The 7% online that counted the clothing stores, not the clothes

An off-white research slide with a green masthead, “Harrow & Vane Retail Property · Research desk” on the left and “Channel mix by store category · 2022” on the right, headlined “Not one store category sells as much as 1 dollar in 14 online. The shop is still where the money is.” A subhead says the chart shows the share of each kind of business’s 2022 sales made through e-commerce, from the U.S. Census Bureau’s Annual Retail Trade Survey benchmarked to the 2022 Economic Census, that every published store category is shown, the axis is linear and starts at zero, each value is printed beside its bar, and the dashed line is the Census figure for all retail trade, 13.9%. The main panel, “E-commerce as a share of sales, by kind of business · 2022”, is a horizontal bar chart on an axis from 0% to 15% with six green bars, sorted: Clothing & clothing accessories stores 7.0% ($19.8B of $281.0B), Sporting goods, hobby, musical instrument & book stores 6.0% ($5.6B of $94.3B), Furniture & home furnishings stores 3.4% ($4.6B of $136.8B), Electronics & appliance stores 3.0% ($2.6B of $87.0B), Food & beverage stores 2.9% ($27.2B of $942.3B) and Building materials & garden equipment dealers 0.8% ($4.0B of $527.4B). A dashed orange line at 13.9% is labelled “All retail trade”, well to the right of every bar. Three cards sit on the right. “The headline number” reads “6 of 6 categories below the all-retail rate” and notes that the most online category, clothing & accessories, does 7.0% of its sales through e-commerce, about half the all-retail rate, while building materials & garden does 0.8%. “Read-out for the desk” says that across every category the firm leases to, about 93 cents in every dollar is still taken somewhere other than an online order, that even apparel is 93% offline, and recommends holding the in-line apparel and home-furnishings allocation at current weights for the 2027 acquisition cycle. “Basis of measurement” cites the Census table, says the categories the Census Bureau withholds (D or S: motor vehicles, health & personal care, gasoline, general merchandise and miscellaneous stores) are not shown, and says nonstore retailers (72.7%) are not shown because they are a channel, not a store category. Five ticks along the bottom read “One source, one year”, “Zero-based linear axis”, “Every published store category”, “Every value printed” and “National benchmark drawn on”. A footnote credits the Census Bureau’s 2022 benchmarked Annual Retail Trade Survey tables, says the shares were computed from the published dollar figures, and says the slide is a demonstration drawn by Misleading Charts: Harrow & Vane, the desk and the recommendation are invented, and every input figure is the Census Bureau’s. — Showing the misleading chart

A cream page, kicked “The same Census tables · the rows the slide left out · one ruler in dollars” and headlined “85% of U.S. retail e-commerce is filed under the row the slide set aside.” A standfirst says every bar on the slide is right but what it measures is not what it says: the Census Bureau files each store under its main line of business and every dollar that store takes goes with it, so a row called “clothing stores” counts the stores, not the clothes; jeans bought from a web-only retailer, or from a chain’s separately run web arm, are filed under electronic shopping and mail-order houses, part of the nonstore row the slide set aside, and that row holds $827.9B of the $969.4B of 2022 retail e-commerce, which is why all six categories can sit below the national 13.9%. The top panel, “Where the $969.4B of 2022 retail e-commerce is filed”, is one bar drawn to scale: a long gold segment for electronic shopping & mail-order houses, $827.9B, 85.4%; other nonstore retailers, $2.0B, too thin to draw; a short green segment for the six store rows on the slide, $63.8B, 6.6%; and a grey segment for withheld or suppressed store rows, $75.7B, 7.8%. The main panel, “Online sales of the same merchandise, in the two rows that hold it · 2022, $ billion”, pairs a green bar for the online sales of each store row the slide drew with a gold bar for the online sales of that merchandise by web and mail-order retailers, on one zero-based axis from $0 to $140B: Clothing & accessories $19.8B and $129.8B, 6.6×; Furniture & home furnishings $4.6B and $124.6B, 27×; Electronics & appliances $2.6B and $81.6B, 31×; Sporting goods, hobby & books $5.6B and $79.4B, 14×; Food & beverages $27.2B and $36.6B, 1.3×; Building materials & garden $4.0B with a note that there is no matching merchandise line because hardware and garden goods sit inside “Other merchandise”. Four cards run down the right. “The tell was on the slide” says an average cannot sit above every one of its parts, so when it does the total contains something the parts don’t — here a single row, nonstore retailers, 72.7% online and carrying 85.6% of all retail e-commerce. “What the slide actually measured” says it measured how much of a store’s own sales come in as online orders, which across every store-based kind of business is 2.4%, $139.5B online of $5,832.5B, a useful number about shops (chains that run their web sales as a separate business are filed under 4541) but not the share of clothing, furniture or electronics people buy online. “Furniture, read both ways” repeats the paired figures. “What these tables can’t give you” says a true share of clothing bought online needs clothing sales across every kind of store, department stores and warehouse clubs included, which the survey does not publish, so the honest move is to say the tables stop short. A footnote names both Census tables and the retrieval date, explains that the sporting/hobby/book pair adds the merchandise lines for sporting goods ($31.9B), toys, hobby goods & games ($27.3B) and books ($20.2B), lists the withheld or suppressed rows, notes that the 4541 electronics line excludes the suppressed computers-and-phones line and that the rows follow NAICS 2017, and says both drawings are ours, every input figure is the Census Bureau’s and the sums, residuals and shares are ours. — Showing the misleading chart

