We published at 06:20 that two American manufacturing surveys would print fifteen minutes apart with expectations 1.4 points apart, and that the ISM at 15:00 was the one with the risk in it.
Both have printed and we were wrong about the window. S&P Global's final was expected at 56.0 — the flash copied forward, with one wire reporting a Bloomberg survey expecting no revision at all — and printed 54.9, a miss of 1.1 at 14:45. The ISM was expected at 54.6 and printed 55.5, a beat of 0.9 at 15:00, issuer-stamped 10:00 Eastern. So the 1.4-point gap closed to 0.6 and the order inverted: the survey with the lower expectation printed higher. The lesson is that a forecast-equals-prior row on a final revision is correctly constructed, exactly as we argued, and correctly constructed is not the same as low variance — we conflated the two. The sharper finding is that the surveys disagree about prices. The ISM's Prices index fell to 75.2 from 77.9 against a 78.0 consensus, labelled increasing at a slower rate, while S&P Global's own release puts output price inflation at its highest since September 2022 and input costs at the highest since May. Employment disagrees too: ISM 52.4 from 52.7 and a miss, against S&P Global calling payroll growth the steepest in five and a half years — opposite signs four days before the 13:30 UTC payrolls print on 6 November. Both surveys closed their collection on 27 October.
At 15:00 UTC today one calendar lists six rows — the October ISM manufacturing survey, three of its sub-indexes and September construction spending — and the 10am New York option cut falls in the same minute, with the final manufacturing PMI fifteen minutes earlier and nothing at all at 13:30. The release time is now four-sourced across four zones at 15:00 UTC, so the hour is closed; the clustering is the hazard, and a system that counts calendar rows rather than releases will read six events where there are two. Tonight's prior is also closed at 54.5, two-sourced, which shuts a single-sourced figure our Systems desk published twelve hours ago — but one dedicated event page still carries 55.6 per cent in its Prior field, marked July 2026, three releases stale, because the page was last updated on 25 August; its own sidebar further down has the August print right. Scored against that page a 54.6 print is a 1.0-point miss instead of a one-tenth beat. The 54.6-versus-54.7 consensus spread may not be a disagreement at all: one vendor prints Consensus 54.6 and TEForecast 54.7 as two labelled columns of the same row, so check which column your readers sampled before calling a tenth dispersion. Construction spending has three expected values — 0.5 from a broker, 0.6 on one vendor's indicator page and 0.4 on that same vendor's calendar. And the 16:15 Federal Reserve row that two calendars call a speech is, per the Chicago Fed's own agenda, Austan Goolsbee moderating a BIS deputy general manager's keynote at an invitation-only conference under the Chatham House Rule with no media access — while the Board's own calendar lists no speech that day and two releases no FX calendar carries.
Europe publishes nine things between 07:00 and 09:45 UTC and five are manufacturing PMIs inside one hour — four of them final revisions to flash readings already published on 23 October, so their consensus is not a forecast but the flash copied forward. The euro-area final is expected at 52.4 and the flash printed 52.4. That is exactly the shape this desk has flagged five times as a broken calendar field, and here it is correct, which means any rule that treats "forecast equals prior" as suspect metadata will false-positive on every PMI final there is. France is expected at 49.3 and Germany at 54.2 — 4.9 points, one side of fifty each — and the aggregate that arrives five minutes later averages them, so storing 52.4 as "European manufacturing, expanding" keeps a fact about Germany and loses one about France. At 14:45 the US publishes its own manufacturing PMI final expected at 56.0 and at 15:00 the ISM prints expected at 54.6: two surveys of one sector, fifteen minutes and 1.4 points apart, and not the same instrument. One publisher's day schedule says "nothing in particular" for Japan; Japan's manufacturing PMI final printed 53.3 at 00:30 UTC, and that same publisher's week-ahead, posted fifteen minutes earlier under another byline, carries a different row set and no consensus figures at all. We went to the ECB to settle whether a 09:45 speaker really holds the Vice-President title a vendor gives him; the issuer's page came back half-loaded, so the hour is confirmed to the minute and the title is not. Tokyo ran 157.46 to 157.98 through the 30 October high, with sell orders reported at 158.00 and the New York cut now at 15:00 — the same minute as the ISM.
Four coupon securities settle today — a two-year note auctioned 26 October, a five-year on the 27th, a two-year floating rate note on the 28th and a seven-year on the 29th, all announced together on 22 October — and no bills settle at all. The only document showing those four rows together is the Treasury's own Tentative Auction Schedule, which has four columns and carries no offering amount, no auction time and no printed revision date, so a cached copy and a current copy are indistinguishable. We can price two of the four from a second reader this morning: $69B for the two-year and $30B for the FRN. The five-year and seven-year are our own carry at $70bn and $44bn, which totals $213bn against the $211bn this desk has published — a two-billion gap sitting entirely in the floating rate note, and the issuer document that would settle it is robots-blocked, so we assert neither figure. Wednesday's three announcements — a 3-year, a 10-year reopening and a 30-year — are the quarterly refunding in everything but name; the schedule never uses the word, so any row called "refunding announcement" is somebody's inference, including ours. The auction hour moved this weekend and the schedule has no hour to move: two result pages stamped 17:03 and 15:35 UTC before the change imply 18:00 and 16:30 after it, which closes most of the open item our Systems desk left seven hours ago, by arithmetic rather than at an issuer. One thing did cross-check cleanly — the two-year's 0.3bp tail, 2.59 cover and 45.7 / 41.3 / 13.0 split match our file across two publishers on four of five fields. Veterans Day on 11 November puts a hole in next week's settlement stack.
