Tobacco Connection

The five headline numbers, each with the window it is measured over and, where a gate fired, the reason it reads as it does instead of the number.

• One release carries most of this change — likely a data-coverage artifact, not store openings.

Every quarterly observation the index holds for this brand, as a raw count and rebased to 100.

• Indexed values rebase the series to 100 at the first vintage carrying a non-zero count, which is what makes a 300-store chain and a 3,000-store chain the same shape. The raw count beside it is the figure to quote.
• A count is what the index observed on that date. It can lag a chain's own reporting, and a rise between two vintages is not by itself a store opening — see the openings table below, which counts them one at a time.

Each individually verified opening, newest first, with the city and state the live map holds and cannot display.

• 1 openings.
• Each row cleared both halves of the openings gate: a place id the index had never seen, created inside the window. The ingest stamp is printed so a batch load is visible — a supplier loading many records on one day produces the same signature as a chain opening stores.

The step array every pace and every verdict on this page is computed from, printed so the arithmetic can be checked.

• Pace is openings ÷ step length × 91, the same formula the latest-quarter verdict uses, printed here so it can be checked against the two numbers beside it.
• Batch-suspect is the share of a step's records that arrived on an ingest day or in a concentration the build flagged. A step over 50% batch-suspect withholds the COMPARISON to the step before it, not the step's own count: batchiness impeaches timing claims, not existence claims.
• A step is one interval between two quarterly vintages, so this table has one fewer row than the quarterly series above and its rows must not be read against that table's line by line.

The same records, grouped. This is the legible form of the live map's heat layer.

• Grouped from the dated records above, including the ones past the rendered cap.
• The interactive map draws these as an H3 hexagon ramp. A cell id is not a place a reader can act on, so the same signal is grouped by state and city here instead.

• A city is the locality string on the record plus its state code. Records the index could not localise are counted in their own row rather than dropped, so this table sums to the total above.

Ranked by acceleration — opening faster now than they were a year ago.

Tobacco Connection is not in the credible ranking for this category — its pattern looks like a coverage artifact rather than store openings. The two numbers that flagged it are in the key-figures footnotes above.

Rankings exclude companies whose entire rise lands in a single quarter — that pattern is a change in data coverage, not expansion. They also exclude domains whose index records are departments and facilities rather than distinct stores, and the two identity classes that do not lease a retail box.

• Acceleration is stores per quarter recently minus stores per quarter earlier in the window. Positive means they are opening faster than they were — a chain adding 40 a quarter and holding steady scores 0.
• Biggest quarter and Rising quarters are the two numbers the credibility gate is decided on, and in the interactive view they appear only inside a hover tooltip on excluded rows. A row passing needs at most 70% of its change in one quarter and a count that rose in at least 85% of quarters.
• A growth percentage is withheld on any row whose index coverage is still maturing: the count movement is real, but dividing by a base the index was still discovering produces a number rather than a measurement. The counts beside it are unaffected.

The most locations added, in absolute terms. A different ordering of the same pool, on this page rather than behind a sort control — and the same twelve columns, because a row that appears here and not in the ranking above must not lose a field to being in the shorter table.

The biggest increase against their own starting count. A different ordering of the same pool, on this page rather than behind a sort control — and the same twelve columns, because a row that appears here and not in the ranking above must not lose a field to being in the shorter table.

Growth beyond the median for their own category. A different ordering of the same pool, on this page rather than behind a sort control — and the same twelve columns, because a row that appears here and not in the ranking above must not lose a field to being in the shorter table.

The eight rows the interactive comparison lays out side by side, for this brand.

• The interactive view compares up to five chains on these same eight rows, indexed to 100 at the first quarter so a 300-store chain and a 3,000-store chain are the same shape.

The chains immediately above and below this one by location count. Nearest-by-count rather than top-by-count, so every page in a category carries a different set.

Ordered by acceleration, and gated on the two credibility columns plus positive net growth — all four of which are printed here and none of which the interactive rail displays.

• To appear here a chain needs at most 70% of its change in one quarter, a count that rose in at least 85% of quarters, and net growth above zero. Without those three a 'fastest growing' list confidently links to data artifacts.

The states with enough locations to have their own page — at least 3.

• The same gate applies to the link and to the page behind it, so no row here points at a state page that does not exist.