San Antonio Bar and Restaurant Revenue

Those two facts together are the most useful thing in this dataset, and neither is visible alone. Revenue alone shows a catastrophe; open businesses alone shows almost nothing happening. On the same timeline you see the actual shape: the businesses were still there, still permitted, still filing, and reporting almost no sales. That is why the two charts are drawn separately rather than stacked on one pair of axes. Dollars and business counts are different units, and a chart with two y-scales lets whoever drew it decide how dramatic the comparison looks by choosing where the axes cross.

It is also worth knowing what the recovery does not tell you. A permit that stops filing leaves this dataset silently, and a new permit at the same address arrives as a different business. The venue line is not a count of survivals, and this record cannot tell you how many places open in Feb 2020 are the same ones open today.

Every figure is a grouped aggregate computed by the State’s own API and reconciled before publication. Nothing downloads the underlying rows, which matters because the platform silently caps a grouped result at a thousand rows unless you tell it otherwise, and a cap like that does not announce itself.

Two independent reconciliations run on every rebuild: the citywide total must equal the sum of the monthly series, and the per-ZIP window must equal the same months in that series. If either disagrees by more than a rounding margin the build publishes nothing and the previous good copy stays put. Every number in this prose reads from the same file as the charts, through a shortcode, so the essay cannot drift out of agreement with the data. Last rebuilt 26 September 2026. The CSV and the documented JSON API are linked from the panel above, and the API’s quality endpoint returns the county-code trap, the disclosure floor and the reporting lag as structured data.