UK retail sales index food stores: the grocery aggregate explained, and how it arrives

The query sounds like one series and lands as a family of them. Inside Britain's official read on consumer spending - the Office for National Statistics' Retail Sales Index - the grocery cut carries its own aggregate, Predominantly Food Stores (AGG1): retailers whose main line is food, measured as average weekly sales in value and volume terms, seasonally adjusted and non-seasonally adjusted, on a 2023=100 base.

Two named products carry it. The UK ONS Retail Sales Index - Retail Industry is the full matrix - 622 series keyed by four-character CDIDs, drawn from a monthly survey of roughly 5,000 retailers representing about 75.7% of known retail turnover - and the UK ONS Retail Sales Index - Reference Tables resolves the grocery block explicitly: AGG1 holding GBP 201,285 million of 2023 sales at a 37.57 weight, supermarkets inside it as non-specialised food stores under SIC 47.11 (GBP 184,104 million, weight 34.57), specialist food stores - butchers, bakers, greengrocers - as AGG26 (GBP 13,803 million, weight 2.26), and Alcoholic Drinks, Other Beverages and Tobacco (AGG27) reported beside them (GBP 3,378 million, weight 0.74).

The chained-volume aggregate answers to CDID EAPT, which doubles as the join key between the matrix and the reference tables. On Datadory's rubric the matrix scores 9 out of 10 against a 7.81 catalog mean across 1,744 datasets.

One boundary worth stating once rather than absorbing silently: the geography is Great Britain - England, Wales and Scotland - with Northern Ireland and the islands outside the survey frame, so any panel labelled "UK grocery" built on these series inherits that perimeter.

Read straight off the July 2026 edition, laid out as the rows arrive:

table : CPSA # value of retail sales at current prices, % change year-on-year
2026 May : Predominantly_Food_Stores +1.3 | Non_specialised_Food +1.9 | Specialist_Food -6.1 | Alcohol_Tobacco -1.7
2026 Jun : Predominantly_Food_Stores +0.7 | Non_specialised_Food +1.0 | Specialist_Food -4.9 | Alcohol_Tobacco +7.1
2026 Jul : Predominantly_Food_Stores -0.6 | Non_specialised_Food -0.9 | Specialist_Food +0.2 | Alcohol_Tobacco +10.6

# column header metadata riding with every series
Predominantly_Food_Stores: sales_2023 GBP 201,285m | weight 37.57 | code AGG1 | identifier EAPT

Three months, four different grocery stories. In July 2026 the food aggregate fell -0.6% year on year while total retailing including fuel slowed from +3.8% to +1.6% and predominantly non-food stores cooled from +7.0% to +3.8% - groceries decelerating harder than goods. Beneath the aggregate the split argues with itself: supermarkets down -0.9%, specialist food stores turning positive at +0.2% after two deep negatives (-6.1% in May, -4.9% in June), and the small alcohol-and-tobacco category swinging from -1.7% in May to +10.6% in July.

That dispersion is precisely why the store-type breakdown exists. An all-retailing headline flattens every one of those moves into a single number; the AGG1 family keeps them apart.

Deeper than almost anything else monthly in the catalog. Monthly observations begin January 1988, annual reference tables reach the same year, and quarterly cuts open at 1996 Q1 - roughly 38 years of grocery-demand history spanning every downturn, the e-commerce transition and dozens of complete seasonal cycles. The internet era has its own seam: internet sales by store type start in November 2006, and the all-retailing internet-share ratio J4MC has run monthly since January 2007.

These are first estimates, revised as later responses land - summary, quality and revision-triangle material travels with every edition. One episode belongs in any model's memory: the August 2025 edition slipped to September 2025 while a seasonal adjustment error affecting January-May 2025 was corrected, so extracts taken before the correction disagree with today's figures.

Datadory handles that the unglamorous way. Every value ships stamped with its reporting period and vintage, so a restated number reads as a new edition of the same series instead of a contradiction, and cross-vintage comparisons become analysis rather than archaeology.

Four, and each one punishes naive handling.

1. The accounting calendar. Monthly periods consist of four weeks except March, June, September and December, which have five - so raw month-on-month moves mix trading-period effects with real demand. Compare like months year on year, or take the pre-computed growth rates instead of differencing the indices yourself.

3. Adjustment basis. Seasonally adjusted and non-seasonally adjusted series sit side by side, and differencing across that line produces noise that looks like signal. The all/large/small business-size splits exist only in the unadjusted tables, so size-class work pins you to a specific basis.

