RiskyEats

Semiannual · Statistical risk modeling · ORSM Phase 2

Model forecast A probabilistic projection—the future, not the present. Definitions →

Semiannual Closure-Pressure Outlook

Aggregate six-month outlook with model limits and named results withheld.

Florida restaurant outlook, next six months: 4,151 licenses in elevated/high inspection-pattern bands. The current baseline scores 102,084 licenses across four analytical bands (high=267, elevated=3,884, moderate=13,534, low=84,399). Named closure forecasts are withheld while RiskyEats completes its model card, record-completeness rules, editorial threshold, and correction workflow.

Updated 2026-09-01 14:33 UTC

By RiskyEats / FRI Editorial Desk

ORSM high-risk band 267 RiskyEats calculation · current published snapshot · Florida
ORSM elevated 3,884 RiskyEats calculation · current published snapshot · Florida
ORSM moderate 13,534 RiskyEats calculation · current published snapshot · Florida
ORSM low 84,399 RiskyEats calculation · current published snapshot · Florida
Modeled licenses ≥50% 0 RiskyEats calculation · current published snapshot · Florida
NIM forecast source NV-Tesseract Forecast NIM RiskyEats calculation · current published snapshot · Florida

The six-month outlook is built on the Phase 2 ORSM baseline: 102,084 scored licenses distributed across four risk bands. 267 licenses score in the high-risk band (anomaly_score > 0.75), 3,884 in elevated (0.50–0.75), 13,534 in moderate (0.25–0.50), and 84,399 in low (< 0.25). But this edition is not a score table. The report reads the driver stack behind every named anomaly and explains why that license belongs on the desk's watch board.

The Phase 2 baseline is operator-tuned and explainable: severity_z, emergency-order counts, and ownership-change filings are not decorative fields. They are the factual spine of the editorial read — why the venue is abnormal, what kind of abnormality it is, and what the next reporting question should be.

Source dataset: the published RiskyEats data snapshot (Phase 2), the published RiskyEats data snapshot (Phase 3).

TSE Anomaly Overlay

The TSE overlay adds signal-level anomaly context above the ORSM license baseline. These rows are current anomaly detections, not closure findings.

TSE time-series anomaly Hot Pot (cuisine) DBPR inspection severity-point activity per week, eligible cuisines typical range 26 of 100 2026-08-24
TSE time-series anomaly Pinellas (county) Foursquare-reported county DBPR inactive/delinquent-license rate per week, eligible Florida counties typical range 23 of 100 2026-07-27
TSE time-series anomaly Latin American (cuisine) DBPR inspection severity-point activity per quarter, eligible cuisines typical range 24 of 100 2026-07-01
TSE time-series anomaly American (cuisine) Foursquare-reported cuisine DBPR inactive/delinquent-license rate per month, top 30 eligible cuisines typical range 24 of 100 2026-07-01
TSE time-series anomaly Pinellas (county) Foursquare-reported county DBPR inactive/delinquent-license rate per month, eligible Florida counties typical range 23 of 100 2026-07-01
TSE time-series anomaly Statewide Strip-mall restaurant DBPR inactive/delinquent-license rate per week (unit/suite address) typical range 24 of 100 2026-06-29
TSE time-series anomaly Gainesville (city) Foursquare-reported city DBPR inactive/delinquent-license rate per week, eligible Florida cities typical range 23 of 100 2026-06-29
TSE time-series anomaly American (cuisine) Foursquare-reported cuisine DBPR inactive/delinquent-license rate per week, top 30 eligible cuisines typical range 25 of 100 2026-06-08
TSE time-series anomaly D5 (district) Inspector activity per month, per district typical range 24 of 100 2026-06-01
TSE time-series anomaly Jacksonville (metro) DBPR emergency-order and closure events per 10,000 accommodation and food-service jobs, per metro typical range 23 of 100 2026-06-01
TSE time-series anomaly Statewide FB/news scraped closure mentions per week typical range 28 of 100 2026-05-04
TSE time-series anomaly Daytona Beach (city) Foursquare-reported city DBPR inactive/delinquent-license rate per month, eligible Florida cities typical range 25 of 100 2026-05-01

Source dataset: the published RiskyEats data snapshot via tse_overlay.get_anomalies_for_signal().

