# Verity beta 3 — September 14, 2026 - Fixed the alcohol exclusion to match the exact alcoholic-beverages category. Non-alcoholic sodas and drinks can now receive disclosed estimates. - Plain single-ingredient rice with a rice category is treated as a cereal in estimates, despite the source's broad seeds/oilseed classification. Mixed rice/seed foods are not automatically reclassified. - Lundberg Black Pearl Rice 1 lb (073416542600) uses its matching manufacturer package reference in the estimate: per 45 g dry, 160 kcal, 1 g fat, 0 g saturated fat, 0 g sugar, 3 g fiber, 4 g protein, 10 mg sodium, 33 g carbohydrates. The manufacturer package shows USDA Organic. Original OFF data is retained. This is not verification of a user's specific bag or a cooked serving. - Reference: https://www.lundberg.com/products/organic-black-pearl-rice-1lb and its linked back-panel image, reviewed September 14, 2026. - Black/whole-grain rice ingredient terms now match the ordinary culinary classification. This is not a blanket ingredient safety guarantee. - Estimates are still withheld for unidentified records, too few usable nutrition inputs, unsupported special-purpose foods and explicit prepared-only data. An asterisk is an estimate, not a claim of measured accuracy or an official Yuka rating. ## Previous releases # Estimate beta 2 — September 13, 2026 This section supersedes beta-1 behavior below for the cases described here. - A product named vodka pasta sauce with tomatoes in its ingredient list is evaluated as a food in the estimate, even when mistakenly categorized as spirits. Its original database category is preserved. This describes sauce as sold, without added cream, butter or pasta. - A named protein bar with a corresponding bar-category tag can receive a provisional general-food estimate despite a supplement tag. Powders, other supplements, infant/medical products and actual alcoholic drinks remain excluded. - Salt/sodium disagreements no longer trigger an uninformed midpoint across the entire salt scale. When both are valid numbers, the working estimate uses the higher supplied salt-equivalent, and the possible span includes both supplied values. This is a cautious assumption, not a validated correction. - Conflicting kcal and kJ values are both represented in the possible span, including effects on protein eligibility. The working estimate continues to prefer listed kcal. - Nutrition-only database errors do not independently invalidate ingredient assessment. Missing ingredients, ingredient errors, and unknown compounds still remain unresolved. Yellowfin tuna, albacore tuna, skipjack tuna and squid are recognized culinary terms; this does not assess mercury, contaminants, allergies or individual seafood suitability. - Safe Catch scanned UPC 859480006596 uses the manufacturer's matching Wild Ahi Yellowfin 5 oz can label as a reference: 110 kcal, 0.5 g fat, 0 g saturated fat, 0 g total sugar, 0 g fiber, 26 g protein and 220 mg sodium per 85 g. The page does not explicitly identify the scanned barcode, so the package match remains provisional. Source reviewed September 13, 2026: https://safecatch.com/buy-safe-catch-tuna/wild-ahi-yellowfin-tuna/ ## Reconciling the latest test scans Ligo remains 53*. Its recorded sodium, sugar and protein determine the nutrition component; being canned does not itself create a penalty. Its ingredient list is missing. No exact Yuka score was verified, so the number was not lowered to fit an expectation. Carbone Spicy Vodka sauce now receives 31*, using food rather than beverage rules. The scanned 32 oz identity is supported by the retailer listing: https://www.shoprite.com/sm/delivery/rsid/350/product/carbone-world-famous-spicy-vodka-sauce-32-oz-id-00850024735453 Safe Catch moves from 26* to 85*. The former estimate discarded conflicting salt data and substituted ten salt penalty points, which also removed protein credit. Nutrition errors also blocked its simple ingredient list. The new reference and error handling address both issues. The revised result is not a verified Yuka match. Kirkland Chocolate Peanut Butter Protein Bars now receive 70*, with a wide possible span because the kcal/kJ data conflict and ingredients remain incompletely reviewed. Protein and fiber help the working nutrition score; this is not an endorsement of every ingredient. Costco identifies the product as protein bars: https://www.costco.com/p/-/kirkland-signature-protein-bars-chocolate-peanut-butter-chuck-212-oz-bars-40-count-4-pack-of-10/100423984 ## What a Yuka comparison can establish Yuka publishes 60% nutrition, 30% additives and 10% organic weights, plus its own nutrition conversion and ingredient assessments. Sharing these weights does not make our estimates equivalent. We use a provisional midpoint for incomplete ingredient assessments; this is an important source of differences. Our score does not include a product-specific mercury analysis or personalized dietary advice. Published Yuka methodology: https://help.yuka.io/l/en/article/ijzgfvi1jq Nutrition conversion: https://help.yuka.io/l/en/article/owuc9rbhqs No exact live Yuka record or product score for these four UPCs was verified. The public simulator loaded with broken labels and was not used as a reference result. User recollections of Yuka scores remain comparison leads, not target values used to tune the formula. ## Earlier beta-1 policy (where not superseded above) # Food Scanner scoring policy Release: estimate-beta-1, September 13, 2026. Complete assessments use overall-beta-4. This is an independent experimental comparison tool, not Yuka's score, a validated health measurement, or an allergy assessment. ## What the number means Nutrition contributes up to 60 points, reviewed ingredients up to 30, and verified organic certification 10. The conversion and ingredient weights are our design choices. Missing organic evidence earns no bonus; that is not a claim that the food is non-organic. Products with D/E nutrition are capped at 49 overall. Compare similar foods; portion size and your individual needs matter. A number without an asterisk has complete inputs under the current checks. It is still based on potentially outdated community records. A number with an asterisk is a provisional working estimate. Its expandable explanation lists the assumptions and sources. It does not mean a professional reviewed your package. ## How estimates work 1. Retain any nutrition or ingredient component that already passes the complete assessment. 