Every month, a single spreadsheet from the US Census Bureau moves currency markets, resets analyst models and generates a wave of headlines that frequently contradict each other. The monthly retail sales release is one of the most widely quoted indicators in American commerce, and one of the most consistently misread. The gap between what the release actually measures and what the coverage claims it proves is wide enough to lead retailers, investors and operators into decisions the data never supported.
This guide walks through the release the way a data analyst reads it: what is in the sample, which vintage of the number you are looking at, why the control group exists, how seasonal adjustment reshapes the headline, and where inflation quietly does half the work. None of it requires an economics degree. It requires knowing which line to read and which caveats travel with it.
In short
- The monthly retail sales report is published by the US Census Bureau as the Advance Monthly Retail Trade Survey, typically in the middle of the month covering the previous month, and it measures dollar sales at retailers and food service businesses rather than total consumer spending.
- The first number you see is an advance estimate drawn from a subsample, and it is revised at least twice before being benchmarked against the annual survey, so the headline you react to is rarely the number that ends up in the historical record.
- The control group excludes autos, gasoline, building materials and food services, and it exists because those four categories are volatile enough to disguise the underlying trend that feeds into GDP consumption estimates.
- The headline series is nominal, meaning it is not adjusted for price change, so a month of rising prices and flat volumes still prints as growth unless you deflate the series yourself using price data from the Bureau of Labor Statistics.
- Read alongside inflation data, personal spending figures and company results, the release is a useful directional signal; read alone as a verdict on consumer health, it is misleading by construction.
What the release measures, and what it leaves out
The report most people mean when they say “retail sales” is the Advance Monthly Retail Trade Survey, known as MARTS, published by the US Census Bureau. According to Census documentation, it estimates total dollar sales for retail trade and food services establishments in the United States, based on a sample of firms that report their sales for the prior month. The Bureau publishes it alongside the fuller Monthly Retail Trade Survey and the Annual Retail Trade Survey, which together form the revision and benchmarking chain described later in this article.
The first thing to internalise is the unit of measurement. The survey collects dollar sales, not units shipped, not transactions, not customers. A retailer that sold the same number of items at higher prices reports a larger number, and that larger number lands in the headline as growth. This single fact explains most of the confusion in any month where inflation is moving.
The second thing is coverage. Retail trade and food services is a large slice of the American economy, but it is not the whole of consumer spending. Services such as rent, healthcare, insurance, travel, education, utilities and professional services are outside the survey frame, and the Bureau of Economic Analysis has long shown that services account for the larger share of household outlays. When a headline says “consumers pulled back”, the honest version is that spending at goods retailers and restaurants moved in a particular direction that month.
What is inside the frame
The survey covers establishments classified under retail trade in the North American Industry Classification System, plus food services and drinking places. That includes motor vehicle and parts dealers, furniture stores, electronics and appliance retailers, building material and garden suppliers, grocery and beverage stores, health and personal care retailers, gasoline stations, clothing retailers, sporting goods and hobby stores, general merchandise chains, miscellaneous retailers, nonstore retailers, and restaurants and bars. Nonstore retailers is the category that carries most e-commerce activity, which matters for anyone tracking the online channel specifically.
If you want the wider structural context for how these categories relate to each other, our explainer on what the retail industry today actually covers maps the same segments to the businesses that operate in them. The classification is not academic. It determines which company results are comparable to which line of the release.
What sits outside it
Business to business wholesale activity is measured separately, in the wholesale trade survey. Most services spending is captured by the Bureau of Economic Analysis in its personal income and outlays release rather than by Census. Purchases of homes, financial products and used items traded privately are not retail sales in this sense. Understanding these exclusions is what stops an analyst from treating one monthly print as a full read on household behaviour, a habit that the broader coverage of how retail news shapes the global e-commerce industry tends to reward rather than correct.
Why the frame matters commercially
For an operator, the practical value of knowing the frame is comparability. If your business sells physical goods online, the nonstore retailers line is closer to your reality than the headline total. If you operate quick service restaurants, food services and drinking places is your line, and it behaves differently from goods categories in almost every cycle. Reading the wrong line and then benchmarking your own performance against it produces a false sense of underperformance or safety.
Advance estimate versus revised data
The number that generates headlines is the advance estimate, and it is the least settled version of the month. Census describes MARTS as being based on a subsample of the firms in the fuller monthly survey, which is what allows the Bureau to publish quickly. Speed is bought with precision, and the trade is explicit in the Bureau’s own documentation rather than hidden.