TSE Forecast Overlay

The forecast overlay shows signal-level trajectories beside the ORSM outlook. Forecast rows are model projections and remain separate from public-record facts.

city DBPR inactive/delinquent-license rate weekly city=jacksonville -0.001 0.001 0.000
city DBPR inactive/delinquent-license rate weekly city=miami -0.000 0.001 0.000
city DBPR inactive/delinquent-license rate weekly city=tampa -0.000 0.000 0.000
city DBPR inactive/delinquent-license rate weekly city=orlando -0.000 0.000 0.000
city DBPR inactive/delinquent-license rate weekly city=kissimmee -0.000 0.000 0.000
city DBPR inactive/delinquent-license rate weekly city=new port richey -0.000 0.000 0.000
city DBPR inactive/delinquent-license rate weekly city=oviedo -0.000 0.000 0.000
city DBPR inactive/delinquent-license rate weekly city=sarasota -0.000 0.000 0.000
city DBPR inactive/delinquent-license rate weekly city=daytona beach -0.000 0.000 0.000
city DBPR inactive/delinquent-license rate weekly city=fort myers -0.000 0.001 0.001
city DBPR inactive/delinquent-license rate weekly city=opa-locka -0.001 0.001 0.000
city DBPR inactive/delinquent-license rate weekly city=tallahassee -0.000 0.000 0.000

Source dataset: the published RiskyEats data snapshot via tse_overlay.get_forecast_for_signal().

Scenario Rollup — Elevated/High Concentration by Metro

The metro rollup answers where regulatory pressure is concentrating. Miami carries the highest concentration of elevated/high-band licenses at 1.8%, which means the forward-look is more uncertain there than in metros with thinner high-band stacks.

This is not a closure count. It is an aggregate of per-license anomaly scores — the same scores that power the per-license table above — rolled up to metro geography. A high concentration here means the model sees more individual licenses behaving outside their own historical norm, which the editorial desk reads as a pressure signal.

Source dataset: the published RiskyEats data snapshot grouped by metro county definitions.

Metro-level ORSM band concentration

Miami 5391 95 1.8% 95 of 5391 scored licenses in the elevated+ band
Fort Lauderdale 2474 46 1.9% 46 of 2474 scored licenses in the elevated+ band
West Palm Beach 2097 29 1.4% 29 of 2097 scored licenses in the elevated+ band
Port St. Lucie 851 7 0.8% 7 of 851 scored licenses in the elevated+ band
Key West 253 6 2.4% 6 of 253 scored licenses in the elevated+ band
Tampa 5655 69 1.2% 69 of 5655 scored licenses in the elevated+ band
Orlando 7256 111 1.5% 111 of 7256 scored licenses in the elevated+ band
Jacksonville 2599 51 2.0% 51 of 2599 scored licenses in the elevated+ band
Fort Myers 3479 55 1.6% 55 of 3479 scored licenses in the elevated+ band
Tallahassee 4050 78 1.9% 78 of 4050 scored licenses in the elevated+ band

Raw counts mislead -- Putnam County losing 4 places is different from Miami losing 4. Use the rate or per-capita column for defensible comparisons. Methodology details linked above.

Source dataset: the published RiskyEats data snapshot.

NIM Forecast Scenario Rollup

The NV-Tesseract Forecast NIM produced 24 metro×target forecast pairs for this semiannual edition. Each pair compares the last observed value against the forecast-horizon endpoint, labeled as a directional trend rather than a point estimate.

south-florida · severity_sum: 1864.0 → 865.3, trend down 53.6% over the forecast horizon.

south-florida · eor_count: 2.0 → 2.8, trend up 37.7% over the forecast horizon.

south-florida · total_inspections: 184.0 → 95.6, trend down 48.1% over the forecast horizon.

tampa-bay · severity_sum: 750.0 → 307.2, trend down 59.0% over the forecast horizon.

tampa-bay · eor_count: 2.0 → 2.2, trend up 8.1% over the forecast horizon.

tampa-bay · total_inspections: 101.0 → 41.3, trend down 59.1% over the forecast horizon.

Source dataset: the published RiskyEats data snapshot. Model: NV-Tesseract Forecast NIM.

