October 22, 2026: retail sales levels for June to August 2026
The September 30 edition combined two Census files that did not agree. Its monthly history ran through May 2026, and its June, July and August 2026 values came from the September 16 advance-release workbook, whose adjusted levels were about 4.5% higher than the monthly history for the same months (for example, August 2025 was 730,037 million dollars in the advance workbook and 700,218 million in the history). The splice created a spurious jump of about 5% in June 2026 and overstated year-over-year retail growth for June to August, shown as 10.5% for August.
The October 22 edition uses the October 15 advance workbook, whose values match the monthly history for every overlapping month. August 2026 is 739,322 million dollars, up 5.6% from a year earlier, and September 2026 is 745,021 million, up 6.3%. The September 30 edition and its files remain preserved unchanged as evidence; the release-change comparison lists these as changed values.
Report a specific observation
Email [email protected] with the page address, series name, reference period, displayed value and the source you believe differs. If possible, include the snapshot ID, transformation and seasonal-adjustment basis. These details let the publisher reproduce the figure rather than compare two superficially similar measures.
A discrepancy can result from different releases, adjustment flags, units, rounding or a source revision. It can also be a processing error. We first identify whether the two figures represent the same statistic and information set. A later official estimate does not automatically make a correctly preserved earlier snapshot erroneous.
Review and correct the affected outputs
The review traces the displayed number to its source code, acquired file checksum and transformation. If a processing error is confirmed, the correction must identify the affected indicator, reference periods and outputs. Charts, tables, downloads and saved views are all checked because they should derive from one approved dataset.
A corrected version receives a distinct snapshot identity. An earlier version is not silently rewritten. If an old file must be withdrawn because it is materially misleading or cannot legally be retained, the archive should explain the withdrawal and point to the correction. Rollback can restore the last approved build while a candidate is reviewed.
Separate producer revisions and site errors
Agency revisions change a producer’s estimate for the same period. Site corrections fix errors introduced by this publication. Method changes can alter comparability. Keeping these event types separate helps readers understand whether a reported economic movement changed or the site had mislabeled a figure.
A large movement alone is not proof of an error. The publisher checks source units, decimal scales, missing flags, definitions and release notes before deciding whether a candidate should be approved. Unpublished or suppressed values must remain unavailable; they are never replaced by zero to make a chart continuous.
Correction log
The September 30 prelaunch review recorded three data/presentation corrections: the ECI changed from a private-industry source to the intended civilian source; the empty unadjusted unemployment variant was replaced with valid code LNU04000000; and the GDP article’s millions-to-billions label error was corrected in a separate export. Original files remain preserved and explicitly identified.
The release-change workspace documents the affected identities, original and corrected files, and values. Updated BEA estimates are classified separately as producer revisions. Selected-adjustment citations, invalid-comparison handling and saved snapshot views were also corrected in the interface. Source limitations remain documented on the status page.
October 4: consistent earnings deflation and research views
The economy dashboard and since-2019 research page previously divided seasonally adjusted earnings by unadjusted CPI. Both now use the stored seasonally adjusted CPI-U series CUSR0000SA0, matching CES0500000003 earnings. The August 2019–August 2026 real-earnings calculation changes from 2.74% to 2.80%; the matched twelve-month dashboard result changes from −0.30% to −0.26%. Original source snapshots and the original derived export remain preserved; the corrected inputs have a separate research-2026-10-04 distribution.
Unavailable saved snapshot/indicator queries now retain their explicit error through later control changes. Archived definition/source links expose the corresponding archived metadata. Directory numbers show transformations and seasonal adjustment. Current freshness notices distinguish confirmed producer publications from the frozen approved edition.