The challenge
The founders saw a market saturated with marketing claims and very little objective signal. Existing apps tended to score products through opaque rules that treated every ingredient identically regardless of context, role or concentration. The result was that two products with very similar formulations could be rated wildly differently across services, which eroded consumer trust in any rating at all.
The brief was to build something the team could defend on the science: a curated ingredient database with sourced evaluations, a scoring model that accounted for an ingredient's role and typical concentration in the formulation, and a product layer that turned that into a clear grade for the consumer without hiding the reasoning behind it.
Our approach
We built the platform as a multi-tenant SaaS product through our custom SaaS development practice, with the analysis pipeline scoped by our machine learning solutions team. The data foundation was the first priority — an ingredient database with versioned evaluations, citations to primary sources where they exist, and a clear escalation path for ingredients flagged as ambiguous.
The scoring model combined a deterministic core with AI assistance. Deterministic rules governed safety-critical evaluations and known interaction effects, so the same input always produces the same output. AI was used in two places: ingestion of product label data into a normalised representation that the rules could operate on, and the natural-language explanation that helps the consumer understand why a product received the grade it did. Every AI step shipped with evaluation data and human review queues.
The roadmap was sequenced through our AI integration consulting approach so the team could ship a credible v1 quickly, then layer in personalisation and brand-side analytics on the same data spine.
Outcome
The platform launched with a database large enough to grade the great majority of mainstream products in the target region, and with a grading model the team could explain in detail to a sceptical journalist or a brand owner. Early consumer feedback consistently highlighted the transparency of the grade rationale rather than the grade itself, which is exactly the trust signal the founders had set as the success criterion.
The platform is now positioned to extend into B2B analytics for brands that want to understand how their formulations compare on the same model.
Stack used
Duration
approximately 7 months for v1