NIM forecast metro×target summary

south-florida severity_sum 1864.0 865.3 down 53.6%
south-florida eor_count 2.0 2.8 up 37.7%
south-florida total_inspections 184.0 95.6 down 48.1%
tampa-bay severity_sum 750.0 307.2 down 59.0%
tampa-bay eor_count 2.0 2.2 up 8.1%
tampa-bay total_inspections 101.0 41.3 down 59.1%
orlando severity_sum 352.0 261.4 down 25.7%
orlando eor_count 1.0 1.3 up 34.2%
orlando total_inspections 38.0 32.0 down 15.7%
jacksonville severity_sum 301.0 142.9 down 52.5%
jacksonville eor_count 2.0 0.9 down 52.8%
jacksonville total_inspections 29.0 15.5 down 46.7%
florida-keys severity_sum 14.0 67.8 up 384.3%
florida-keys eor_count 0.0 0.0 new activity from a zero prior base
florida-keys total_inspections 2.0 6.8 up 240.0%
treasure-coast severity_sum 110.0 54.2 down 50.7%
treasure-coast eor_count 0.0 0.2 new activity from a zero prior base
treasure-coast total_inspections 17.0 7.5 down 56.1%
southwest-florida severity_sum 130.0 145.6 up 12.0%
southwest-florida eor_count 0.0 0.5 new activity from a zero prior base
southwest-florida total_inspections 25.0 24.4 down 2.3%
north-florida severity_sum 113.0 51.9 down 54.1%
north-florida eor_count 0.0 0.5 new activity from a zero prior base
north-florida total_inspections 17.0 7.9 down 53.6%

Source dataset: the published RiskyEats data snapshot.

Aggregate 90-day closure-pressure model

The model currently places 0 active licenses above the 50% analytical threshold. Top-decile precision is 27.8%. Named-business results are not published while identity, record completeness, threshold governance, evidence linking, and correction safeguards are under review.

RiskyEats model output; see Methods for sources and limitations.

High-Pressure Cities Ranking

Fort Myers leads the city pressure ranking with 96 DBPR delinquent/inactive (status 45/46) license records against 1,144 active-at-start records, a 8.4% delinquency-pressure read. Delinquency is a public-record regulatory state, not by itself proof a business has closed. Named city records include SOUL BOWL ACAI in Fort Myers, and MIKUU RAMEN in Fort Myers.

The city ranking is a pressure atlas, not a restaurant review. It tells the desk where DBPR license-delinquency concentration is dense enough to justify assignment work before the next six-month outlook.

Source dataset: the published RiskyEats data snapshot, cities_ranked.

Top 20 high-pressure cities

1 Fort Myers 1144 96 8.4% 8.4% closure pressure; percentile 100.0
2 Fort Lauderdale 1096 91 8.3% 8.3% closure pressure; percentile 98.0
3 Tallahassee 822 68 8.3% 8.3% closure pressure; percentile 96.0
4 Oakland Park 192 15 7.8% 7.8% closure pressure; percentile 94.0
5 Naples 1108 84 7.6% 7.6% closure pressure; percentile 92.0
6 Sunrise 324 24 7.4% 7.4% closure pressure; percentile 90.0
7 Miami 4898 360 7.3% 7.3% closure pressure; percentile 88.0
8 Sebring 141 10 7.1% 7.1% closure pressure; percentile 86.0
9 Orlando 4520 318 7.0% 7.0% closure pressure; percentile 84.0
10 Sarasota 971 68 7.0% 7.0% closure pressure; percentile 82.0
11 Delray Beach 380 26 6.8% 6.8% closure pressure; percentile 80.0
12 Punta Gorda 148 10 6.8% 6.8% closure pressure; percentile 78.0
13 Coral Springs 307 20 6.5% 6.5% closure pressure; percentile 76.0
14 Jacksonville 2864 186 6.5% 6.5% closure pressure; percentile 74.0
15 Oviedo 174 11 6.3% 6.3% closure pressure; percentile 72.0
16 Miami Beach 723 45 6.2% 6.2% closure pressure; percentile 70.0
17 Davie 246 15 6.1% 6.1% closure pressure; percentile 68.0
18 Key West 394 24 6.1% 6.1% closure pressure; percentile 66.0
19 Bradenton 712 43 6.0% 6.0% closure pressure; percentile 64.0
20 Destin 239 14 5.9% 5.9% closure pressure; percentile 62.0

Raw counts mislead -- Putnam County losing 4 places is different from Miami losing 4. Use the rate or per-capita column for defensible comparisons. Methodology details linked above.

