Two retail-sales stories ran on the same Friday and looked like they contradicted each other. One reported a growth streak. The other reported a decline. Both were accurate. They were measuring different things.
The National Retail Federation's release carried the headline 'CNBC/NRF Retail Monitor Shows 10th Month of Sales Growth in July' [1]. By that gauge, retail sales excluding automobile dealers and gasoline stations rose 0.32% month over month, seasonally adjusted, and 5.15% year over year [1]. 'Retail sales maintained their steady upward momentum in July as consumers kept shopping despite ups and downs in other economic indicators,' NRF President Matthew Shay said [1].
The Census Bureau's advance estimate, out the same week, moved the other way. As Yahoo Finance reported the figures, 'Total seasonally adjusted sales came in at $763.6 billion in July, down from a revised $768.1 billion in June,' a 0.6% monthly decline, while 'U.S. retail and food services sales fell 0.6% in July' [2]. Trading Economics recorded the same move: retail sales 'fell 0.6% month-on-month in July 2026, sharply missing expectations for a 0.1% rise and reversing June's 0.2% gain' [3].
The decline checks out arithmetically. A step from $768.1 billion to $763.6 billion is a drop of 0.59%, which rounds to the reported 0.6% [2]. That is a month-over-month change in the total, autos and gasoline included.
The two are not in conflict; they are different instruments. The NRF Retail Monitor is built from actual card-transaction data and strips out automobile dealers and gasoline stations, two of the most volatile categories [1]. The Census advance number is a survey-based early estimate that keeps them in [2]. On the measure the two share, the year-over-year trend, they nearly line up: about 5.15% for NRF and 5.0% for Census [1][2]. Coverage that set 'a 10th month of growth' against the official monthly drop as if one debunked the other was comparing a card-data, ex-autos-and-gas index with a survey-based total. The disagreement is in the method, not in the economy.
The NRF numbers are internally accurate; nothing here says the Retail Monitor is wrong. The error is downstream, in reading two purpose-built measures as one scoreboard. When the categories and the methods differ, the month-over-month signs can differ too, while both series still describe a consumer spending more than a year ago and a little less than the month before.