Each month therefore exists in several vintages. The advance estimate arrives roughly two weeks after the month closes. It is then revised when the fuller monthly survey results arrive, revised again the following month, and eventually benchmarked to the Annual Retail Trade Survey in the Bureau’s annual revision cycle. By the time a month reaches its final form, the growth rate can look materially different from the print that moved markets.
| Vintage | Approximate timing | What changes | Best used for |
|---|---|---|---|
| Advance estimate | About two weeks after month end | First published figure, based on a subsample of reporting firms | Directional read, market reaction, early trend check |
| First revision (preliminary) | The following month’s release | More reporting firms included, late responses added | Confirming or discarding the advance signal |
| Second revision | Two months after the advance | Fuller survey response, corrected reports | Building month-over-month trend series |
| Annual benchmark | Annual revision cycle | Series benchmarked to the Annual Retail Trade Survey, seasonal factors re-estimated | Historical analysis, year-over-year comparisons, modelling |
The practical habit this creates
Serious readers of the release check two numbers, not one. The first is the current month’s change. The second is what happened to the prior month’s figure in the same release, because a strong headline sitting on top of a downward revision to the previous month can mean the level of sales is roughly where it was before. Coverage almost never leads with the revision, which is precisely why it is worth your attention.
A useful discipline is to record the advance estimate the day it lands, then compare it to the same month two releases later. Doing this for six months teaches more about the reliability of the headline than any amount of commentary. In most cycles, the direction survives revision more often than the magnitude does.
Sampling error, stated plainly
Census publishes standard errors and coefficients of variation alongside the estimates, and the Bureau states directly that small month-over-month changes may not be statistically distinguishable from zero. This is the single most ignored line in the entire release. A change of a few tenths of a percent in a month, reported as a decisive move by the consumer, can sit inside the survey’s own margin of error, and the current published error bands should be checked in the release documentation for the month in question rather than assumed.
The control group and why analysts watch it
Buried below the headline is a line usually labelled “retail sales, control group” or “core control”. It removes four categories: motor vehicle and parts dealers, gasoline stations, building materials and garden equipment, and food services. Each is excluded for a specific reason, and once you know the reasons the line stops looking arbitrary.
Autos are large ticket, lumpy and heavily influenced by financing conditions, incentive programmes and supply availability. Gasoline is a price story more than a volume story, because the dollar value moves with the pump price rather than with the number of gallons bought. Building materials track construction and weather. Food services is a category the Bureau of Economic Analysis handles differently in its consumption estimates.
The result is a series designed to feed cleanly into the goods component of personal consumption expenditures. That is why economists who care about GDP nowcasting watch it more closely than the headline. It answers a narrower question, and narrower questions produce more stable answers.
| Series | Excludes | Primary audience | Main weakness |
|---|---|---|---|
| Headline total retail and food services | Nothing within the survey frame | Media, general commentary | Dominated by autos and gasoline swings |
| Retail sales excluding autos | Motor vehicle and parts dealers | Retail analysts | Still exposed to gasoline price moves |
| Retail sales excluding autos and gas | Vehicles and fuel | Sector analysts, retail investors | Still includes volatile building materials |
| Control group (core control) | Autos, gasoline, building materials, food services | Economists, GDP modellers, central bank watchers | Excludes restaurants, so it misses a real part of consumer demand |
| Core retail as defined by trade bodies | Varies by publisher, commonly autos, gas and restaurants | Industry associations, retail press | Definition differs between sources, so figures are not interchangeable |
The last row deserves emphasis. Trade organisations publish their own “core retail” definitions that differ from the Census control group, and both get quoted as “core retail sales” in the same news cycle. Two accurate articles can therefore report different core growth rates for the same month. Checking which definition a figure uses takes seconds and prevents a false contradiction.
Seasonal adjustment in plain language
Retail is violently seasonal. December is enormous, January is small, and nobody learns anything from a report that says holiday shopping was busier than the following month. Seasonal adjustment exists to strip out the pattern that repeats every year so that the remaining movement carries information.
Census applies seasonal adjustment using the X-13ARIMA-SEATS methodology, the standard the Bureau documents publicly. In practice the software estimates the recurring monthly pattern from history, then divides it out. What remains is the part of the month that was not explained by the calendar.
Adjusted and unadjusted are two different questions
The release publishes both adjusted and unadjusted figures, and they answer different questions. The seasonally adjusted series tells you whether this month was better or worse than the season would predict. The unadjusted series tells you how much money actually moved through the tills. An operator forecasting warehouse staffing wants the unadjusted reality. An economist judging momentum wants the adjusted one.
Calendar effects the adjustment cannot fully absorb
Some distortions are not neatly seasonal. The number of weekends in a month, the timing of Easter, a shifted holiday promotion window, severe weather, or a major sales event moving between months all leave residue in the data. Adjustment models handle trading day effects to a degree, but an unusual calendar year still produces months that look stranger than the underlying demand was.
