Trade Data Revisions: A Procurement Planning Control
“A statistic without its release date can become a different number next month. Record the version before you record the decision.”
What does the MTIS report combine?
The current MTIS release builds its estimates from three Census surveys: monthly retail trade, monthly wholesale trade, and manufacturers’ shipments, inventories and orders, known as M3. Its adjusted figures account for seasonal variation and trading-day differences, with holiday variation adjusted for sales only, and none of them are adjusted for price changes; the release also carries unadjusted tables. For a buyer, a rise in a sales or inventory figure can therefore reflect prices as well as physical volume, and the report alone cannot separate the two.
One release can also mix revision states. According to the release notes, wholesale and manufacturing data are carried unrevised from the latest component reports, while retail data are revised and shown in more detail than in the advance indicators report. Its tables label preliminary figures separately from revised ones, and the seasonal factors are applied to the current and previous month and, for retail and wholesale, to estimates a year earlier as well. A combined inventory-to-sales ratio taken from one release can therefore rest on component estimates at different stages of revision, and the label belongs in any record that reuses it.
Why do the component series revise on different dates?
The MTIS revision notice sets out the sequence. Retail revisions for sales and inventories were released on September 28, 2026 and are reflected in the latest report. Wholesale revisions are scheduled for October 26, 2026 and are due to appear in the September 2026 MTIS release planned for November 17, 2026. The notice attributes the revisions to historical corrections and to results from the 2023 and 2024 Annual Integrated Economic Survey. Until the wholesale release appears, its date is an announcement, not evidence that wholesale history has already changed.
Manufacturing follows a third path. The same notice states that, because of schedule and resource constraints, revised historical M3 data and the M3 seasonal adjustment models will remain unchanged for the remainder of 2026. A team comparing manufacturing shipments or inventories with retail or wholesale series this year is setting an M3 history that is held fixed beside retail estimates that have already been revised and wholesale estimates that are scheduled for revision.
Each component also revises on its own routine. For wholesale, the Census methodology says preliminary estimates are published approximately 40 days after the data month and revised approximately 30 days later. Monthly wholesale estimates are benchmarked to annual survey estimates that reflect the latest Economic Census results, and seasonal adjustment uses the X-13 ARIMA-SEATS program. Each step can change a value a planner has already copied into a forecast.
What can change the basis of a series, not just its latest month?
Annual benchmarks and classification changes reach further back. The AIES program page says the Annual Integrated Economic Survey replaced and integrated seven annual business surveys in 2023 and now serves as a benchmark for indicator programs, including retail, M3 and wholesale. The retail methodology records that estimates were restated to the 2017 NAICS beginning with the 2025 benchmark report released on April 25, 2025, and that nonemployers were removed from the time series as of the April 2025 annual revision to align with the AIES.
Sample redesigns can also restate a series. Census says wholesale sample revisions are performed approximately every 5 to 7 years and that, in most cases, estimates from the prior and new samples have been put on a comparable basis. The wholesale sampling frame also excludes manufacturers’ sales branches and offices, so wholesale inventories describe merchant wholesalers rather than every stock position in a supply chain. Recording which sample and classification a value came from is one more reason to keep the release with the number.
What should a procurement data-use register record?
The control is simple: never store a Census value without the version that produced it. The register below is our method for doing that. It records enough to rerun a decision when a later release changes the input, and it keeps announced revisions separate from observed ones.
| Register field | What to record | Why it matters |
|---|---|---|
| Series and component | Program (MTIS total, retail, wholesale or M3), industry line and unit | Components revise on different schedules |
| Release vintage | Release name, publication date and retrieval date | The same month can carry different values in later releases |
| Value as used | The exact figure copied, with a saved copy or extract of the release | A release URL may not identify a vintage on its own |
| Observation period | The data month or year the value describes | Keeps release dates and reference periods apart |
| Revision status | Advance, preliminary, revised or benchmarked, using the release’s own label | Separates first estimates from later ones |
| Adjustment basis | Seasonally adjusted or not, and whether values are in current dollars | Adjusted history moves as seasonal factors update; current dollars mix price and volume |
| Announced changes | Scheduled revision dates and frozen models, marked as announced | An announcement is not evidence of a changed value |
| Decision bound to it | The forecast, buffer setting, index clause or category plan that used the value, and the change that would alter it | Shows which decisions to revisit and when a revision matters |
| Refresh owner and date | Who rechecks the value, and when, keyed to the Census release calendar | Makes the next review deliberate |
An authored control, not a Census product. It sets no materiality threshold and does not predict the size or direction of any revision.
Use the register at two moments. When a value enters a plan, log its row and decide in advance what size of change would alter that decision. When Census publishes a revision, compare the new vintage with the logged one against that threshold. Limit the routine to categories where a public series actually informs a contract, a buffer or a forecast, and read it alongside your supplier risk assessment so a data change and a supplier change are judged together.
Which buyer records does a Census series not replace?
- Purchase order lines, receipts and prices for the categories the series is meant to inform.
- Supplier lead times, capacity statements and delivery performance.
- Plant and warehouse locations, including stock held outside the domestic merchant wholesale population.
- The contract clause, model or calculation that uses the public value, with its own effective date.
Each of these records is specific to a buyer and its suppliers; national estimates are not. A national ratio can trigger a review of a category, but it should not feed a safety-stock formula directly, and it cannot confirm a supplier’s inventory, a lead time or a price. If a contract index refers to a revisable series, the clause should name which vintage governs and how later revisions are treated. Where a supplier relationship depends on shared forecasts, agree in writing which release vintage both sides will use and when it will be refreshed.
Why treat a revision calendar as a planning control?
Sources
- Manufacturing and Trade Inventories and Sales — U.S. Census Bureau, 2026. Current empirical evidence (official report): MTIS composition, staggered retail and wholesale revision schedule, unchanged M3 history and seasonal models.
- Manufacturing and Trade Inventories and Sales, August 2026 — U.S. Census Bureau, 2026. Current empirical evidence (official report): Composition, adjustment basis, component revision status inside one release, preliminary and revised labels.
- Monthly Wholesale Trade: Monthly Methodology — U.S. Census Bureau, 2026. Foundational evidence (official report): Preliminary-to-revised timing, benchmarking and seasonal adjustment for wholesale estimates.
- Monthly Retail Trade Survey: How the Data Are Collected — U.S. Census Bureau, 2026. Foundational evidence (official report): NAICS restatement and removal of nonemployers from the retail time series.
- About the Annual Integrated Economic Survey (AIES) — U.S. Census Bureau, 2026. Foundational evidence (official report): AIES as the annual benchmark for retail, wholesale and manufacturing indicator programs.
- Monthly Wholesale Trade Survey: Sample Revision FAQs — U.S. Census Bureau, 2026. Foundational evidence (official report): Sample redesign cadence and Census's statement that prior and new samples are put on a comparable basis where possible.