Verified CalFresh Store Lists Double Measured Grocery Distance Compared With Commercial Map Directories
We compare raw business map lists with verified SNAP store lists when mapping local food access.
County food teams build access dashboards from commercial map directories, USDA CalFresh retailer lists, county health permits, and census surveys. Each data source uses different counting rules and update schedules. Switching from raw business map tags to verified CalFresh EBT stores changes measured trip distance even when zero stores closed.
Checking county map lists against 514 verified CalFresh EBT stores and 755 county health permits keeps 81.7% of top-need areas (67 of 82 tracts) in the same top tier.[1][2] Raw map queries return 6,613 food spots (such as cafes that do not take CalFresh), so switching to verified grocers doubles mean trip distance by +97% (0.29 to 0.573 miles) with zero store closures.[1][2] Checking map feeds and 5-year census surveys against monthly permits and caseloads keeps grants aimed at real local gaps.[2][3]
Table 1. Four Data Rule Shifts That Look Like Big Policy or Market Shifts
We compare four real data rule shifts across map tags, SNAP lists, permit records, and survey waves.
| Data Check Step |
Source 1: Map Tags
Commercial Map Listings
|
Source 2: EBT Filter
Broad Spots vs. SNAP EBT
|
Source 3: Permits
Deduplicated Health Permits
|
Source 4: Rolls
5-Yr Survey vs. Monthly Rolls
|
|---|---|---|---|---|
| Unfiltered Data Feed | 344 stores1 old tag | 6,613 broad spots0.29 mi | 8,555 raw rows987 old est | 27,068 ACS 2020 HH-4.94% dip |
| Verified Agency List | 6,613 storesmulti-tag pull | 514 SNAP / 755 DEH0.573 mi | 7,335 sites755 / 724 grocers | +56% to +71% CDSS CalFresh |
| Apparent Shift From Data Rule | +1,822% Store Jump | -92.2% Stores+97% Trip | -14.3% Counter Duplicates-1,220 | -4.9% Survey vs. +71% Rolls |
| Real Change on the Street | 0 new storestag fix only | 0 stores closedEBT filter | 0 grocers closedmerged depts | 5-yr wave watered down 2020 |
| Cross-Check & Rank Match | 8.76x higher than 755 permits | 81.7% top-tier steady67/82 | 97.5% retail accuracy477/489 | ACS rose +17.15% in 2022-23 |
| Match Across Data Lists | Google matched 42.1% of retail | 65.2% same tier98.0% in 1 tier | Removed 35 chain cafes & 3 bars | 28,476 (2019) vs 28,880 (-1.4%) |
Table 1 Notes & * Modeled Estimates
Expand notes
Table 1 Notes (Mode B Observed Data Rule Shifts): Each column uses its own group base and is not pooled across columns. Shift 1 counts Santa Clara County food map spots (344 single-tag vs. 6,613 multi-tag). Shift 2 tracks distance for 408 Santa Clara census tracts (6,613 spots vs. 514 SNAP stores, checked against 755 refined / 724 confirmed DEH permits). Shift 3 tracks Santa Clara County DEH food permit deduplication (DEHDENOM_UPDATES_SUMMARY.md: 8,555 raw permit rows minus 1,220 multi-department duplicates = 7,335 unique sites, yielding 755 refined / 724 confirmed SNAP-eligible grocers vs. 987 under an older un-deduplicated keyword rule). Shift 4 compares ACS 5-year survey homes with CDSS monthly caseloads (PHASE6_FINAL_DELIVERABLE.md). In Shift 2 across N = 408 tracts (EXTENDED_VULNERABILITY_SNAP_VERIFIED_RESULTS.md), 266 of 408 tracts (65.2%) stay in the exact same quintile, 400 of 408 (98.0%) stay within ±1 quintile, and 67 of 82 Q5 tracts (81.7%) stay in Q5.
Column Scope. Commercial Map Listings. Base: SCC Store Spots.. Broad Spots vs. SNAP EBT. Base: 408 SCC Tracts.. Deduplicated Health Permits. Base: 8,555 SCC Permits.. 5-Yr Survey vs. Monthly Rolls. Base: Survey vs. Rolls..
Two-Ruler Benchmark Summary. Mode B Store-List Check on 408 Santa Clara Tracts (Permit Deduplication & Survey Window Checks in Table 1). Ruler 1: Broad Food Map List (6,613 Raw Spots): 0.29 Miles Mean Trip (6,613 Spots). Pulling broad food map tags returns 6,613 spots (such as cafes, snack shops, and mini-marts that do not take CalFresh EBT). That shrinks mean tract distance to 0.29 miles (0.25 mi median). Ruler 2: SNAP EBT & Health Permit List (514 EBT / 755 Permits): 0.573 Miles (+97% Trip; 81.7% Top-Tier Steady). Keeping only 514 USDA SNAP EBT stores (checked at a 97.5% accuracy against 755 DEH permitted grocers) doubles mean distance to 0.573 miles. Still, 81.7% of top-need tracts (67/82) and 98.0% within ±1 tier stay put.