Source dataset: the published RiskyEats data snapshot.

High-Attrition Neighborhoods

Atlantic Beach in Jacksonville leads the neighborhood pressure board with 24 DBPR delinquent/inactive (status 45/46) license records against 43 active records, or 55.8%. Delinquency is a regulatory state, not confirmed closure. The dominant cuisine printed for that geography is Italian & Pizza.

Neighborhoods are where the model starts to look less like a spreadsheet and more like a real-estate story.

Source dataset: the published RiskyEats data snapshot, neighborhoods_ranked.

Top 30 high-attrition neighborhoods

1 Atlantic Beach Jacksonville 43 24 55.8% Italian & Pizza
2 Coral Springs Coral Springs 233 19 8.2% Italian & Pizza
3 Sunrise Sunrise 189 12 6.3% Mexican
4 Davie Davie 178 11 6.2% Japanese, Thai & Sushi
5 Doral Doral 196 11 5.6% Japanese, Thai & Sushi
6 Pembroke Pines Pembroke Pines 309 16 5.2% Fast Food
7 Hialeah Hialeah 249 12 4.8% Cuban
8 Hollywood Hollywood 286 11 3.8% Italian & Pizza

Raw counts mislead -- Putnam County losing 4 places is different from Miami losing 4. Use the rate or per-capita column for defensible comparisons. Methodology details linked above.

Source dataset: the published RiskyEats data snapshot.

Cuisine-Level Forecast

The cuisine forecast is blunt: cuisines above the statewide median DBPR inactive/delinquent-license rate of 5.3% are treated as shrink candidates unless operator churn is low enough to argue for stability. Donuts, Sandwiches, Fast Food, Ice Cream, and Desserts sit near the top of the pressure board in the current atlas.

Pizza remains structurally important even when it is not the monthly rate leader because the address-churn file keeps pulling it back into the story.

Source dataset: the published RiskyEats data snapshot and the published RiskyEats data snapshot.

Cuisine-level six-month forecast

Sandwiches projected shrink 12.6% 0.4% 12.6% DBPR inactive/delinquent-license rate and 0.4% operator-churn rate
Desserts projected shrink 9.7% 0.7% 9.7% DBPR inactive/delinquent-license rate and 0.7% operator-churn rate
Fast Food projected shrink 8.9% 0.2% 8.9% DBPR inactive/delinquent-license rate and 0.2% operator-churn rate
Juice/Smoothie projected shrink 6.3% 0.0% 6.3% DBPR inactive/delinquent-license rate and 0.0% operator-churn rate
Bagels projected shrink 6.2% 0.0% 6.2% DBPR inactive/delinquent-license rate and 0.0% operator-churn rate
Cafe projected shrink 5.6% 0.0% 5.6% DBPR inactive/delinquent-license rate and 0.0% operator-churn rate
Cocktail Bar projected shrink 5.4% 0.0% 5.4% DBPR inactive/delinquent-license rate and 0.0% operator-churn rate
Sports Bar projected shrink 5.3% 0.0% 5.3% DBPR inactive/delinquent-license rate and 0.0% operator-churn rate
Caribbean projected growth/stability 5.2% 0.0% 5.2% DBPR inactive/delinquent-license rate and 0.0% operator-churn rate
Mexican projected growth/stability 5.0% 1.2% 5.0% DBPR inactive/delinquent-license rate and 1.2% operator-churn rate
Vietnamese projected growth/stability 4.9% 0.0% 4.9% DBPR inactive/delinquent-license rate and 0.0% operator-churn rate
Bakery projected growth/stability 4.7% 0.9% 4.7% DBPR inactive/delinquent-license rate and 0.9% operator-churn rate
Diner projected growth/stability 4.5% 0.4% 4.5% DBPR inactive/delinquent-license rate and 0.4% operator-churn rate
Pub projected growth/stability 4.4% 0.0% 4.4% DBPR inactive/delinquent-license rate and 0.0% operator-churn rate
Health Food projected growth/stability 4.4% 2.0% 4.4% DBPR inactive/delinquent-license rate and 2.0% operator-churn rate
Deli projected growth/stability 4.2% 0.0% 4.2% DBPR inactive/delinquent-license rate and 0.0% operator-churn rate
Cuban projected growth/stability 4.0% 1.1% 4.0% DBPR inactive/delinquent-license rate and 1.1% operator-churn rate
Lounge projected growth/stability 4.0% 0.0% 4.0% DBPR inactive/delinquent-license rate and 0.0% operator-churn rate
Bar projected growth/stability 3.9% 1.3% 3.9% DBPR inactive/delinquent-license rate and 1.3% operator-churn rate
American projected growth/stability 3.9% 1.3% 3.9% DBPR inactive/delinquent-license rate and 1.3% operator-churn rate

