Census Economic Indicators Time Series (EITS)

One service holds the recurring half of the U.S. government's economic dashboard: eighteen indicator programs - monthly retail trade and food services (mrts), its advance cousin (marts), monthly wholesale trade (mwts), manufacturers' shipments, inventories and orders (m3), the advance durable-goods report, manufacturing and trade inventories and sales, the quarterly services survey, the quarterly financial report, quarterly state and local tax summaries, the quarterly survey of public pensions, business formation statistics, international trade in goods and services, housing vacancies and homeownership, new residential construction, new home sales, the manufactured homes survey and construction spending.

What makes the family unusually tractable is the shared spine. Every program returns observations keyed the same way - an industry category, an item type, a period, a seasonally-adjusted flag - with the reading itself in cell_value. Learn one table shape and you can sweep the whole indicator suite: a retail-sales model, a wholesale inventory cycle, a housing-starts pipeline and a business-formation leading indicator arrive as siblings, not strangers. Category detail runs deep enough to matter operationally: retail rows split to store-group level, so an apparel platform can benchmark itself against category 448 - clothing and clothing accessories stores - rather than against all of retail.

Get a sample of this dataset and we will return rows cut to the programs, categories and periods you name.

Rows come back as flat JSON arrays of arrays, headed by the six working columns of the retail-trade program:

# Monthly Retail Trade and Food Services (mrts) - tabular JSON
["cell_value", "data_type_code", "time_slot_id",
 "seasonally_adj", "category_code", "error_data"]

cell_value     : <reading>    # the indicator value for the period
data_type_code : <code>       # the measure - e.g. sales or inventories
time_slot_id   : <period id>  # the month or quarter the reading belongs to
seasonally_adj : yes          # or no - both variants exist side by side
category_code  : 448          # clothing and clothing accessories stores
error_data     : no           # sampling-error variants flagged when asked for

A single-year slice of the retail program alone returns hundreds of rows once categories, measures and both seasonal variants multiply - which is precisely why the fixed column order matters: the same parse works whether you pulled one category for one month or the entire suite for a decade. Period selection is equally uniform - a single month, an explicit from-to range, or an open-ended window starting from a given year - so backfilling a long history and keeping a current feed current use the identical shape.

Twelve documented columns, verified against the Bureau's own variable documentation - four are selectors you set, the rest travel with every row so a merged multi-program table never loses its labels:

• Geography: United States, national level on every program; several programs additionally support finer geography selections, and the trade program reports the goods-and-services balance with world partners.
• Temporal: decades-long histories on the flagship series - the retail pair is demonstrably queried back to 2004-05 and earlier - selectable by year, by month range, or as an open window from a start year to the latest print.
• Granularity: one observation per industry category per item type per period, monthly on most programs and quarterly on the surveys, with seasonally adjusted and non-adjusted variants carried side by side.

Set against the wider Datadory catalog - where the average quality score across all 1,744 datasets is 7.81 - this slice scores 9/10, carried by fully verified field documentation and the sheer breadth of the indicator family.

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. Bulk files suit analysts who want twenty years of retail and wholesale history loaded once and joined to their own macro tables. Structured payloads suit products that surface a single indicator - this month's category reading, say - inside an app or a dashboard. Warehouse delivery suits teams running nowcasts in SQL against a dozen indicator programs at once. Cadence changes are a settings conversation, not a re-integration project.

• Retail demand tracking - the advance and revised retail pairs separate early signal from final print, and store-group category detail turns 'consumer spending softened' into 'clothing accessories moved X direction last month'.
• Wholesale and inventory-cycle work - wholesale sales and inventories alongside manufacturing-and-trade inventories expose the inventory-to-sales behavior that leads restocking and freight demand.
• Manufacturing pipeline reads - shipments, inventories and orders, plus the advance durable-goods report, give the bookings picture behind capital-goods and industrial theses.
• Housing and construction monitoring - residential construction, new home sales, manufactured homes and total construction spending chain the shelter pipeline end to end, feeding mortgage and home-improvement models.
• Business-formation leading indicators - new business applications are a forward-looking input for lenders, payments providers and anyone sizing small-business market entry.
• Macro nowcasting - trade in goods and services, quarterly services and corporate financial reports plug directly into high-frequency GDP-tracking exercises.

Investors and quants get official-print inputs for nowcasts and regime detection - the release calendar stops being a surprise because the underlying readings are already in the warehouse. Data scientists get one uniform spine across eighteen heterogeneous programs, so feature engineering is a groupby rather than eighteen bespoke parsers. Developers and builders get a single table contract to integrate once, then swap programs in and out without touching downstream code. Competitive-intel and product teams get the authoritative baseline their company's own numbers should be quoted against - category-level, revised, and citable. E-commerce operators get category demand baselines to separate their own share shifts from the market's tide.

Three things worth knowing upfront. First, advances are not finals: the advance retail estimates exist precisely because the revised series lag, so any model mixing the two needs an explicit revision policy rather than a quiet overwrite. Second, the seasonal flag is a fork, not a footnote - adjusted and non-adjusted readings of the same measure will disagree by design, and picking one variant per analysis is the difference between a trend and an artifact. Third, the codes are compact on purpose: category and item-type identifiers resolve through the Bureau's published data dictionaries rather than an inline lookup, and sampling-error variants ride along only when the error flag is raised - both come documented with your sample, so nobody reverse-engineers a code table from memory.

Notes that pair well with this page:

• Consumer finance data hub - the full pooled view of the industry, from federal statistics to complaint archives.
• vs CFPB Consumer Complaint Database - macro spending prints against individual complaint records, two altitudes on the same consumer.
• Monthly Retail Trade Survey (MRTS) and MARTS advance estimates glossary entries - the two series behind the headline retail numbers, defined.
• U.S. Census Bureau source profile - the statistical agency behind the indicator family, and what else it publishes.