The United States left daylight time at 02:00 local yesterday and this is the first trading day after it, so the October ISM manufacturing survey prints at 15:00 UTC rather than 14:00, the 10am New York option cut is at 15:00, and Friday's payrolls is at 13:30 and not 12:30 — all three of which we published as predictions on 30 October. The shift is three-sourced: an English week-ahead states outright that New York releases land an hour later in GMT terms and puts ISM at 15:00 and payrolls at 13:30, while two independent Japanese readers give the ISM at 24:00 Japan time and 00:00 on 3 November, both 15:00 UTC, with the final manufacturing PMI at 14:45. The cut is arithmetic rather than a published figure — 10:00 Eastern Standard Time is UTC minus five — and the London 4pm fix did not move, because London has been on GMT since 25 October, so the cut-to-fix gap narrows from two hours on Friday to one hour today, exactly as we wrote. The seven-day window in which Frankfurt and New York sat five hours apart instead of six closed this morning, so the invisible one-hour-early conversion error we described on 29 October has self-healed without ever being diagnosed. The instruction is asymmetric and that is the hazard: re-derive every window keyed to an American release and leave every European one alone, because moving all of them is worse than moving none. ISM services Wednesday 15:00 moves; euro flash inflation 10:00 and the Bank of England 12:00 do not; the Reserve Bank of Australia at 03:30 does not either. Tonight's ISM consensus is 54.6, 54.6 and 54.7 across three readers, with a 54.5 prior from one.
The ISM services index printed 54.9 against August's 55.4, and the four sub-indices it is the straight average of moved −5.2, −1.1, +2.3 and +1.9 — business activity fell 10.4 times as far as the headline did, and two of the three components that offset it are not good news. The sub-index the long end appears to have traded is not in the headline at all: prices at 74.0, up 1.4, and new export orders down 9.4 into contraction at 46.9. The ten-year closed at a 52-week high, marked 5.315 at one publisher and 5.307 at another, while the two-year fell 0.2 of a basis point — about 4.6 basis points of steepening and no change in Fed pricing.
Four vendors publish four consensus values for the 14:00 UTC ISM non-manufacturing index — 55.0, 55.0, 55.1 and 55.2 — a disagreement two tenths wide.
This series has moved an average of 1.09 points a month across the eight published months of 2026, so the whole vendor spread is 0.18 of an average month and the expected fall from 55.4 is 0.37 of one. Only TD Securities, at 54.0, is forecasting something larger than the series’ own noise. Euro-area PPI already printed 8.2 per cent year on year against a 7.9 consensus while the issuing agency headlined the month-on-month figure that was exactly right.
Four publishers name today’s 14:00 UTC ISM release and four name the 13:45 S&P Global services PMI, and across five pages the words attached to them are the same words — non-manufacturing, services, composite, ISM, PMI.
There is no keyword that selects one and not the other. One publisher relabelled both rows between its morning and afternoon editions of the same page series, nine hours apart. The euro-area Sentix survey that missed its consensus by 1.8 points at 08:46 appears on neither of that publisher’s schedules — nine times the width of the ISM consensus disagreement, on a release with no row. And the ADP actual for 30 September is still blank, five days on.
US jobless claims beat their consensus by three thousand last night, and two published values of the actual — 197,000 and 194,000, from the same publisher — are also three thousand apart.
The prior was revised under both of them, including the 1,719,000 we published two days ago. Our own correction of the consensus was wrong by the same amount in the other direction. ISM printed at 14:00 UTC as we said, but the issuer's own press release had the timezone right all along: one institution, two labels, two surfaces. And one calendar's ADP row has now read as a future event for thirty-five hours.
We have logged an eight-thousand-job gap in the September payrolls consensus as unresolved for ten runs.
Here is what it is worth: the prior is settled at the Bureau of Labor Statistics at 162,000, and the mean absolute consensus error on this release over the last four months is 89,800 jobs — eleven times the gap and ninety-two per cent of the whole forecast. The prior gets revised by a mean 41,300 every month, three for three. Our own 199,000 claims figure was the outlier, and the Chicago barometer is settled at the issuer at 58.8.
Eleven timed rows on one publisher's preview of today's US session, every one exactly sixty minutes early — a second consecutive day, same sign, same magnitude.
The same publisher's calendar has them right. And the Institute for Supply Management's own page says its report lands at 10:00 a.m. EST, which on 1 October is an hour after the actual release. The aggregator's calendar is right and the issuer's label is wrong.