Method
How a public DBPR inspection becomes a county-adjusted brand score, and what the numbers can and cannot tell you.
Data
Every food service inspection file DBPR publishes on its public records page, fiscal years 2016-17 to 2026-27 (the last one in progress), plus the weekly emergency closure extracts and the licence files. Latest inspection in the data: 2026-11-09. Built 2026-11-10. Nothing is edited by hand; the site is rebuilt from the files by one command.
The measure
High-priority violations at a routine inspection: the items DBPR treats as most directly linked to foodborne illness. InspectLens never describes a store or brand as anything other than its high-priority violations per routine inspection, relative to its county.
The unit
A routine inspection can have callback visits that re-check what was cited. Callbacks record far fewer violations, so counting them would lower the average of exactly the stores that needed one. Brand and store scores use the first visit of each routine inspection. County figures on the home and county pages count every routine visit, as DBPR's file does, so anyone can reproduce them: FY2024-25, 116,597 routine inspections statewide, state average 0.98.
The baseline
For each county and fiscal year (July to June), the average high-priority violations per first routine visit. A score is observed minus that baseline. County was chosen over DBPR's 7 district offices because it explains more of the variance: 6.6% vs 3.9% (omega squared, which corrects for county having 201 groups to district's 21).
Brand scores
- Window: FY2023-24 to FY2025-26, the last 3 complete fiscal years.
- Score: the mean of observed minus baseline over the brand's routine inspections.
- 95% interval: bootstrap with 2,000 resamples that resample stores, not inspections, because inspections of the same store are related.
- Ranked only with at least 30 stores and 50 inspections, and only inside a segment (burgers, pizza, casual dining...), because menu and format differ.
- Raw averages are shown only next to the baseline the brand's stores face.
Store scores
A store gets one to three routine visits a year, so its own average is mostly noise. Store scores, computed only for chain stores with at least 2 inspections, are shrunk toward the brand mean in proportion to how little data the store has (empirical Bayes). The site reports their distribution per brand, not individual stores.
Brand matching
DBPR records a business name, not a brand. Names are matched to 108 brands with reviewed aliases after store numbers are removed. On a hand-labelled sample of 400 names, precision is 99.0% (301 matched names) and recall 97.1% (310 true brand stores). Co-branded units (KFC with Taco Bell, Dunkin' with Baskin-Robbins) count for neither brand. A store converted to another brand keeps its licence; only visits under the current brand count for it.
Coverage gap
Florida splits food inspection between DBPR (restaurants and other food service) and FDACS (grocery, many coffee shops, convenience food). Brands whose stores are mostly inspected by FDACS, such as Starbucks, are excluded rather than scored on an unrepresentative subset.
Chains vs everyone else
| Group | vs county | 95% interval | Raw | Baseline | Stores | Inspections (n) |
|---|---|---|---|---|---|---|
| Reviewed chains | -0.38 | -0.39 to -0.37 | 0.77 | 1.15 | 12,599 | 48,703 |
| All other licences | +0.08 | +0.07 to +0.09 | 1.24 | 1.16 | 57,422 | 232,519 |
Restaurant lookup
The lookup lists every licence DBPR inspected since July 2023 (80,823 restaurants, 477,262 visits), chain or independent, and shows facts only: each visit's date, type, DBPR's result, the violation counts and the codes cited. It gives no score and no rank, because a single restaurant has one to three routine inspections a year, too few for a fair comparison. The one piece of context is the county average: high-priority violations at a first routine visit in the same county and fiscal year, the same baseline the brand scores use. "Passed, no follow-up needed" and the other plain-English results translate DBPR's own outcome text, which is shown underneath. Emergency closure reasons come from DBPR's weekly extracts and cover 2025-12-29 to 2026-11-06 only. Name and address are the most recent DBPR recorded for the licence. A licence with no inspection for 18 months is marked as possibly closed.
What a score is not
A snapshot inspection count is not a food-safety rating, and a difference between counties is more likely to reflect local inspection practice than kitchens. That is the reason for the county baseline. Violation code descriptions are DBPR's, from its file layout notes.