ISM Report On Business (PMI)

Two numbers move distribution desks every month, and both come from this survey. The Institute for Supply Management polls a national panel of purchasing and supply executives - the people whose job is knowing whether orders, materials and inventories are turning - weighted by each industry's contribution to GDP. The answers become the Manufacturing PMI, released on the first business day of every month, and the Services PMI, released on the third business day, both at 10:00 a.m. Eastern.

Everything in the reports is a diffusion index: the share of respondents reporting higher conditions, plus half those reporting no change. That construction makes 50 the line between expansion and contraction on every series, so a composite at 55.6 says something different from one at 50.4 even though both sit above water. Manufacturing stacks a five-part composite - New Orders, Production, Employment, Supplier Deliveries, Inventories - with six further measures beside it; Services runs its own composite on Business Activity through Prices. Respondent commodity mentions and verbatim comments by NAICS industry ride along with the tables, which is where channel-level color hides inside a national print.

Within Datadory's catalog of 1,744 datasets across 159 viable industries, this record scores 6/10 on the quality rubric with field definitions verified - marked honestly for national-only granularity rather than penalized for the numbers themselves, which have been quoted by every market desk since before most of them existed. Get a sample of this dataset.

One row per index per report month, with the prior month riding beside it so direction needs no lookup:

# verified rows from the July 2026 Manufacturing report
report_month : 2026-07
index        : Manufacturing PMI
value        : 55.6     # composite, five components rolled into one
prior_month  : 53.3     # June - the move itself is the signal

report_month : 2026-07
index        : New Orders
value        : 56.7
prior_month  : 56.0     # demand still building, just slower

report_month : 2026-07
index        : Production
value        : 58.5
prior_month  : 52.2     # factories catching up to orders

report_month : 2026-07
index        : Employment
value        : 52.8
prior_month  : 49.7     # back above water after contracting

report_month : 2026-07
index        : Prices
value        : 71.1
prior_month  : 73.0     # input-cost pressure easing but hot

Read the anatomy rather than any single digit. Every row keys on report_month and index, so the whole report is one filter away and a twelve-month history is a group-by. The prior_month column makes momentum a subtraction rather than a second join. And the spread between indexes tells the story the composite averages away: July's Production leap of 6.3 points against New Orders gaining only 0.7 says factories were filling a backlog faster than new demand arrived - exactly the destocking-versus-restocking call distributors make with it.

Twelve documented fields carry the record, definitions verified during research, split across the two report families:

• Manufacturing PMI — the composite, rolling New Orders, Production, Employment, Supplier Deliveries and Inventories into one diffusion number where 50 is breakeven. The headline every desk quotes.
• New Orders Index — share-weighted read on respondents reporting higher order volumes; the front edge of the cycle and the sub-index nowcast models reach for first.
• Production Index — output reported higher month over month; fills orders already booked.
• Employment Index — hiring intentions turned index; the labor side of capacity decisions.
• Supplier Deliveries Index — slower deliveries push it up, which is why it reads backwards: higher often means suppliers straining to keep pace, not failing.
• Inventories Index — raw-material stock change as buyers see it; the direct input to distributor restocking calls.
• Prices Index — input price changes paid; the inflation whisper that arrives weeks before official price deflators.
• Customers' Inventories Index — respondents' view of how stocked their own customers run; too low means restocking demand is queued.
• Backlog of Orders Index — unfilled-order drift; production already committed.
• New Export Orders Index — foreign demand on the same diffusion scale.
• Imports Index — imported-materials change; the trade-side companion to New Export Orders.
• Services PMI / Business Activity Index — the third-business-day counterpart covering Business Activity, New Orders, Employment, Supplier Deliveries, Inventories and Prices for the services economy.

Each field ships with its example values, the response distribution behind it and a trailing 12-month history in the delivered schema - mapped to your use case rather than flattened into a screenshot.

The report surface extends past the twelve core indexes; those extensions get mapped to your use case at sampling:

• Per-index response distributions — the share reporting better / same / worse behind each diffusion figure, so your models can weight breadth against magnitude themselves instead of accepting the pre-collapsed number.
• Trailing 12-month histories — each index's recent path as columns rather than chart pixels, ready for seasonal comparison without a scraping pass.
• Commodity price mentions — the named materials respondents flagged rising or falling in the month, the earliest anecdotal read on input costs.
• NAICS-segmented respondent comments — verbatim channel color organized by industry, the qualitative layer behind the quantitative panel.
• Seasonal adjustment factors — the adjustment machinery beside the raw diffusion readings, for teams that prefer to run their own smoothing.
• International counterparts — Taiwan's PMI/NMI issued jointly with CIER and Taiwan's NDC, where the same methodology runs offshore.

