Guide

ISM Manufacturing PMI explained

Harbor Manufacturing's operations desk treated the first-business-day ISM print like a earnings surprise: a headline miss of 49.8 versus 50.2 consensus triggered an immediate capex freeze email. Two weeks later, Federal Reserve industrial production rose 0.4% and factory orders beat. The ISM had not lied — the desk had misread what a diffusion index measures. Purchasing managers reported slightly more firms contracting than expanding; that is not the same as output falling, and a 0.4-point headline miss is often noise inside survey standard error.

The ISM Manufacturing PMI (formerly the NAPM survey) polls purchasing and supply executives at roughly 300 U.S. manufacturers each month. The headline number summarizes whether more respondents see activity improving than deteriorating. It leads hard data by one to three months when you read the right subindexes — especially new orders and customer inventories — but it is a breadth gauge, not a production volume index. After Harbor rebuilt its release-day playbook around subindex spreads and confirmation rules, false alarms dropped and the team stopped equating 49.9 with recession. This guide covers diffusion math, the major subindexes, comparison with S&P Global PMI, links to GDP and inflation data, the Harbor refactor, a technique decision table, pitfalls, and a production checklist.

What the ISM Manufacturing PMI measures

The Institute for Supply Management publishes the survey on the first business day of each month, covering conditions in the prior calendar month. Respondents answer whether a given metric — new orders, production, employment, and so on — is higher, the same, or lower than the previous month. ISM converts those shares into a diffusion index:

Index = P₁ + 0.5 × P_same

where P₁ is the percentage reporting improvement and P_same is the percentage reporting no change. The index ranges from 0 to 100. 50 is the breakeven: equal shares improving and worsening (with no-change responses split evenly). Above 50 means net expansion breadth; below 50 means net contraction breadth. It does not mean output grew 52% or fell to 48% of prior levels.

Because the survey asks about direction, a sector can report PMI above 50 while absolute output is still below pre-crisis peaks — firms are improving from a low base. Conversely, PMI can slip below 50 while year-over-year production remains positive if the pace of improvement is slowing. That distinction matters when pairing ISM with Fed G.17 hard data and capacity utilization.

Headline PMI and the main subindexes

New Orders

The new orders subindex is the primary leading indicator inside the report. Order books turn before production schedules and shipments. Markets watch month-over-month changes and the spread versus headline PMI: when new orders lead headline lower for several months, factory output often softens with a lag. Harbor weights new orders at 40% in its internal manufacturing momentum score.

Production

The production subindex tracks current output schedules. It tends to move concurrently with industrial production and is less leading than new orders. Large gaps between new orders and production can signal backlog drawdown or front-loaded shipments.

Employment

The employment index reflects hiring intentions and headcount changes at surveyed plants. It often leads manufacturing payrolls in the establishment survey by a month but is volatile and can diverge when large plants outside the panel drive national jobs data.

Supplier Deliveries

Supplier deliveries measures whether lead times are lengthening or shortening. Unusually, slower deliveries (longer lead times) raise this index because they indicate supply bottlenecks and strong demand pressure. During 2021–2022, elevated supplier deliveries inflated the headline PMI even as some demand components cooled. Analysts often read supplier deliveries above 50 as inflationary pressure on goods pipelines.

Inventories and customer inventories

Inventories reports whether respondents' own stocks are rising or falling. Rising inventories with falling new orders can signal unwanted stockbuilding ahead of production cuts. Customer inventories asks whether downstream customers have too much or too little stock. Historically, very low customer inventories (subindex well below 50) preceded restocking waves that boosted production even when headline PMI was soft.

Prices Paid

Prices paid tracks input cost pressure reported by purchasing managers. It leads PPI goods components with a short lag and is a useful check on goods-side inflation when services CPI dominates headlines. ISM also publishes an order backlog and new export orders series in the full release.

Reading the report: levels, momentum, and spreads

Three lenses beat staring at the headline alone:

Level vs 50 — sustained stretches below 50 align with manufacturing recessions in hard data; one print at 49.7 is weak signal.
Three-month momentum — the change in headline and new orders over 90 days tracks better with IP growth than any single month.
Internal spreads — new orders minus inventories, production minus new orders, and prices paid minus supplier deliveries each tell different stories about demand, pipeline, and inflation.

A classic late-cycle pattern: new orders fall below 50 while production stays above 50 as plants work through backlog. A classic early-recovery pattern: new orders rebound first while employment lags because firms extend hours before hiring. Mapping these sequences onto the business cycle reduces whipsaw reactions to headline surprises.

ISM vs S&P Global (Markit) Manufacturing PMI

S&P Global publishes a competing U.S. manufacturing PMI from a different panel and methodology (GDP-weighted by sector, five-level response options converted to diffusion indexes). The two series usually directionally agree but can diverge by 2–4 points in a given month because of sample composition: ISM overweight smaller firms and traditional industries; S&P Global weights large multinationals and uses different seasonal adjustment.