Method & Data Audit (TooEarlyToSay.com): EBT Retailer Verification Method → · Transit Stop Deduplication Claim Record → · GTFS Feed Validation Workflow in Python →
Table 2. How to Check Data Feeds So Rule Shifts Do Not Misguide Grants
We set clear check rules for map APIs, SNAP store lists, permit registries, and 5-year census waves.
| Choice Factor | Commercial Map Listings | Broad Spots vs. SNAP EBT | Deduplicated Health Permits | 5-Yr Survey vs. Monthly Rolls |
|---|---|---|---|---|
| Why the Data Shift Happened | Google retired one old store tag, so a single-tag script pulled just 344 stores until multi-tag pulls found 6,613. | Map APIs do not check if a spot takes USDA CalFresh EBT cards. | Raw health permit logs list separate deli, bakery, and meat counters inside one store as separate rows. | The 2020 ACS 5-Year wave blended four pre-COVID years with 2020, creating a -4.94% dip. |
| Risk of Acting on the Raw Shift | Comparing the 344-store run to the 6,613-store run looks like a +1,822% store boom with zero new stores. | Switching from broad map lists to SNAP stores doubles measured trip distance (+97%) with zero store closures. | Using un-deduplicated permit rows overstates store supply by 14% to 30%. | Using 2020 ACS 5-Year tables alone makes CalFresh need look down (-4.94%) when state rolls jumped +56% to +71%. |
| Why Priority Ranks Held Steady | Raw map pulls shift across software updates and should not serve as the base. | Poverty and SNAP hold 55% of index weight (distance 20%), so 81.7% of top-need tracts stayed put. | Merging counter permits by street address gives a clean 755-store denominator (97.5% accuracy). | Once pre-2020 years rolled off, ACS counts rose +17.15% to match state rolls. |
| Data Check Rule for Teams | Check Map Pulls Against Permits. Match map pulls against County DEH health permits. | Use Verified SNAP EBT Lists. Define food access using USDA SNAP EBT retailers (514 stores in Santa Clara). | Deduplicate Permit Addresses. Merge bakery and deli counter rows at the same address before counting stores. | Pair 5-Yr ACS With Monthly Rolls. Use ACS 5-Year tables for tract rates, and track fast shocks with CDSS rolls. |
Table 2 Notes: Tier Shift Matrix, Permit Deduplication & Provenance
Expand notes
• Column Denominators, Tier Shift Matrix & Permit Deduplication (Mode B, N = 408 Santa Clara Tracts): Each column keeps its own base (store spots, 408 Santa Clara tracts, 8,555 Santa Clara DEH permit rows, and survey vs. state rolls). In EXTENDED_VULNERABILITY_SNAP_VERIFIED_RESULTS.md, the 5-component index weights poverty at 30%, SNAP at 25%, store distance at 20%, no-vehicle share at 15%, and density at 10%. Switching from 6,613 spots (0.29 mi mean) to 514 SNAP stores (0.573 mi mean) leaves 266 of 408 tracts (65.2%) in the exact same quintile, 400 of 408 (98.0%) within ±1 quintile (only 8 tracts or 2.0% shift by ±2 quintiles), and 67 of 82 Q5 tracts (81.7%) in Q5. In DEHDENOM_UPDATES_SUMMARY.md and STORE_COUNTS_FOR_ANALYSIS.md, consolidating 8,555 raw DEH permit records removes 1,220 multi-department duplicates (14.3%) to yield 7,335 unique facilities, 724 confirmed SNAP-eligible grocers (97.5% accuracy), and 755–766 refined grocers (vs. 987 under an older un-deduplicated keyword rule). Note that a separate companion TooEarlyToSay transit-stop article (49.0% on 24,421 stops across 8 agencies vs. 12.0% on 64,060 Cal-ITP stop rows across 9,039 state tracts in mobility_deserts_metadata.json and mobility_deserts_calitp_metadata.json) is flagged in FoodSecurity/analysis/MOBILITY_DESERTS_STATUS.md as incomplete and under reconciliation, so it is not used as a policy trigger here.
• Data Provenance: Exact check tables come from FoodSecurity/EXTENDED_VULNERABILITY_SNAP_VERIFIED_RESULTS.md (L1-150), FoodSecurity/STORE_COUNTS_FOR_ANALYSIS.md (L1-120), FoodSecurity/DEHDENOM_UPDATES_SUMMARY.md (L1-135), and FoodSecurity/PHASE6_FINAL_DELIVERABLE.md (L1-110).
• Extended Cell Context. Why the Data Shift Happened (Broad Spots vs. SNAP EBT). Map APIs do not check if a spot takes USDA CalFresh EBT cards or holds an active County DEH food permit. · Risk of Acting on the Raw Shift (Deduplicated Health Permits). Using un-deduplicated permit rows (8,555 rows or 987 keyword matches) overstates store supply by 14% to 30%. · Data Check Rule for Teams (Commercial Map Listings). Match map pulls against County DEH health permits (755 stores, 97.5% accuracy).
