdWallet and rebuilt through every trading day. One surface carries two layers, and keeping them separate is the whole trick of using it well.

The first layer is ten national product averages - 30-, 20-, 15- and 10-year fixed, 30-year FHA, 30-year VA, and 3-, 5-, 7- and 10-year adjustable-rate mortgages - each quoted as an interest rate and as an APR, beneath a headline chip tracking the 30-year fixed with a one-week change. These are attributed to Zillow and stamped to the minute.

The second layer is a lender offer panel: roughly eleven partner cards rendered against a named borrower scenario - purpose, credit-score band, location, loan amount, down payment, term, property type - each carrying its own APR, interest rate, estimated monthly payment, total fees, NMLS identifier and an editorial star rating with pros-and-cons commentary. Offers arrive via the Mortech rate engine and ship flagged as not binding.

Within the Consumer Finance shelf this is the quoted-market record - what borrowers were actually shown, product by product, day by day. It scores 6/10 on our rubric against a catalog average of 7.81 across 1,744 datasets: carried by verified field definitions and minute-stamped rows, held back by single-page breadth and an offer panel that rotates its partners. Get a sample of this dataset cut to the products you track.

Both grains in one capture, from the August 21, 2026 verification pass:

# national averages table - one row per product, stamped August 21, 2026
product            interest_rate    apr
30-year Fixed      6.62%            6.63%
30-year Fixed FHA  5.38%            6.11%
30-year Fixed VA   6.09%            6.43%
15-year Fixed      5.93%            5.95%
5-year ARM         6.86%            6.69%

# headline chip above the averages table
30-year fixed apr 6.63%            weekly_change +0.09

# lender offer cards - one row per partner, rendered for one borrower scenario
scenario       : Purchase, Good (720-739) score band, 30-year fixed,
                 single-family, primary residence
apr            : quoted per card   interest_rate : quoted per card
est_mo_payment : 2463              total_fees    : 5536
nmls           : 2059741           stars         : 4.5/5
stamp          : 2026-08-21T09:40:00-04:00

Read the spreads rather than the levels. The FHA row quotes 5.38 against a 6.11 APR - a near-three-quarter-point gulf that is the insurance-and-fee load folding into the APR, visible nowhere in the bare rate. The VA row sits below the conventional 30-year on rate yet closes much of the gap once its fee load lands. The 5-year ARM inverts the usual order - rate above the 30-year, APR below it - which tells you the adjustment assumptions, not just today's teaser. On the offer side, one card prints a 2,463 monthly payment against 5,536 in fees for the Good-band purchase scenario, and the NMLS identifier stays constant even when the card itself rotates out. Every row ends at the stamp, which is what turns a retail screen into a citable observation.

Ten documented core fields, verified during the August 2026 research pass - nothing inferred from labels alone. Four carry the averages layer: product, interest_rate, apr and weekly_change. Four carry the offer layer: est_monthly_payment, total_fees, nmls and nerdwallet_star_rating. Two bind the grains together: scenario_parameters - purpose, score band, ZIP code, amount, down payment, term, property type, occupancy - without which offer rows are incomparable, and rate_timestamp, which makes any row citable to a minute.

Everything deeper folds out as additional fields on request: the full commentary text behind each star rating, the historical points behind the on-page charts across windows from one month to ten years, alternative scenario renders, and fee detail beyond the printed total. Their shape depends on the cut you specify, so they confirm with the sample rather than being promised blind.

Geography: United States throughout. National averages hold regardless of where the reader sits; the offer panel localizes to ZIP-code level once the scenario supplies one, alongside the credit-score band and property inputs that drive the cards.

Temporal: current-day snapshots stamped intraday - the verification pass caught August 21, 2026 at 9:40 AM Eastern. History lives on the page as chart series across windows from one month to ten years, filterable by purpose and product, so a trend question never requires waiting for an archive to build.