Raw counts mislead -- Putnam County losing 4 places is different from Miami losing 4. Use the rate or per-capita column for defensible comparisons. Methodology details linked above.

Source dataset: the published RiskyEats data snapshot.

The 90-Day Watch List

Watch list pending — closure predictions not yet scored.

Source dataset: the published RiskyEats data snapshot.

Six-Month Rear-View — Closed-Loop Model Validation

Of 1,450 licenses that moved to DBPR delinquent/inactive status (45 or 46) in the last six months, 0 currently score in the elevated or high ORSM band (0.0%). Delinquency is a public-record regulatory state, not by itself confirmation a business has closed. This is a rear-view validation: the same Phase 2 scoring that produces the forward-look above catches a material fraction of these delinquency events in its elevated+ bands. A higher catch rate means the model is surfacing real pre-delinquency signal; a lower rate tells the desk that delinquencies are dominantly structural (retirement, lease expiry, franchisor decisions) rather than inspection-pressure events — useful context for interpreting the forward-look.

No rear-view delinquency matches found — either no licenses went delinquent in the window or the model does not yet score delinquent licenses.

Source dataset: the published RiskyEats data snapshot joined to the published RiskyEats data snapshot delinquent-status (45/46) records.

Editorial Verdict — Where to Look First

Most pressured metros: Key West, Jacksonville, Fort Lauderdale lead the elevated/high-concentration board. The editorial desk should assign coverage resources to these geographies first — they carry the densest stacks of individual licenses behaving outside their own historical norms, which is the Phase 2 definition of elevated regulatory pressure.

Most pressured cuisine cohorts: Sandwiches, Desserts, Fast Food top the DBPR inactive/delinquent-license rate board. These are the cohorts where the forward-look is least certain — when a cuisine is already losing units at an above-median rate, the Phase 2 per-license signals in that cuisine carry higher editorial weight because the structural backdrop supports the per-license read.

Highest-pressure cities: Fort Myers, Fort Lauderdale, Tallahassee lead the statewide city ranking. The city pressure is a complement to the metro-level ORSM rollup — a city with high closure pressure AND a high ORSM elevated-band concentration is the strongest editorial signal in this brief.

Highest-attrition neighborhood: Atlantic Beach in Jacksonville with 24 DBPR delinquent/inactive (status 45/46) license records against 43 active records (0.0%). Neighborhoods are where the model starts to look less like a spreadsheet and more like a real-estate story — the editorial desk should cross-reference the per-license table above against neighborhood-level delinquency to identify corridor-level patterns.

Source datasets: the published RiskyEats data snapshot, the published RiskyEats data snapshot, the published RiskyEats data snapshot.

Methodology

Methodology: ORSM Phase 2 baseline produces per-license anomaly scores from operator-tuned weighted-max subscores (severity spike, cohort deviation, EOR history, ownership churn, license-expiry pressure, administrative distress, external closure signals). Bands: high (>0.75), elevated (0.50-0.75), moderate (0.25-0.50), low (<0.25). Scenario rollups aggregate band concentrations by metro county definitions. NIM forecast layer (NV-Tesseract) provides metro-level severity/EOR/inspection-volume projections when the forecast parquet is present. Closure predictions are Phase 3 logistic regression outputs, included here as complementary forward-look. All drivers are printed in plain English so the editorial desk can audit every call.

Data sources: Statistical risk modeling · ORSM Phase 2. Built from DBPR public inspection records, Florida Sunbiz business filings, and Foursquare/OSS location data. All counts are derived from public records — no estimates or projections unless labeled as forecast.

Share: Facebook X LinkedIn Email