Why Checking Data Rules Protects Both Program Checks and Grant Siting
Health teams often build dashboards from map APIs, store lists, permit logs, and rolling census surveys. Each source can change its own rule without notice. A map API drops a store tag (jumping from 344 to 6,613 spots). An EBT filter removes cafes (-92.2% store count, +97% distance). A permit registry lists bakery and deli counters inside one supermarket as separate rows (1,220 duplicates out of 8,555 rows, or 14.3%). Or a 5-year survey waters down a 1-year shock (-4.94% in 2020 ACS vs. +56% to +71% in state CalFresh rolls).
Teams should treat any sudden dashboard jump as a data-rule shift until checked against state or county rolls. Deduplicating county health permits by address (755 refined / 724 confirmed grocers at 97.5% accuracy), pairing 5-year census baselines (-1.4% gap vs. 2019 CalFresh admin rolls) with monthly caseloads, and checking rank stability (where 65.2% of tracts stay in the same quintile and 81.7% of top-need tracts stay in the highest-need tier) keeps county funds aimed at real local need.
Data Sources, Model Inputs, Notes & Endnotes
(Numbered endnotes [1]–[4], state data sources & math rules)
Expand details
Case Study Focus: Study Credibility & Measurement · Instrumentation Shifts (Mode B Multi-Instrument Audit)
Numbered Endnotes & Main Sources
- [1]USDA Food and Nutrition Service (FNS): SNAP Retailer Locator (514 Verified EBT Retailers across 408 Santa Clara County Census Tracts).
- [2]Santa Clara County Department of Environmental Health (DEH): Active Retail Food Facility Permit Registry (755 Permitted Grocers, 97.5% Concordance).
- [3]California Department of Social Services (CDSS): Monthly CalFresh Caseloads vs. U.S. Census Bureau ACS 5-Year Rolling Windows.
- [4]U.S. Census Bureau ACS 5-Year Estimates
Mode B Multi-Ruler Audit: All store and permit counts (344 single-tag, 6,613 broad map spots, 514 SNAP-verified stores, 8,555 raw DEH rows, 7,335 consolidated sites, 1,220 or 14.3% department duplicates, 724 confirmed / 755 refined grocers, 97.5% classification accuracy, 42.1% Google coverage), trip distances (0.29 mi vs. 0.573 mi), tier shifts (266/408 = 65.2% same quintile, 400/408 = 98.0% within ±1 quintile, 67/82 = 81.7% Q5 stability), and ACS vs. CDSS counts (28,476 vs. 28,880 in 2019 = -1.4%; 27,068 in 2020 = -4.94%; +17.15% in 2022-23) come from verified FoodSecurity files on disk. Companion TooEarlyToSay pages (including the seven-county EBT correlation and the 49% to 12% transit-stop series flagged in FoodSecurity/analysis/MOBILITY_DESERTS_STATUS.md) remain under reconciliation and are excluded from grant-siting rules.
Main Sources: We use USDA FNS SNAP Retailer Locator and Santa Clara County Department of Environmental Health (DEH) Retail Food Facility Registry. We also use California Department of Social Services (CDSS) CalFresh Data Tables and U.S. Census Bureau ACS 5-Year Estimates.
Companion Method & Code Guides (TooEarlyToSay.com)
- The Retail Density Paradox: Why More Stores Mean Worse Data. Proposed EBT acceptance verification across 7,023 California food retailers; seven-county counts, matching thresholds, and the r = -0.828 correlation are under reconciliation.
- Transit Data Quality Under Reconciliation. Article-reported shift from 49% to 12% of tracts classified as mobility deserts; the raw versus deduplicated stop counts behind it are under reconciliation.
- GTFS Validation in Python: An Article-Reported Workflow. Six-layer instructional GTFS validation workflow; the seven-county, 11-feed routing results are article-reported and the project scripts are not public.
Common Questions & Method Notes
- Why did filtering from 6,613 broad spots to 514 SNAP stores keep 81.7% of top-need tracts (67 of 82) in the highest-vulnerability quintile?
- In the 5-component Extended Vulnerability Index, economic factors (30% poverty rate and 25% SNAP participation) hold 55% of the index weight while normalized store distance holds 20%. Even though mean distance doubled from 0.29 to 0.573 miles, 65.2% of tracts (266 of 408) stayed in the exact same quintile and 81.7% of Q5 tracts stayed in Q5.
- Why did the 2020 ACS 5-Year survey show a -4.94% dip in SNAP homes during COVID-19?
- The 2020 ACS 5-Year table averages 60 months from January 2016 to December 2020, so 80% of the sample came before COVID-19. Also, 2020 survey collection faced low response rates among low-pay homes. Monthly CDSS rolls tracked the real +56% to +71% jump right away.
- How did the County DEH health permit list check the store count and why was deduplication needed?
- Raw Santa Clara County DEH permit logs contain 8,555 rows because supermarkets hold separate permits for bakery, deli, and meat counters at the same address (1,220 duplicate rows, or 14.3%). Consolidating by address (7,335 unique sites) and classifying retail grocers (97.5% accuracy) yields 755 refined / 724 confirmed SNAP-eligible grocers, proving the 6,613 map count was inflated 8.76x by non-grocery spots.