4. Weighting discipline. Every column publishes its 2023 sales value and percentage weight against the 2023=100 base, so contribution arithmetic comes built in - AGG1's 37.57 weight means a one-point food move shifts all-retailing by roughly 0.38 points. Weights differ between tables, so confirm which weighting applies before rolling sub-sectors into a custom composite. That single habit prevents most aggregation errors people blame on the data.

Three official instruments bracket the question, and they measure different things in different units.

The US Census Monthly Retail Trade Survey (MRTS) reports monthly US sales in millions of dollars by NAICS kind of business from January 1992, with the 445 food-and-beverage-store family resolved beneath the headline line, adjusted and unadjusted side by side; the July 2026 advance put total retail and food services at USD 763.6 billion. The Eurostat Data Browser - Apparel, Textiles & Trade indexes European momentum instead: NACE G47 net-turnover and volume-of-sales series across 42 European geographies from January 1991 in three adjustment states, with HICP food-price series cut to COICOP subclasses such as bread and cereals and meat, verified current to June 2026.

Three mismatches need handling before any cross-border chart: the UK series run as indices on a 2023=100 base while MRTS prints dollar levels and Eurostat indexes turnover to 2021=100; Great Britain is not the United States and neither is the EU-27; and the classifications disagree - AGG codes with SIC 2007 divisions on one side, NAICS kinds of business in the middle, NACE G47 nodes on the other. The second table below lines the three instruments up.

The index measures spending, not shelves. It carries no product-level prices, no assortment, no store locations and nothing about what is actually in the basket. Those gaps sit in named records on the same shelf:

Demand index out, composition and access and price texture in - that pairing is the whole grocery stack, and all of it arrives under one contract in the food-retail pool.

Ranked by how directly the AGG1 family answers the day job:

1. Market researchers and consultants size grocery categories and test whether spend is shifting between food and goods using the aggregate against its own splits; the workflow continues in market researchers use cases and market-sizing.
2. Investors and quant researchers trade the monthly grocery line as a consumer-staples signal and hold the revised history as ground truth; see investors & quants and quant-backtesting.
3. Data scientists and ML engineers build nowcasting features from 38 years of seasonally adjusted store-type panels keyed on stable CDIDs; see data scientists use cases and demand-forecasting.
4. E-commerce and grocery operators benchmark like-for-likes against the sector index and track channel mix from the internet splits; see e-commerce operators.
5. Journalists and academics cite an accredited official statistics series whose methodology, weights and revision record publish beside the values - citation-grade by construction.

Because the number was never the hard part - the packaging is. Three column schemas live inside a single edition: the store-type layout, a collapsed four-commodity-group view, and the growth sheets with their own format. Row-one labels are really column names. Weights ride in header rows waiting to be promoted into fields. Six hundred-plus series carry identifiers doubling as join keys. Accounting months stretch to five weeks four times a year. And the first-estimate regime revises its own past.

Datadory normalizes all of it upstream: typed rows keyed on stable CDIDs such as EAPT, weights and 2023 sales values promoted into their own fields, adjustment basis declared per series, and vintage stamps on every figure so consecutive editions diff cleanly. When a new month lands, it arrives as new rows under the same dictionary - no re-integration project.

Name the aggregates, the splits and the periods when you request a sample - the AGG1 trio back to 2005, supermarkets versus specialists across the last decade, the alcohol-and-tobacco swing months, or the value-volume-deflator trio for a single quarter - and the extract comes back cut to that scope, with the complete field dictionary and coverage statement attached. You keep all three regardless of what happens next.

That is the point of sample-first: the thing you evaluated is the thing that ships, drawn from the same delivery path as the standing feed rather than a fenced-off demonstration surface. Fields beyond the verified core get confirmed cell-by-cell against your nominated scope before the pipeline locks naming for your side.

Start with the dataset pages behind the figures - UK ONS Retail Sales Index - Retail Industry, UK ONS Retail Sales Index - Reference Tables and UK ONS Retail Sales Index - Automotive Fuel and Non-Store Time Series - each carrying sample rows, a field dictionary and coverage chips.

For how the grocery cut fits the wider stack, open the Food Retail Data Guide and the food-retail data hub; the scorecard lives at best food-retail datasets.

Reading next: UK retail sales index monthly data walks the monthly grain series by series, monthly grocery sales data by NAICS covers the American equivalent, and EU food price statistics (HICP) follows the European price half. The definitions underneath the series sit in the glossary entries for retail sales index, ONS CDID and implied price deflator.