Monthly Wholesale Trade Survey (MWTS): Sales & Inventories

One number answers "how much product moved through American warehouses this month" and a second answers "how much is still sitting on the shelves." MWTS is the federal government's monthly estimate of both: dollar sales by US merchant wholesalers, their end-of-month inventories, and the inventories-to-sales ratio that turns the pair into a single restocking signal. Coverage runs across NAICS sector 42 - every kind of business from motor vehicle parts (4231) to beer, wine and distilled spirits (4248) - with manufacturers' sales branches explicitly excluded so the read stays on true distribution businesses.

The history is what makes it usable. Adjusted monthly values reach back to January 1992, when total merchant wholesaler sales ran about $143 billion a month; the June 2026 preliminary print sits at roughly $794 billion. Thirty-four years on one axis means the 2008-09 inventory collapse and the 2021-22 restocking squeeze appear as data, not anecdotes.

Datadory delivers the full panel as a production feed - sales, inventories and ratios side by side, shaped for whatever your models or dashboards expect.

One column per kind of business, one row per month, dollars in millions:

month     : June p   year: 2026
  42   (total wholesalers)          : 794101
  423  (durable goods)              : 397510
  4231 (motor vehicle & parts)      :  54697
  4232 (furniture & furnishings)    :  10828
  4233 (lumber & constr. materials) :  18321
  4234 (prof. & comm. equipment)    :  64158

month     : January  year: 1992
  42   (total wholesalers)          : 142980
  423  (durable goods)              :  67861
  4231 (motor vehicle & parts)      :  12652
  4232 (furniture & furnishings)    :   2533

Two rows, thirty-four years apart, same schema. That stability is the point: nothing about the layout changed while the sector grew more than fivefold, so a backtest written against the 1990s rows runs unchanged against last month's. Request a sample and you get rows in exactly this shape, extended across whichever industries and years you name.

Three identifying fields and twenty-two measure columns define every row, and each measure column maps one-to-one to a NAICS kind of business - which is why joins to any other NAICS-keyed dataset never drift. The dictionary below covers the core schema.

Temporal - monthly, from January 1992 to the present in the historical time series, with annual revision reports available back to 1997. Every month carries two estimate vintages: the preliminary (p) figure in the first release and the revised (r) figure a month later, flagged right in the row label.

Geographic - United States, national totals only. There is no state or metro breakdown; the value here is depth of history and industry cut, not map views.

Granularity - one row per month, twenty-two industry columns spanning three levels of NAICS 42: subsector totals (Durable Goods 423, Nondurable Goods 424), all seventeen four-digit component industries, and one selected five-digit line - Computer and Computer Peripheral Equipment and Software wholesalers (42343). Three measures ship for the whole panel: sales, end-of-month inventories and the inventories-to-sales ratio.

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

• Demand forecasting - train on a monthly series where the seasonal pattern repeats for thirty-four years without a schema break; the p/r flags let you model publication lag honestly instead of pretending the first estimate was the final one.
• Inventory-cycle positioning - the inventories-to-sales ratio is the classic early-warning gauge: when ratios rise while sales flatten, distributors are stocking ahead of demand, and that turn shows up here before it shows up in anyone's quarterly report.
• Sector benchmarking - compare a distributor's growth against its own NAICS line rather than the wholesale aggregate; a machinery distributor up 4% in a market up 9% is losing share, and only the four-digit cut reveals it.
• Macro and policy work - wholesale trade feeds directly into quarterly GDP estimation, so the series doubles as a high-frequency check on the income side of the national accounts.
• Credit and risk underwriting - track the financial health of the distribution channel a borrower sells into, industry by industry, month by month.

Data scientists and ML engineers get a fixed-schema, NAICS-keyed monthly panel long enough to train on across multiple inventory cycles. Investors and quant researchers get the restocking signal that leads distributor earnings, with the p/r structure preserved for realistic backtests. Competitive intelligence and strategy teams get the category baseline their internal numbers should be judged against. Market researchers and consultants get a citable federal benchmark behind every chart. Journalists, academics and students get one authoritative series for the wholesale economy instead of stitching press releases.

Three things worth knowing upfront. First, coverage is national only - if your question needs state or metro resolution, this dataset answers a different question than the one you're asking. Second, values are dollars in millions, not units sold, so pairing against physical-volume data requires a price deflator on your side. Third, the numbers get revised: each month's first estimate is flagged preliminary and superseded a month later, and Census has announced a comprehensive historical revision incorporating 2023-2024 economic census results - series values may shift when it lands, which is precisely why the vintage flags travel with the rows.