Liquidity LlamaLiquidity LlamaMarkets · Macro · Crypto
← Back to all reportsPMI Playbook · Aug 2026
Playbook · PMI · Archived

PMI Playbook — Buys & Sells · August 2026

The ISM Report On Business is one of the highest-signal macro releases most people never read. What this month’s numbers say, which sectors that favours, and where to read the full report yourself — free.

Archived — August 2026. This is the August 2026 edition, kept at a permanent address. The current one is the latest PMI Playbook.
ISM Report On Business, August 2026 edition
ISM Manufacturing sub-indexes for July 2026, each plotted against the 50 line
Every sub-index in the July report, against the 50 line. Dashed tick is the prior month. Colour reflects whether the reading is good or bad for the economy — which is why customer inventories at 40.7 is green and prices at 71.1 is red.

The buys and sells

From the July report, released 3 August — manufacturing 55.6, a four-year high, with 15 of 18 industries expanding — the framework’s tilt leans harder into the cycle. Research starters, not advice.

▲ Overweight / Buy
  • Industrials 52w high, PMI 55.6
  • Semiconductors +14% off the lows
  • Financials 52w high, steep curve
  • Materials inventories too low
  • Data-center / power capex theme
▼ Underweight / Sell
  • Long-duration bonds still at 52w lows
  • Rate-sensitive housing permits ▾
  • Real Estate contracting
  • Defensives wrong side of the cycle

What the read means

55.6Manufacturing, from 53.3
58.5Production — highest since Nov 2021
52.8Employment, from 49.7
15 / 18Industries expanding
56.7New orders
40.7Customer inventories — too low

Manufacturing at 55.6 is the strongest print since May 2022 and the seventh straight month of expansion. It beat the 54.0 consensus by a wide margin. ISM’s own translation: a 55.6 reading historically corresponds to real GDP growing at about 2.8% annualised.

Three things in the release matter more than the headline:

The awkward line is supplier deliveries at 58.9, up from 57.4. Slower deliveries are usually a demand signal, but they also mean cost pressure — and combined with hot input prices, that shapes the tilt toward cyclicals that can pass those costs on rather than cyclicals generally.

The services print landed — 54.1, and one line in it matters. Services came in at 54.1 against 54.5 expected (54.0 prior) — a 25th straight month of expansion, with business activity jumping to 59.1 and new orders to 57.2. But services employment fell back into contraction at 47.4, down from 51.2, after a single month above 50. So the labour picture is now split: factory employment expanding for the first time in 33 months, services employment shrinking again — and services is by far the larger employer. That makes Thursday’s payrolls print the tiebreaker rather than the confirmation. A soft print would say the manufacturing signal was the outlier.

The verdict

✓ What’s good
  • Employment crossed 50 after 33 months. The longest factory-jobs contraction since the 1980s just ended. Nothing else in the release matters as much.
  • Production at a four-year high with backlogs jumping to 55.0. Firms aren’t clearing old orders — new work is arriving faster than they can build it.
  • Customer inventories at 40.7. Stock levels this thin force restocking. That’s demand you can see coming rather than demand you have to forecast.
  • Breadth, not a spike. 15 of 18 industries expanding means this isn’t one sector flattering the average.
✕ What’s bad
  • Long bonds still at 52-week lows. The one signal that hasn’t joined the party. A cyclical upturn financed at rising long rates is a materially worse trade than one financed at falling rates.
  • Input prices remain hot and supplier deliveries slowed again to 58.9. Slower deliveries flatter the headline PMI while describing a cost problem.
  • Real estate is still contracting, and building permits are trending down. Housing has led every meaningful cycle turn of the last forty years — it is not confirming this one.
  • One month is one month. The employment print reverses 33 months of contraction. Base rates say wait for the second one.
◆ What to keep an eye on
  • Next payrolls. The single confirmation that matters. Factory employment expanding in the survey but not in the hard data would kill the thesis.
  • Whether TLT stops making lows. That would remove the last bearish signal and take the regime score materially higher.
  • Services employment at 47.4 — printed 5 Aug, back in contraction after one month above 50. The labour recovery is now a manufacturing-only claim until payrolls says otherwise.
  • Prices paid rolling over. Falling input costs turn a demand recovery into a margin recovery, which is a different and better trade.
  • Building permits. If they turn up, the one genuine hole in the bull case closes.