Seasonal factors are themselves revised
A detail that surprises people: the seasonal factors are re-estimated periodically, which means historical adjusted figures can change even when the raw data did not. If you keep a spreadsheet of adjusted monthly numbers and never refresh it from source, your history will silently drift away from the official series. Pull the series fresh from the Bureau rather than trusting a stored copy.
Nominal versus real: the inflation adjustment step
The retail sales headline is a nominal figure. It is measured in current dollars and it is not deflated for price change. In a period of meaningful inflation, this makes the release a poor guide to whether people bought more things, which is usually the question the reader actually has in mind.
The correction is straightforward. Take the nominal series and deflate it using an appropriate price index published by the US Bureau of Labor Statistics. Using the all items consumer price index gives a rough real series. Using category level price indices, for example food at home against grocery sales, gives a considerably better one because price behaviour differs sharply between categories.
Gasoline is the clearest illustration. When pump prices climb, gasoline station sales rise in dollar terms even if drivers buy the same fuel or slightly less, which inflates the headline while telling you nothing good about demand. The same mechanism runs in reverse when energy prices fall, which is why our analysis of why fuel and energy costs can outrank tariffs in retail guidance matters for anyone modelling category level results rather than headline totals.
A simple worked approach
- Take the seasonally adjusted nominal sales figure for the category you care about.
- Find the matching consumer price index series for that category from the Bureau of Labor Statistics.
- Divide the sales figure by the price index and rebase both to the same starting period.
- Compare the resulting real series against the nominal one across at least twelve months.
The gap between those two lines is the story most coverage omits. In months where the two diverge sharply, the nominal headline and the real consumer experience are pointing in different directions, and only one of them predicts unit volumes for your supply chain.
Category detail worth reading beyond the headline
The category table is where the release earns its keep for anyone operating in commerce. The headline is an average of businesses with almost nothing in common, and averages of dissimilar things are rarely actionable. The categories, read individually, describe a set of separate markets moving at different speeds.
Nonstore retailers
Nonstore retailers is the closest proxy the release offers for e-commerce, and it is usually among the faster growing lines. It is not a perfect e-commerce measure, because established retailers book much of their online revenue within their own store categories. For a cleaner online read, Census publishes a separate quarterly e-commerce estimate, which is worth consulting before quoting any online share of retail figure.
Food and beverage stores against food services
The relationship between grocery sales and restaurant sales is one of the more informative signals in the whole release. When restaurant growth outpaces grocery growth over several months, households are typically comfortable spending on convenience and experience. When the balance flips, it often shows up in packaged goods demand and private label share well before it shows up in commentary.
Clothing, electronics and general merchandise
Discretionary goods categories tend to lead the cycle rather than follow it. Clothing and electronics respond quickly to confidence and to promotional intensity, while general merchandise absorbs trade-down behaviour when shoppers consolidate trips. Reading these three together gives a better read on discretionary appetite than the headline ever will.
Gasoline stations as a distortion detector
Treat the gasoline line as a diagnostic rather than a demand signal. If the headline moved and gasoline moved in the same direction by a similar magnitude, check pump prices before writing anything about consumer strength. This one check prevents a large share of the misreadings that circulate every month.
Using the release alongside other retail indicators
No single indicator carries a month. The retail sales release becomes reliable when it is read as one input in a small panel of sources that measure overlapping things through different methods. When several disagree, the disagreement itself is information about what is happening at the margins.
| Indicator | Publisher | What it adds | Cadence |
|---|---|---|---|
| Advance monthly retail sales | US Census Bureau | Dollar sales at retailers and food services | Monthly |
| Personal income and outlays | Bureau of Economic Analysis | Full consumption including services, plus a price index | Monthly |
| Consumer price index | Bureau of Labor Statistics | The deflator needed to convert nominal sales into real sales | Monthly |
| Employment and earnings data | Bureau of Labor Statistics | The income base that funds future spending | Monthly |
| Company reported results | Individual retailers | Actual traffic, ticket, margin and comparable sales | Quarterly |
| State sales tax collections | State revenue departments | Independent transaction based signal at state level | Monthly or quarterly |
Company results are the sharpest cross-check available, because they are audited and they report the operating detail the macro series cannot. A grocery chain reporting its comparable sales tells you what one real business experienced, in one real basket, with real pricing decisions attached. Our coverage of how a major grocer’s quarterly comparable sales bar is set shows how differently a company frames the same demand environment that the macro release summarises in a single percentage.
State level tax collections deserve a mention because they measure taxable transactions rather than surveyed sales, giving a genuinely independent read. They come with their own complications, since what is taxable and who is obliged to collect varies by jurisdiction, a topic covered in our guide to sales tax nexus for online sellers state by state. Used carefully, collections data is a useful sanity check when a monthly survey print looks out of line with what operators are reporting on the ground.
