How We Score
Every Australian product in Trulu's database receives a score between 0 and 100. Here's exactly how that number is calculated - tuned to local labels and FSANZ standards.
The four scoring dimensions
Trulu's model evaluates four dimensions for every product. Each dimension contributes to the final score, weighted according to the product category.
AI assists with ingredient parsing, cross-referencing scientific literature, and generating draft analyses at scale. The Trulu team audits scores, reviews flagged products, and refines the rules whenever we find a gap.
1. Ingredient Quality
Every ingredient in a product is classified against a database of thousands of individual ingredients. Each is assessed for health impact based on peer-reviewed research and regulatory guidance from FSANZ and EFSA. Whole foods, quality fats, and unprocessed ingredients push scores up. Hydrogenated oils, high-fructose corn syrup, and ingredients with established health concerns pull them down.
2. Additive Flags
Additives are evaluated individually, not as a group. Ascorbic acid (vitamin C) is benign. Certain azo dyes are not. We flag preservatives, artificial colours, emulsifiers, and flavour enhancers based on the strength of evidence for health concerns. Each flag affects the score proportionally to the severity of the evidence.
3. Nutritional Profile
We evaluate sugar, sodium, saturated fat, fibre, and protein content. These are weighted relative to the product category — the acceptable range for a condiment is different from a breakfast cereal. Products with favourable nutritional profiles for their category score higher.
4. Processing Level
We draw on the NOVA food classification system to assess processing level. But processing alone doesn't determine the score. A moderately processed product with excellent nutrition can outscore a minimally processed one with poor nutritional content. Processing is one signal among several.
Category-relative scoring
Scores are relative to the product's category. A score of 75 for a pasta sauce means it's among the better pasta sauces in our database — not that it competes with raw vegetables. This keeps comparisons meaningful: products are always compared within their own category.
Category weighting also adjusts how much each dimension matters. For breakfast cereals, sugar content is heavily weighted. For supplements, bioavailability of the active ingredient matters more. For personal care products, the additive safety profile dominates.
Score updates
Scores are recalculated when product formulations change, when new research becomes available, or when we refine the model. See our changelog for a record of every meaningful update.
Honesty about our limits
We do everything we can to score products fairly and to reduce errors - but we want to be straight with you: scoring tens of thousands of products at scale means we can't guarantee that every score, ingredient breakdown, or analysis will be perfect at every moment in time.
Trulu is built and maintained by a small Australian team managing a database that's growing by the day. Our pipeline uses AI to do the heavy lifting on parsing and analysis, and we prioritise human review for flagged products, edge cases, and high-impact categories. Realistically, it isn't possible for us to manually inspect every single item before it goes live - and we'd rather be honest about that than pretend otherwise.
When we get something wrong, we want to fix it quickly. If you spot an inaccurate score, a missing ingredient, an out-of-date label, or anything that doesn't look right, please use the Report an issue or Send feedback buttons in the app. Every report gets reviewed and we do our best to get it corrected.
Help us keep it accurate
A project of this scope - thousands of products, constantly reformulating, across food, drinks, supplements, beauty, and personal care - really only works with help from the people using it. The community is part of how Trulu stays accurate.
If you scan something and the data looks off, flag it. If you know a category well, tell us where the model is missing nuance. If a product has reformulated and the score hasn't caught up, let us know. The more eyes on the database, the better it gets for everyone.
For more on how we handle reports, see our corrections policy.