2. Otherwise recalculate nutrition from usable product amounts with the 2023 general-food, beverage, cheese, red-meat or fats/nuts/seeds thresholds. If a category is inferred, disclose it. Mayonnaise uses general-food rules. Listed kcal is preferred and converted to kJ. Do not use explicitly prepared amounts. 3. Missing negative factors use the rounded midpoint of their possible penalty points. Known total carbohydrate or total fat can narrow the corresponding sugar or saturated-fat bound. Missing fiber and protein earn no favorable credit in the working estimate. No nutrient amount is written into the source record. 4. Use an explicitly supplied produce percentage when available. A source-estimated percentage is identified as an assumption. Otherwise the working number grants no produce credit and the possible span covers all percentage bands. An existing exhaustive range from the complete calculator uses equal weight for each band in its provisional nutrition component. Equal weights are a policy choice, not estimated probabilities. 5. If ingredient review is incomplete, the unresolved ingredient component uses the midpoint between zero and its remaining maximum, after any known caution. With no known caution this is 15/30. This neither proves harm nor grants a safety clearance. Completing the review can move the score up or down. 6. Show the possible total span from unresolved scoring points, with each scenario's grade cap. This is conditional on the listed inputs and assumed category. It does not capture every possible error in the food record and is not a statistical confidence interval. The confidence label is a heuristic. Low means a span over 30 points, fewer than six usable core nutrition inputs, or a retailer-label correction. Moderate describes other estimated cases. Neither label is a measured accuracy rate. No AI model or comparable-product model has been trained or validated for these estimates. Unidentified records, fewer than three usable core nutrition inputs, explicitly prepared data, supplements, infant/medical foods, alcohol and pet food do not receive an automatic estimate. This release therefore does not guarantee a score for every barcode. Store filters cannot supply missing nutrition. ## Reviewed product evidence - Bachan's Original Japanese Barbecue Sauce, 34 oz, scanned UPC 850006883059: the matching Costco listing gives 35 kcal, 7 g total sugar, and 520 mg sodium per 19 g. The database record has conflicting values. The estimate uses the retailer's gram-based serving conversion, with its inconsistent tablespoon wording disclosed. The original database facts remain visible and explicitly identified as conflicting. Review date: September 13, 2026. https://www.costcobusinessdelivery.com/p/-/bachans-japanese-barbecue-sauce-original-34-oz/100458754 - R.W. Knudsen Carrot Ginger Turmeric, UPC 074682200300: the manufacturer identifies this UPC and USDA Organic certification. The estimate includes that certification bonus, while nutrition remains from the database and ingredient differences remain unresolved. Review date: September 13, 2026. https://rwknudsen.com/products/p/veggie-juices/organic-carrot-ginger-turmeric-beverage - Kewpie's manufacturer lists its 24 fl oz USA version at Costco. Its database ingredient review remains incomplete; the estimate does not treat MSG, EDTA or natural flavors as fully assessed. https://www.kewpieshop.com/collections/kewpie-mayonnaise/products/kewpie-japanese-style-mayonnaise-24oz-product-of-usa These source reviews are tied to identified products, not applied to every product from a brand. They are bundled with app releases and do not automatically refresh from retailer websites. ## When scores change Ordinary barcode lookups and Refresh product data request the latest database record and rerun the current policy. Better inputs can replace the estimate with a complete score automatically. There is no saved score history or scheduled background recheck. Reviewed catalog entries and retailer evidence change when reviewed and published in an app update. Release labels distinguish the current website from older versions. ## Ingredient evidence and method The complete assessment uses reviewed rules for phosphates, lecithins, xanthan, calcium carbonate, citric acid, gelatin and maltodextrin. Each identified rule has its source and review date in the product's ingredient evidence. Unreviewed families remain explicitly provisional. Organic status, NOVA processing, and nutrition are separate concepts; none establishes personal safety. Published 2023 method: https://www.santepubliquefrance.fr/nutrition-et-activite-physique/article/nutri-score-2023-update Formula thresholds were cross-checked against Open Food Facts' implementation: https://github.com/openfoodfacts/openfoodfacts-server/blob/main/lib/ProductOpener/Nutriscore.pm Data: https://world.openfoodfacts.org/ (ODbL / DbCL). Store filters use community store tags, which can be incomplete or outdated. They are not stock checks and never alter a product's score.