Allergen management for food manufacturers

Catch allergen cross-contact in the schedule, not on the swab.

SavorCline checks every changeover in your production plan against what the previous run leaves on the line and the cleaning recorded in between. Your QA team sees unsafe sequences, missing cleaning records and a safer run order before the shift starts.

Interactive preview · runs in your browser on a fictional bakery's sample data

The problem

Allergen risk lives in the changeover.

Most allergen controls are designed product by product. The risk appears when products share a line: a hazelnut brownie at 06:00, a plain loaf at 08:00, and a cleaning record that may or may not say what was done in between.

    Changeover anatomy

    Run A · 06:00Hazelnut praline brownie
    Run B · 08:00Plain white loaf
    left on line = allergens(A) − declared(B)
    1. Was there a clean? Any record between 07:30 and 08:00 on this line.
    2. Was it the right clean? Dry brushing or sanitising does not remove allergen protein.
    3. Was it verified? Sign-off and a passing allergen swab.

    Product

    Five checks between the supplier spec and the first run.

    Each capability works on the same allergen profiles and schedule, so a change in one place is re-checked everywhere it matters.

    How it works

    Start with documents you already have.

    No sensors, no factory hardware, no MES integration to begin with. SavorCline works from specifications, schedules and cleaning logs, and people make the final call.

    1. 01

      Add products and specs

      Upload or paste supplier specifications and recipe sheets. Allergens are extracted, including hidden sources such as whey, spelt or tahini, and shown for review.

    2. 02

      Confirm allergen profiles

      A QA team member confirms each product's "contains" and "may contain" allergens. Confirmed profiles become the reference for every check.

    3. 03

      Check the week's schedule

      Every changeover on every line is assessed against the cleaning records. Risky transitions are ranked high, medium or controlled.

    4. 04

      Approve and report

      Accept or reject the suggested run order, then generate a dated report with findings, required cleans and a sign-off block.

    Solutions

    Built first for bakery and confectionery.

    Shared ovens, depositors and enrobers, five or more allergens in rotation, and several changeovers per shift make bakery and confectionery the hardest case. Patterns proven there carry over to other categories.

      Technology

      What runs today, and what we're building on AWS.

      Working today

      In-browser risk engine

      The changeover assessment, sequencing search, cleaning verification and report generator are deterministic JavaScript with an automated test suite. The demo runs entirely in your browser; nothing you enter leaves it.

      • 14 UK/EU regulated allergens, configurable
      • Exhaustive sequence search up to 8 runs per line
      • Rule-based ingredient synonym matching
      Proposed architectureNot yet deployed
      AWS servicePlanned role in SavorClineStatus

      Pricing

      Pilot programme in preparation.

      Pricing is not yet published. SavorCline is preparing a pilot with bakery and confectionery sites, where pricing will be agreed per production site. Until then, the demo workspace is free to use with your own product list.

      • Allergen profiles for your product range
      • Weekly schedule checks across your lines
      • Cleaning record gap review
      • Report format agreed with your QA team

      FAQ

      Questions QA teams ask first.

      Does SavorCline replace our HACCP plan or allergen risk assessment?

      No. It applies your allergen data and cleaning rules to the schedule consistently and shows where they are not met. Your HACCP team still owns the controls, the validation of cleaning methods and the final decision on every run.

      Does the AI decide the production order?

      No. SavorCline suggests a run order that needs fewer allergen cleans. A planner or QA lead accepts or rejects it. In the planned AI document pipeline, extracted allergens are likewise shown for human confirmation before they are used.

      Which allergens are covered?

      The demo uses the 14 allergens regulated under UK and EU food information law, including hidden sources such as whey (milk), spelt (gluten) and tahini (sesame). The list is configurable for other markets.

      Is SavorCline available now?

      SavorCline is building its initial product. The demo workspace on this site is fully interactive and runs on sample data or data you enter, stored only in your browser. Document upload and the AWS-hosted platform are in development.

      What data do we need to start?

      A product list with allergen information (or the supplier specifications to extract it from), a production schedule per line, and your cleaning records for the same period.