Gera Recall Watch Index (GRWI) — Methodology

1. Data source

The GRWI is computed exclusively from the FSA Food Alerts open REST API (Food Standards Agency, UK). The API is publicly accessible without authentication at data.food.gov.uk/food-alerts. All data is published under the Open Government Licence v3.0.

• Fetch URL: https://data.food.gov.uk/food-alerts/id.json?_limit=2000 (returns up to 2,000 alerts; current DB size 1328 alerts)
• Last fetched: 2026-06-20
• Total alerts in FSA database: 1,328 (all published since 2018-01-09)
• Alert types included: AA (Allergy Alerts), PRIN (Product Recalls), FAFA (Food Alerts for Action), FAWL (Food Alerts for Awareness)

2. Rolling 24-month window

Only alerts with a created date between 2024-06-20 and 2026-06-19 are included in the index. This rolling window (320 alerts) represents the most recent 24 months of food-safety enforcement. Older alerts are excluded because recall risk reflects current supply-chain and labelling practice, not historical incidents.

3. Category classification

Each alert is classified into one of 13 product categories by keyword matching on the alert's title / shortTitle field. The first matching category wins; alerts that match no keyword fall into “Other Food Products”. Category keywords are:

4. Allergen classification

Allergens are identified by keyword search across the alert's title and the problem.riskStatement field. All 14 EU-regulated allergens (as listed in EU Regulation No 1169/2011, retained in UK law post-Brexit) are covered. Allergen notation follows the FSA's own allergen vocabulary at data.food.gov.uk/codes/alerts/def/allergen. Alerts implicating no named allergen are classified as “Other/Unknown” and excluded from allergen hub counts.

5. GRWI formula

Where:

• alert_count_24mo — total FSA alerts (any type) in the rolling 24-month window for this category.
• allergen_alert_fraction — proportion of alerts that are Allergy Alerts (type AA), the highest consumer-risk class (0–1). Weight of 20 reflects that AA alerts pose immediate risk to allergic consumers.
• allergen_diversity — count of distinct named allergens implicated across all alerts in the category (excl. Other/Unknown). Higher diversity = broader risk surface.
• trend_factor — alerts in the last 6 months (Jan– Jun 2026) divided by alerts in the prior 6 months (Jul–Dec 2025). A rising trend inflates the score. Set to 1.0 if the prior period has zero alerts (division-by-zero guard).
• GRWI — min–max normalised so the highest-raw category scores 100 and the lowest scores 0. All categories with GRWI > 0 have at least one real FSA alert in the window.

6. Limitations

• GRWI reflects reported FSA alerts, not absolute product safety. Categories with fewer SKUs (e.g. Sandwiches & Wraps) may score low simply because fewer products carry FSA alert obligations, not because they are inherently safer.
• Category assignment is keyword-based and may misclassify multi-ingredient products. The FSA API does not carry a structured product-category field.
• Allergen extraction is text-based; complex free-text risk statements may miss allergens not named in the title. Counts labelled “Other/Unknown” represent alerts where no named allergen was detected.
• The GRWI is a relative risk index for the 13 categories represented in the 24-month FSA window, not an absolute food- safety measure.

7. Reproduction

To reproduce this index independently:

1. Fetch: curl https://data.food.gov.uk/food-alerts/id.json?_limit=2000
2. Filter to alerts where created ≥ 2024-06-20 (rolling 24mo).
3. Classify each alert to one of the 13 categories above via title keyword match.
4. For each category: count total alerts, AA alerts, distinct allergens. Compute 6-month sub-windows for trend_factor.
5. Apply the raw formula and min–max normalise across all 13 categories.

All computations use publicly auditable FSA data. The formula and weights are fixed as of 2026-06-20; future versions will note any changes.