Be the first to star this exhibit

A property-research slide we drew from the Census Bureau’s retail survey shows six store categories, every one selling 7.0% or less online in 2022 against a national 13.9%. The rows are kinds of store, not kinds of goods: web and mail-order retailers are filed in a row of their own, set aside on the slide as “a channel, not a category”, and that row holds $827.9 billion of the $969.4 billion of 2022 U.S. retail e-commerce.

U.S. Census Bureau, Annual Retail Trade Survey, 2022 benchmarked tables: “Estimated Annual U.S. Retail Trade Sales — Total and E-commerce, 1998–2022” and the merchandise-line table for electronic shopping and mail-order houses (NAICS 4541), total and e-commerce sales, 1999–2022. The slide, the firm and both drawings are ours., 2024Added October 18, 2026

01The claim

Even in the categories everyone assumes have gone online, shoppers still buy in person: not one store category made as much as 7.1% of its sales online in 2022, against 13.9% for retail as a whole.

02The trick

Every bar on the slide is the Census Bureau’s own figure, and the arithmetic is right: clothing and accessories stores made $19.8 billion of their $281.0 billion of 2022 sales online, 7.0%, and furniture stores $4.6 billion of $136.8 billion, 3.4%. The catch is the word under each bar. The Annual Retail Trade Survey, like most business statistics, is counted by establishment: each retailer is given one industry code for its main line of business, and every sale it makes goes into that row. So “clothing and clothing accessories stores” means sales by clothing stores, of anything, and leaves out every pair of jeans sold by a department store, a warehouse club or a web-only retailer. Under the NAICS 2017 codes these tables use, web and mail-order retailers have a code of their own, electronic shopping and mail-order houses — which also takes in chains’ separately run web arms — which sits inside the nonstore row the slide set aside as “a channel, not a store category”. That row is where the online money is: $827.9 billion of the $969.4 billion of 2022 retail e-commerce, 85.4%, against $63.8 billion, 6.6%, for all six rows on the slide together. The slide even shows the tell. An average cannot sit above every one of its parts, so six categories all below the 13.9% line means the total holds something the categories don’t, and here it is one row. Read properly, the chart is answering a narrower question: how much of a store’s own sales come in as online orders. (Harrow & Vane, its research desk and its recommendation are our invention; every input figure is the Census Bureau’s, and the sums and shares are ours.)

03The fix

The redraw puts the two rows that hold each kind of goods side by side on one dollar axis. The Census Bureau publishes what web and mail-order retailers sell by merchandise line, and next to the store rows the slide drew it changes the picture: furniture stores sold $4.6 billion online and web retailers sold $124.6 billion of furniture and home furnishings online, 27 times as much; electronics stores $2.6 billion against $81.6 billion, 31 times (a conservative pairing, since the web retailers’ electronics line leaves out computers and phones, which the Census suppresses); clothing stores $19.8 billion against $129.8 billion of clothing, 6.6 times. Only groceries stay close, at $27.2 billion against $36.6 billion. A single bar above it shows where all $969.4 billion of 2022 e-commerce is filed. The redraw keeps the slide’s real finding and gives it its proper name: store-based retailers took 2.4% of their sales online in 2022, $139.5 billion of $5,832.5 billion, and that is a useful number about shops, if a low one, since chains that run their web sales as a separate business are filed with the web retailers. What the tables cannot give is the share of all clothing bought online, because that needs clothing sales from every kind of store, and the survey does not split them that way. When a product-level claim needs a product-level table that doesn’t exist, say so rather than borrowing a store-level share. And label the bars with what they count: “clothing stores”, not “clothing”.