The compliance & quality model
How PIE evaluates products safely against pluggable regulatory frameworks and completeness profiles, per locale.
PIE treats regulatory compliance and data quality as first-class, evaluable properties of a product — not documentation someone maintains by hand. Both are computed from the product's data against rule sets, per locale, and can gate publishing.
Safe, no-eval evaluation
At the core is a safe condition and validator engine. Rules are expressed as conditions and validators that a purpose-built parser evaluates — there is no eval, ever. This matters because rule sets come from regulatory frameworks and, for quality, from organisation authors: the engine can evaluate untrusted rule expressions without executing arbitrary code. The same evaluation primitives are reused wherever safe rule evaluation is needed, including workflow data conditions.
Frameworks are pluggable
Compliance frameworks — EU FIC, EU Organic, EUDR, FSMA 204, HACCP, IFS/BRC, EU Health Claims, Novel Food, Packaging, DPP/ESPR — are pluggable rule sets. A product is evaluated against the frameworks relevant to its markets, producing per-rule results and an overall score.
Everything is per locale
Because PIE is multilingual and multi-market, evaluation is locale-aware. A product carries required locales derived from its target markets, and it is scored per locale with fallback disabled — a missing translation counts as missing. Compliance is scoped to the market too, so German, Austrian, and Swiss variants evaluate against the right rules. Editing a translation or changing markets re-scores the product, so a score is never stale.
Quality mirrors compliance
Data-quality completeness profiles are authored by organisations and evaluated with the same primitives as compliance. A profile can be blocking, gating publish when a mandatory rule fails in any required locale.
Fail closed
Every gate fails closed. If a required check cannot be evaluated — a malformed rule, a missing snapshot, a pinned-but-inactive profile — the publish is denied rather than allowed through. The safe default is "don't publish."
Related
- Quality & compliance guide — using it day to day.
- Workflow guide — data conditions reuse these primitives.