Granularity: two grains, deliberately kept apart. Ten product-level national averages form the market board; roughly eleven lender offer cards per scenario view form the quoted-offer layer, keyed by scenario parameters and stamp. Neither grain pretends to be the other.

Set against the wider Datadory catalog - average quality score 7.81 across 1,744 datasets - this record scores 6/10, held back by its single-page breadth and partner rotation rather than by documentation, which is complete.

API, files, or your warehouse. Daily, weekly, or hourly.

You choose the channel and the cadence; the field dictionary above travels unchanged across all three. Daily suits teams building the national product-day series off the morning print. Hourly suits desks watching for midday re-quotes. Weekly suits research cycles that only need the trend. Deliveries arrive normalized to the dictionary - averages as product-day rows, offers as scenario-keyed snapshot rows with their stamps attached - so consecutive days diff cleanly in your warehouse instead of arriving as screenshots of a web page. Scope the extract to named products, named scenarios and named windows before anything commits.

1. Secondary-market and mortgage strategists read the national board daily and watch the FHA-to-conventional and ARM-versus-fixed spreads move instead of relying on weekly prints.
2. Lender product and quote teams position their own quotes against the scenario-keyed panel - knowing what a Good-band borrower was shown, and by whom, is competitive intelligence with a timestamp.
3. Quant and ML builders assemble daily rate features for housing and consumer models; see ML model training.
4. Housing economists track government-backed share behavior through the FHA and VA rows sitting in the same table as conventional products.
5. Journalists cite a dated daily figure for the 30-year fixed with the stamp attached; see citation grade research.
6. Homebuying and relocation platforms embed market context that survives scrutiny because every number carries its minute.

Get a sample of this dataset scoped to the products and scenarios your work touches.

Investors and quants turn the daily board into mortgage-market features and screens, with the FHA, VA and ARM legs riding beside the conventional curve (investors quants use cases). Market researchers and consultants quantify how lenders compete for named borrower segments through the scenario-keyed panel (market researchers use cases). Data scientists and ML engineers land model-ready rows with scenario covariates and observation times (data scientists use cases). Journalists, academics and students cite the quoted market to the minute, with rotation and attribution caveats stated up front (journalists academics use cases). Developers building data products inherit a stable ten-field dictionary keyed on product, scenario and stamp.

Within the wider mortgage shelf, the neighbors answer different questions. The FHFA House-Price Index values the collateral underneath every quote here. FRED Mortgage and Housing Data supplies official weekly survey series with deep history - slower, but regulator-published; the weekly benchmark behind it is the primary mortgage market survey. The Freddie Mac Loan-Level Dataset records what actually closed, loan by loan, months after the quoting; FFIEC HMDA captures originations as regulated disclosures. And Bankrate Mortgage & Financial Product Rates runs the nearest retail playbook across several product families. This record's edge is pairing: a same-day national board and a scenario-keyed lender panel in one normalized delivery.

Four honest caveats. First, the offer panel is partner-driven and rotates: the same scenario can render different lenders hour to hour, so treat the card layer as a market read keyed on NMLS identifiers, not a stable quote ledger. Second, the averages carry an upstream attribution and the publisher's own usage terms, which bar republication - analyze and embed the figures, do not republish the table. Third, scenario state is part of the record: an offer row without its score band, location and amount is meaningless, which is why scenario_parameters ships as a field rather than metadata. Fourth, stamping is intraday, so a daily cadence captures the morning print unless hourly pickup is requested. None of these bite unexpectedly; they ship flagged against the analysis you plan to run.

Because the raw artifact is a single client-rendered retail page mixing two grains and a scenario state machine - averages for everyone, offers quoted to one hypothetical borrower, history trapped in chart payloads. Datadory reshapes it into tidy product-day average rows and scenario-keyed offer-snapshot rows, preserves every stamp, attaches scenario parameters as columns, and cuts to the named products, scenarios and windows you specify. Browse the rest of the industry on the Consumer Finance data hub, the ranked shortlist of the best consumer finance datasets, or the adjacent Commercial & Residential Mortgage Finance hub.