Processing
NOVA data and ingredient-list signals help establish the starting point. Label evidence can challenge an incomplete or inconsistent database classification.
Scoring methodology
A useful score needs reasons and limits.
Lychee uses a processing-first food model and a separate beta cosmetic ingredient model. Both depend on the information available; a missing or uncertain label should not produce fake confidence.
Free download · Optional Premium · iOS & Android
Food scoring
Processing comes first. Nutrition still matters.
Lychee’s food Health Score runs from 1 to 100. Whole and minimally processed foods should generally score higher than heavily engineered products. A good-looking macro profile should not erase evidence of ultra-processing.
Ingredients and additives
Additives, industrial formulation markers, seed oils and ingredient quality can affect the result. Front-of-pack health claims do not override the ingredient list.
Nutrition
Sugar, salt, saturated fat, fiber and protein are considered alongside processing. Lychee displays available nutrition, including energy; it is not a calorie-tracking program.
There is no single fixed percentage formula for every product. Category-specific rules, penalties, caps and data checks are part of the model. Available NOVA data and ingredient-based estimates inform processing; Lychee’s own rules determine the food Health Score.
Reading the result
A comparison score, with reasons behind it.
| Score | Label | How to use it |
|---|---|---|
| 75–100 | Excellent | Higher-scoring label profile within this model. |
| 60–74 | Good | Review the breakdown and your own requirements. |
| 35–59 | Average | Inspect processing, ingredients and nutrition tradeoffs. |
| 1–34 | Poor | Review the reasons and compare available alternatives. |
A high food score does not override an allergen alert. A low score is not a diagnosis. Compare products used for similar purposes and read the reasons, not just the number.
Some products do not receive a food score. Insufficient information, supplements requiring dose-specific context, and pure water are examples where a comparative number may not be useful.
Cosmetic scoring · beta
The formula and how you use it both matter.
Cosmetics use a separate model. Lychee recognizes INCI ingredients, explains their functions and concern categories, and considers product-use context such as rinse-off, leave-on, lip, oral, eye-area and fragrance exposure.
- Ingredient coverage is checked before a score appears. Recognized and unknown ingredients remain useful information even when no score is available.
- The general coverage gate requires at least three ingredients and 75% recognition. Additional checks can withhold a score for a short OCR extract, mismatched product data or poor confidence; selected single-ingredient formulas are an exception.
- When ingredient coverage is strong but product type is uncertain, Lychee can use a cautious exposure assumption instead of automatically blocking the score.
- Concentration, the finished formulation and personal sensitivity cannot be fully determined from an INCI list. A flagged ingredient is a reason to read more, not proof of harm.
Data and interpretation
Keep the inputs and interpretation separate.
Product label
Barcode records come from open product databases or text read from a label. Neither proves that the current package has the same formula.
Scoring model
Lychee applies its own rules to available inputs. The number summarizes those rules; it is not a regulator’s approval or a clinical assessment.
AI and estimates
AI explanations and contaminant estimates provide additional context. They cannot measure contamination or establish individual medical risk.
Published laboratory results, when present, should be read separately from estimates and tied to the specific tested product and source. Scoring evolves as data and implementation improve; old saved results may differ from a fresh analysis.
Common questions
Before you scan.
How does Lychee score food?
Lychee uses a 1–100 food score that prioritizes processing. Additives, ingredient quality, sugar, salt, saturated fat, fiber and protein also matter. Available NOVA data and ingredient-based estimates inform processing; Lychee applies its own scoring rules. Missing data can reduce confidence or prevent a score.
Does Lychee scan cosmetics and skincare?
Yes. Lychee scans beauty and personal-care barcodes and ingredient labels. Its beta cosmetic score considers recognized ingredients, concerns and exposure, including rinse-off versus leave-on use. It shows Not enough data when ingredient coverage is insufficient.
Does Lychee use AI to calculate every score?
The food and cosmetic scores use scoring logic and ingredient dictionaries. Optional AI assistance adds explanations and analysis; it is not a laboratory test or a guarantee that a label is accurate.
Can Lychee detect heavy metals or microplastics?
A barcode or label photo cannot measure contaminants. Where Lychee presents estimates or published lab results, those are different forms of evidence. Check the source, tested product, date and limits before drawing a conclusion.
Keep exploring
More detail, when you need it.
Get Lychee
Start with your next label.
Food and beauty ingredient context in your pocket.