The July retail miss is a Prime Day artifact, not a consumer crack

Retail and restaurant sales in July fell 0.6% from June on a seasonally adjusted basis, to $763 billion, while the unadjusted figure that measures money actually spent rose 0.9%, to $784 billion ^[1][2]. That is a 1.5 percentage point spread on the same month of the same data ^[16], and anyone who repriced demand assumptions off the adjusted headline responded to the placement of Amazon Prime Day rather than to a change in consumer behaviour.

Start with the calendar. Prime Day fell in June this year and in July last year ^[3]. In June 2026, ecommerce sales rose 2.4% from May and 16% year over year, to $138 billion unadjusted ^[4]. In July they came off that peak, down 2.0% to $135 billion, still above May, and up 6.2% year over year ^[5] - measured against a July 2025 that contained the event ^[3]. The give-back was also smaller than last year's: ecommerce fell 4.4% in the month after the 2025 event ^[6] versus 2.0% after the 2026 one, less than half the decline ^[7]. Ecommerce is now the largest retail category at 18% of total sales on a 12-month basis, including the online operations of Walmart, Costco, Target and Macy's, so its timing quirks move the aggregate ^[8]. Wolf Richter, whose reading of the Census Bureau data this is, concludes that consumers did not cut back and that the seasonal adjustments made a calendar problem worse ^[19].

The discretionary tell is restaurants. Sales at eating and drinking places rose month to month and 6.0% year over year unadjusted ^[9]. That is the first line households trim under stress, and it is not being trimmed. Breadth held elsewhere too: miscellaneous store retailers up 11.6%, building materials up 5.9%, clothing up 5.4%, general merchandise up 4.5% ^[10]. Even after the adjustment machinery, the year-over-year gain was 5.0% ^[1].

Autos are where the two versions of July diverge most. Unadjusted, dealer sales rose 0.6% from June and 2.0% year over year, to $135 billion; seasonal factors removed $6 billion, yielding $129 billion and a 2.0% monthly decline ^[11]. At the same time, CPI for used vehicles fell 0.2% on the month and 1.8% year over year, and new vehicle CPI dipped slightly on the month ^[12]. So the modest nominal gain at dealers came alongside falling vehicle prices rather than because of them ^[13]. Autos are the soft category in this report, and they are soft on volume-friendly pricing, which is a different problem from a spent-out consumer.

Three things to watch. Seasonal adjustment factors sum to zero over twelve months, so a July pushed down mechanically implies an August or September pushed up ^[14]; treat that rebound with the same scepticism now warranted for this weakness. Second, the August ecommerce comparison flips in the other direction, since August 2025 was itself down 4.4% off a Prime Day month ^[6]. Third, the Census Bureau's X-13 ARIMA-SEATS adjustment leans on trading-day counts that exclude weekends and holidays, an awkward fit for retailers open seven days a week and ecommerce open around the clock ^[15]. The measurement problem is structural, not a one-month glitch, which argues for tracking the unadjusted year-over-year series alongside the headline rather than instead of it.