Census Advance Monthly Sales Report (PDF/Data Tables)

Census Advance Monthly Sales Report (PDF/Data Tables) is the US government's flagship monthly read on the American shopper, delivered by Datadory as structured rows rather than a formatted document. The July 2026 edition put the advance estimate of US retail and food services sales at $763.6 billion - down 0.6 percent from June, up 5.0 percent year over year. Three tables ride behind the headline: Table 1 holds monthly sales levels for every kind of business on not-adjusted and seasonally adjusted bases at NAICS detail; Table 2 holds the month-to-month and year-over-year percent changes; Table 3 holds estimated sampling variability.

The detail is where retail landlords live in it. Kinds of business run from total retail (NAICS 44-45) down to 3-4 digit branches, and food services and drinking places (NAICS 722) posts $107.2 billion not adjusted for July 2026 by itself - the restaurant-and-entertainment draw that separates a mall from a strip. Underneath sits a stratified random sample of roughly 4,800 employer firms, weighted to represent an universe of over three million retail and food services firms and benchmarked to the Annual Retail Trade Survey.

Within Datadory's catalog of 1,744 datasets across 159 viable industries, the retail REIT slice holds four catalogued records averaging 8.0 on our quality rubric against a 7.81 catalog-wide mark; this one rates 8/10, credited to its unmatched history and its stated error bars rather than withheld ones. Get a sample of this dataset scoped to your kinds of business.

Rows from the July 2026 edition (release CB26-131), exactly as delivered:

# ADVANCE MONTHLY SALES FOR RETAIL AND FOOD SERVICES - JULY 2026 (release CB26-131)
headline    : $763.6 billion advance estimate
mom_change  : -0.6% vs June          yoy_change : +5.0%

# Table 1 row - food services & drinking places (NAICS 722), $ millions
ytd_2026      : 711780       ytd_pct_change : +4.2
not_adjusted  : jul_2026 : 107179     jun_2026 : 105284
adjusted      : jul_2026 : 103555     jun_2026 : 103018

# single kind-of-business extract, 2026 monthly values jan-jul
sales           : 99554  100240  100241  101267  102588  103018  103555
seasonal_factors: 0.925  0.912   1.033   1.004   1.078   1.022   1.035

Read what the grid already says. The NAICS 722 row moves from a $105.3 billion June to a $107.2 billion July on a not-adjusted basis while the adjusted series edges up far less - that gap is seasonality made visible, and the seasonal-factor line beneath it prices the effect explicitly. Year-to-date food-services sales sit at $711.8 billion, up 4.2 percent, so the category's growth rate arrives on the same row as its level. Nothing here is reconstructed: every figure traces to release CB26-131 with its vintage flag intact.

Nine fields carry the record, all verified against the July 2026 edition during the August 2026 research pass rather than inferred from headings. Identity comes from naics_code and kind_of_business, a pair that stays aligned so a filter on '722' always resolves to food services and drinking places and never to a relabeled row. Levels come in two bases - sales_not_adjusted for the calendar as retailers lived it and sales_adjusted for the seasonally smoothed view - each stamped with its vintage flag, (a) advance, (p) preliminary or (r) revised.

Change and reliability complete the set: ytd_sales_millions and ytd_pct_change accumulate the year, pct_changes carries the month-over-month and year-over-year moves markets quote on release day, and cv_standard_error publishes how much sampling noise rides on each advance estimate. footnote_flags closes the loop with the honest negatives - (*) for changes too small to clear 90 percent confidence, (S) for values withheld because their coefficient of variation exceeds 30, (NA) for simply unavailable. Wider deliveries fold under additional fields on request so the core nine stay immediately testable.

• Geography - United States, national totals. This record deliberately stops at the national line: it answers what the American consumer did, not what a specific trade area did, and pairing it with a landlord fundamentals panel is how geography re-enters the question.
• Temporal - monthly, reaching back to October 1953 across 848 archived releases, with continuous kind-of-business series from 1992 forward. Seventy-plus years of demand history means the series has already priced every cycle a backtest will meet.
• Granularity - one observation per kind of business per month, from NAICS subsector down to 3-4 digit branch detail, each delivered in both not-adjusted and seasonally adjusted form with sampling variability attached at Table 3.

That span is why this record pairs naturally with the Nareit T-Tracker Quarterly Operating Performance: a demand month lines up against a landlord quarter on calendar periods, so consumer spend charts straight onto occupancy and same-store NOI.

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

You choose the channel and the cadence; the nine-field dictionary travels unchanged through all three. Rows land flattened and typed - codes as strings, sales as numbers, vintage flags as enums - so a consumer-spend table joins against your tenant-mix mapping or factor pipelines on the first attempt instead of after a document-extraction project. Cadence changes are a setting, not a migration, and a sample scoped to your kinds of business proves the shape before anything ships.

• Tenant-demand tracking - map a center's tenant mix to kinds of business and watch the demand line move ahead of occupancy; NAICS 722 isolates the food-and-beverage draw on its own.
• Same-store NOI context - join demand months to landlord quarters and separate a weak quarter caused by the consumer from one caused by the asset.
• Release-day macro reads - the $763.6 billion July headline with its -0.6 and +5.0 percent changes is the fastest national shopper read available.
• Revision-aware analysis - advance, preliminary and revised vintages survive as distinct observations, so point-in-time reconstruction needs no archaeology.
• Category growth screens - year-to-date percent changes across every kind of business rank which categories compound and which contract, monthly.
• Model features with error bars - Table 3's coefficients of variation attach a stated uncertainty to every advance estimate a model consumes.

Investors and quants treat seventy-plus years of monthly demand as the exogenous series behind every retail REIT valuation (data scientists use cases). Analysts attribute a soft quarter to the right tenant category using kind-of-business detail, and cite the vintage they relied on. Competitive-intel product teams build shopper-trend views on a schema that has not moved in decades (competitive intel product teams use cases). Developers ship a consumer-spend widget off one stable field dictionary (developers builders use cases). Journalists and academics quote national figures whose statistical caveats are printed beside them, on a series that predates the institutions reporting it.

The demand layer reads differently depending on its neighbours. The Nareit T-Tracker Quarterly Operating Performance supplies the landlord half - FFO, NOI, occupancy and leverage across thirteen property sectors - and the trade-off between measuring the shopper and measuring the landlord gets its own page at Nareit T-Tracker vs Census Advance Monthly Retail Sales. The Nareit Retail REIT Sector Directory names the roughly thirty companies whose portfolios absorb this spending, and the NASDAQ Stock Screener API widens the frame to the 7,086-security listed universe whenever a screen needs market-cap or IPO-year context beyond retail proper. Joined on period and ticker, the stack answers the question neither half can answer alone - whether the shopper's month shows up in the landlord's quarter.