Five common misreadings of the retail sales report
Most errors in retail sales commentary come from a short list of repeatable mistakes. Recognising them is faster than re-deriving the analysis each month.
- Treating the advance estimate as final. The first print is a fast estimate from a subsample, and reacting to it as settled fact ignores the revision chain the Bureau publishes openly.
- Reading nominal growth as volume growth. Without deflating by a price index, a purely price-driven increase reads as consumers buying more, which it is not.
- Calling small changes significant. Census publishes error bands for a reason, and a move within them supports no confident narrative in either direction.
- Confusing retail sales with consumer spending. Services sit largely outside the survey, so the release describes a slice of household outlays rather than the whole.
- Mixing definitions of core. The Census control group and the various trade body core measures exclude different categories, and quoting them interchangeably creates contradictions that do not exist in the underlying data.
A sixth habit is worth adding: ignoring the prior month revision inside the same release. Two consecutive months are frequently reported as separate stories when they are, in fact, one revised picture of the same trend. Reading the release as a whole document rather than as a single percentage change is most of what separates a useful interpretation from a headline.
How professionals timebox the read
An efficient monthly routine takes about fifteen minutes. Check the headline and the control group, check what happened to the prior month, scan the category table for anything moving more than the noise band, check gasoline against pump prices, and then compare the picture against the most recent inflation print. Anything beyond that is modelling rather than reading, and modelling deserves its own time slot.
For teams that track this monthly, keeping a simple log of advance versus revised figures builds an internal sense of how much weight the first print deserves in your particular categories. That institutional memory is more valuable than any individual month’s number, and it is the habit that most reliably improves how a commercial team reacts to macro data, a theme that runs through much of our wider reporting on how retail news shapes commercial decisions.
A note on how to use this information
This article is general information and education about a public statistical release. It is not financial, investment, tax or legal advice, and it does not account for any individual company’s circumstances. Anyone making investment, pricing, tax or compliance decisions based on economic data should consult a qualified professional, such as a licensed financial adviser, accountant or attorney, for their specific situation.
Every methodological detail described here reflects the published documentation of the relevant statistical agencies as of September 2026. Statistical agencies revise their methods, sample designs, category definitions, seasonal adjustment procedures and publication schedules over time. Before quoting a specific figure, threshold, release date or definition, verify it directly against the current documentation from the US Census Bureau retail trade programme or the relevant issuing agency rather than relying on secondary summaries, including this one.
FAQ on the retail sales report
When is the monthly retail sales report released?
The US Census Bureau publishes the advance monthly retail trade estimate in the middle of the month, covering the previous month, typically in the morning Eastern time. The exact date varies each month, so check the Bureau’s published economic indicator calendar for the current schedule rather than assuming a fixed day.
What is the retail sales control group?
The control group, sometimes called core control, is retail sales excluding motor vehicle and parts dealers, gasoline stations, building materials and garden equipment suppliers, and food services. It is designed to feed into consumption estimates used for GDP, which is why economists watch it more closely than the headline total.
Is the retail sales report adjusted for inflation?
No. The headline figures are nominal, measured in current dollars, and they are not deflated for price change. To assess volume rather than value, deflate the series using an appropriate price index from the Bureau of Labor Statistics, ideally at category level rather than using the all items index.
How much do the retail sales numbers get revised?
Every month is revised at least twice as fuller survey responses arrive, and the whole series is benchmarked periodically against the Annual Retail Trade Survey. Direction survives revision more often than magnitude does, which is why the prior month revision inside each release deserves as much attention as the current month’s change.
Does the retail sales report include e-commerce?
Yes, but not in one clean line. Pure play online sellers largely appear under nonstore retailers, while established retailers book much of their online revenue inside their own store categories. For a dedicated online measure, Census publishes a separate quarterly e-commerce estimate.
Why do different outlets report different core retail sales figures?
Because “core” is not a single definition. The Census control group excludes autos, gasoline, building materials and food services, while trade associations publish core measures that exclude a different mix, commonly autos, gas and restaurants. Always check which definition a quoted figure uses before comparing two sources.
Is retail sales the same as consumer spending?
No. The survey covers retail trade and food services, so most services spending, including housing, healthcare, insurance and travel services, falls outside it. Broader consumption is measured by the Bureau of Economic Analysis in its monthly personal income and outlays release.
Can a small monthly change be considered meaningful?
Often not. Census publishes standard errors with the estimates and states that small month-over-month changes may not be statistically different from zero. Check the published error bands for the month in question before treating a change of a few tenths of a percent as a real shift in behaviour.
Which line should a retail operator actually track?
Track the category closest to your own business rather than the headline. An online goods seller learns more from nonstore retailers, a restaurant group from food services and drinking places, and a grocer from food and beverage stores. Benchmarking your performance against a headline built from unrelated businesses produces misleading conclusions in both directions.