Name what you need with the sample request and the schema extends to match - nothing requires a second format.

Geography — the United States through one national panel, weighted by industry contribution to GDP. There are no state or metro cuts anywhere in the indexes; regional texture exists only as NAICS-segmented respondent commentary. Treat it as the single national read it is - the breadth is the point, not a limitation to patch.

Temporal — two clocks run together. The release clock is fixed and calendrical: Manufacturing lands on the first business day, Services on the third, both at 10:00 a.m. Eastern, so the ingestion schedule can be written a year ahead. The history clock is deep: the Manufacturing lineage reaches to 1931 and Services to 1948, while current-month report pages expose roughly the trailing twelve months of values - the long-run depth confirms per series when your sample is scoped.

Granularity — national composite and sub-indexes, one observation per index per month, with the respondent layer (commodity mentions, NAICS-segmented comments) beneath. Roughly thirty fresh numbers land between the two releases every month, each stamped with the report month and the prior-month comparison.

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

Name the indexes, the report families and the lookback when you request the sample - the full twelve-field surface, a Manufacturing-only strip, or the composite alone feeding a dashboard tile. The sample ships first either way; the ongoing feed lands on whatever cadence your models need, shaped to the scope you named rather than rebuilt from scratch every release morning.

A monthly diffusion panel earns its keep in six jobs:

• Cycle timing for allocation — Manufacturing and Services prints on fixed business days give quants the month's first demand read, hours before any government statistic stirs; see demand-forecasting workflows.
• Restocking versus destocking calls — the Inventories Index against Customers' Inventories flags whether the channel is drained or full, the exact decision a distributor's buying desk makes quarterly.
• Input-cost early warning — the Prices Index moves before producer-price releases do, so margin planning starts a month earlier.
• Citation-grade context — a survey running since 1931 survives any masthead or review committee; see citation-grade research.
• Nowcast features — New Orders, Production and Inventories drop straight into feature stores as typed rows with prior-month comparisons attached.
• Sales-cycle calibration — delivery-times and backlog sub-indexes tell sales teams whether customers are waiting on product or swimming in it, before territory plans get cut.

One editorial property makes all six defensible: the panel methodology has been stable enough to compare prints across decades, which is why a 55.6 means the same thing to a model trained last year and a researcher citing 1990.

Investors & Quant Researchers hold relevance 3/3: the PMIs as core monthly cycle markers landing on known calendar slots, with enough history to test whether the diffusion signals actually led (investors quants use cases). Market Researchers & Consultants (2/3) frame client demand commentary with the same prints everyone else quotes (market researchers use cases). Data Scientists & ML Engineers (2/3) get typed sub-index rows for nowcast features (data scientists use cases). Competitive Intelligence & Product Teams, Sales & Growth Teams, E-commerce Operators and Journalists, Academics & Students match at 3 - the most-cited US Purchasing Managers' Index readings, citable with their long-run lineage intact. Developers & Data-Product Builders hold relevance 1 and use it honestly - channel anecdotes from respondent comments, outreach timed to demand turns, promotional depth set against the Prices Index, and a small stable feed that never changes shape mid-year. Start from the trading companies & distributors data hub, then read what a Purchasing Managers' Index actually measures before wiring anything to it.

Provenance note — compiled and published by the Institute for Supply Management, the professional body for purchasing and supply executives. One editorial hand and one methodology across both reports is why a Manufacturing print and a Services print speak the same diffusion language, and why a 2026 reading compares cleanly against one from decades back.

Methodology note — diffusion indexes measure breadth, not magnitude: 100 would mean every respondent reported improvement, not that activity grew tenfold. Read levels against 50 and moves against the prior month; treat a 60 differently in a services month than a manufacturing one because the panels differ. The Supplier Deliveries Index runs inverted by construction - slower deliveries push it up - so pipelines should flip their sign convention once, deliberately.

Completeness note — Datadory scores this record 6/10, below the catalog mean of 7.81 across 1,744 datasets: the deduction reflects national-only granularity and a proprietary survey whose full historical depth sits behind the publisher rather than any weakness in the numbers. Field definitions are verified, and the five verified July 2026 rows above ship with the record.

Where to go next — the flows these indexes foreshadow live nearby: GDP by Industry — Wholesale Trade grades the demand story in official accounts, Eurostat short-term business statistics reads the same European turnover monthly, and ImportGenius bill-of-lading search names the shipments behind the sentiment. The full ranked slate sits on best trading companies & distributors datasets.