When ISM and S&P Global disagree materially, Harbor checks three things: whether the gap is in new orders (often the driver), whether auto and aerospace respondents drove the ISM move, and whether S&P's flash estimate was revised at final release. Neither survey is “correct” — they sample different slices of manufacturing. Convergence across both plus hard data confirmation is stronger than either alone.

How PMI connects to GDP, jobs, and markets

Industrial production and durable goods

Regression studies show ISM new orders correlate with future IP growth with roughly one- to two-month lead. Durable goods orders — especially core capital goods — provide a second confirmation layer. PMI is timelier; Census factory data is more granular but arrives later and revises.

Equities and cyclicals

Equity futures often move on the 10:00 a.m. ET release. Industrial and materials sectors are most sensitive; pure services names react only if the print shifts Fed path expectations via the prices paid channel. A sub-50 headline with rising new orders has historically produced smaller drawdowns than sub-50 with broad-based weakness across all five major subindexes.

Fixed income and the Fed

Bond markets weight prices paid and supplier deliveries when goods inflation is the policy debate. A soft headline with sticky prices paid may do little to ease two-year yields. Conversely, a strong headline with collapsing prices paid can support easing bets even when manufacturing is a small share of GDP.

Harbor Manufacturing release-day refactor (worked example)

Harbor's legacy playbook sent capex alerts on any headline miss versus consensus. Refactor steps:

Subindex dashboard — auto-plot new orders, production, employment, supplier deliveries, inventories, customer inventories, and prices paid with three-month trails.
Confirmation rule — require two of three signals (new orders below 50, new orders momentum negative, customer inventories above 50 implying downstream plenty) before flagging demand contraction.
Hard-data cross-check — overlay last IP print, durable goods core orders, and weekly railcar loadings where available.
Spread alerts — notify only when new orders minus inventories falls below a calibrated threshold for two consecutive months.
Prices paid filter — route inflation shocks to the procurement desk separately from volume alerts to avoid conflating cost and demand.
Post-release debrief template — one paragraph: breadth (how many subindexes below 50), lead (new orders vs production), and inflation (prices paid direction).

Outcome: false capex freezes fell from four per year to one; the team caught the 2024 mid-year manufacturing soft patch one month earlier than the legacy headline rule because new orders led production below 50 for three months while headline PMI hovered near 49–50.

Technique decision table

Your goal	Prefer	Avoid
Forecast factory output next quarter	ISM new orders 3-month average + IP trend	Single headline PMI vs consensus
Detect goods inflation pressure	Prices paid + supplier deliveries	Headline PMI alone
Time manufacturing hiring	ISM employment with 1-month lag vs payrolls	Assuming instant layoffs on one sub-50 print
Recession watch for industry	Sustained sub-50 breadth across subindexes	One 49.5 headline in isolation
Cross-check survey signal	ISM + S&P Global + durable goods orders	Trading only the first tweet headline

Common pitfalls

Treating PMI like a growth rate — 48 is not −2% output; it is net contraction breadth.
Ignoring supplier deliveries inversion — longer lead times boost the index even when demand is cooling.
Overreacting to 0.1–0.3 consensus misses — survey noise dominates small deltas.
Skipping customer inventories — downstream stock levels often explain production rebounds.
Assuming ISM equals GDP manufacturing — services dominate U.S. GDP; manufacturing is cyclical but smaller.
Comparing ISM to euro-area PMI without adjustment — different panels, seasons, and sector weights.
Using prices paid as CPI forecast — it leads PPI goods, not shelter or services CPI.
Chasing the futures spike — reversals are common once subindex tables are parsed.

Production checklist

Calendar the ISM release (first business day, 10:00 a.m. ET) with S&P Global final PMI dates.
Archive full subindex table, not just headline, in your data warehouse.
Compute three-month moving averages for new orders and headline PMI.
Track new orders minus inventories and production minus new orders spreads.
Cross-check each major ISM move against prior-month IP and durable goods.
Flag prices paid moves above 55 or below 45 for procurement and rates desks.
Document sector anecdotes from ISM comments section for qualitative color.
Reconcile ISM employment direction with next manufacturing payroll print.
Include PMI breadth (count of subindexes below 50) in weekly macro notes.
Review false signals quarterly and recalibrate confirmation thresholds.

Key takeaways

ISM Manufacturing PMI is a diffusion index: 50 marks equal breadth of improvement and deterioration, not zero growth.
New orders and customer inventories lead; production and employment lag; prices paid and supplier deliveries inform inflation.
Headline surprises matter less than sustained sub-50 breadth and three-month momentum.
ISM and S&P Global PMI complement each other — divergence warrants sector-level investigation.
Harbor Manufacturing cut false alarms by requiring subindex confirmation instead of trading every consensus miss.