Names the framework points at

These follow mechanically from the sub-indexes above — each one is here because a specific number in the report says so, not because it looks good on a chart. They are screens to research, not positions. No entry prices, because the level you pay is your decision and the point of this list is the reasoning.

IdeaWhich number drives itWhat kills the thesis
Machinery & electrical equipment
CAT · ETN · PH
Production 58.5 and backlogs 55.0. These are the businesses that convert a backlog into revenue first. Backlogs rolling back under 50 — order books emptying faster than they refill.
Industrial distributors
FAST · GWW
Customer inventories 40.7. Distributors are the most direct read on restocking — they sell the pick-and-shovel consumables first. Customer inventories climbing back above 48; restocking already done.
Materials into restocking
NUE · FCX · XLB
New orders 56.7 with inventories too low. Volume recovery reaches raw inputs before finished goods. China demand disappointing, or the dollar strengthening hard.
Banks on a steeper curve
XLF · JPM
Long yields at 52-week highs while the front end is anchored. Wider deposit-to-loan spread is mechanical, not narrative. Credit costs rising faster than the spread benefit — watch provisions, not net interest margin.
Data-centre power capex
ETN · PWR · VRT
The one theme where the ISM cycle and the AI capex cycle point the same way. Grid connections are the bottleneck. A hyperscaler cutting capex guidance. This is the theme’s single point of failure.
Long-duration Treasuries
TLT · ZROZ
A 55.6 PMI implying 2.8% real GDP is not a backdrop that rallies 30-year bonds. Payrolls missing badly — growth scare bids duration hard and fast.
Bond-proxy defensives
XLU · XLP
Bought for yield when growth is scarce. Growth is no longer scarce, and the yield is competing with a 52-week-high long bond. The employment print reversing — defensives are the correct hedge if this is a false dawn.
Rate-sensitive housing
XHB · DHI · VNQ
Real Estate contracting in the survey, permits trending down, mortgage rates following the long end higher. Permits turning up, or a genuine dovish shift at the long end.
How to actually use this. The buy list is a bet that the employment print was real and gets confirmed. The sell list is the same bet expressed in reverse. If you take one side you are already implicitly taking the other — so size it once, not twice. And if the next payrolls print comes in soft, both halves are wrong together, which is the honest risk in a single-catalyst framework.
Two honest caveats. PMI is a diffusion / sentiment survey, not hard output — and the market may already have priced it. And before trusting “long cyclicals when PMI is rising” with real money, backtest it. A clean narrative and a positive edge are not the same thing.

Read it yourself — it’s free

You don’t need a terminal or a subscription for any of this. The ISM Report On Business is published free, in full, and most people never look at it. Two places to find it:

The full report

ismworld.org — the month is the last part of the address, so you can jump to any month by changing one word:

ismworld.org/…/ism-pmi-reports/pmi/july/

This is the one to read. It carries all 18 industries ranked in order and the verbatim comments from the purchasing managers themselves — the two things that turn a headline number into an actual view, and the two things every news summary throws away.

The same report, faster

PR Newswire — published free the moment it’s released: first business day of the month for manufacturing, third for services.

Identical headline and sub-index figures, plain text, no sign-up. If you only want the numbers, this is quicker.

What to do with it. Read the manager comments first. Sixty seconds of people describing their own order books tells you more than the headline number does — you’ll see tariffs, shortages and hiring freezes in plain language weeks before they show up in official data. Then check whether the sub-indexes agree with each other: production and new orders rising together is a real expansion, production rising while new orders fall is a company working through a backlog it hasn’t replaced.

If you want a second opinion on the interpretation, TD Economics and PNC both publish free same-day notes and will flag anything odd in the seasonal adjustment before you build a view on it.

Not investment advice. The tilts are framework-generated research starters. Sources: July 2026 ISM Manufacturing Report On Business via PR Newswire; TD Economics; PNC Economics Research.

← All reports