Case studyInteractive prototype
Allergen Matching System
An allergy check that shows its working.
A mobile prototype that compares menu ingredients and environmental exposures with a demo allergy profile, then explains each match.
- Project
- Allergy profile and food checking application
- Designed for
- Diners and treating allergy clinicians
- Focus
- Profile provenance, food checks, places, and a personal journal
What it does
Allergen Matching System explores how a person could use their allergy record while choosing food or visiting a place. The prototype brings the profile, food checks, places, and a personal journal into one mobile experience, alongside a separate clinician view of exposure history.
The problem
A warning is useful only when someone can understand what raised it. This project needed to connect an ingredient to the relevant profile entry, show where that entry came from, and distinguish a listed ingredient from a preparation concern such as a shared fryer.
The interface also needed to communicate uncertainty without presenting a meal as verified.
Our approach
We built a deterministic matching engine and an interface that exposes its reasoning. Each flag carries the ingredient, matching profile entry, and confirmation details. Direct ingredient matches and cross-contact concerns appear separately.
Four risk assumptions use words and icons alongside color, with a reminder to confirm food information with staff. Constructed venues and a fictional profile make the different matching cases repeatable in the demo.
Research and context
The FSA/Ipsos survey found that confidence in avoiding problem ingredients varied by ordering setting. Separately, FSANZ distinguishes ingredient declarations from voluntary precautionary warnings. These findings inform the questions around food information; they do not validate this prototype’s checks. [1][2]
Confidence varies by ordering setting
Adults with food hypersensitivity in England, Wales, and Northern Ireland; FSA/Ipsos, December 2024–February 2025. Percentage confident they could avoid ingredients they react to.
Preparing chart. The exact figures are available below.
| Measure | Value |
|---|---|
| Table service | 84% |
| Counter service | 73% |
| Deli or bakery counter | 65% |
| In-store digital ordering | 63% |
| Business website or app | 62% |
| Delivery website or app | 57% |
| Telephone takeaway ordering | 56% |
Weighted survey of 780 adults with food hypersensitivity, including allergy, intolerance, and coeliac disease. Each setting was a separate question.
Perceived confidence is not measured food safety. This sample does not establish Australian prevalence or clinical outcomes, and the chart is not a result of using Allergen.
Source [1]
What was built
A profile with provenance
Confirmed and self-reported entries show their sources and the information behind them.
Places and menu checks
Search, filters, a map, and venue detail screens show constructed menus and environmental exposures matched to the demo profile.
A food checking flow
Type known ingredients or run a labeled sample scan, then inspect the reason for each flag.
A personal journal
Written entries, symptom selections, and completed food checks form a chronological record of what the person entered.
A clinician exposure view
A separate screen presents recorded scans, near misses, and reactions, with event counts and supporting details.
Environmental context
Simulated pollen readings are matched to the profile; air quality appears as separate context.
Product walkthrough
Explore the demo profile
Open My allergens to see confirmed and self-reported entries and the sources behind them.
Run a sample check
Use the labeled simulation to compare a sample dish with the profile, then read the ingredient trace and any preparation concerns.
Review the journal
Return to the personal notebook to see written entries and recorded food checks using fictional demo data.
Chapter 1
Chapter 2
Chapter 3
The outcome
The result is a working showcase of an explainable allergy checking workflow, from profile provenance to the reason behind a food flag. Constructed data illustrates direct matches, preparation concerns, and checks with no matching entry.
Scope and next steps
This prototype uses a fictional patient and constructed venue data. Its risk matrix requires clinical review before real-world use, and a below-risk result does not verify a meal or a kitchen.
A separate tester-gated photo pathway uses a model to read visible information before the existing engine performs the comparison. Real-label reading quality has not been measured. The walkthrough uses the deterministic sample scan.
A production version would require clinical review, reliable data sources, and account and backend systems. The personal journal is separate from the seeded clinician exposure log.
Sources
Food Standards Agency / Ipsos UK · October 7, 2025; updated September 3, 2026
Allergen information for non-prepacked foods: consumer experiences, behaviours and attitudesSection 3.3, Figure 3: confidence by ordering setting and survey methodology.
Food Standards Australia New Zealand · Updated November 26, 2025
Allergen labelling for consumersIngredient declarations, precautionary warnings, and obtaining information for unpackaged food.
Project evidence
The product description comes from reviewed prototype code and the recorded demo shown above. Profiles, foods, venues, and clinician records are fictional. Its risk rules have not been clinically validated; a check is not permission to eat, and preparation details still need confirmation with staff.
Have a similar problem to solve?
Tell us how your business works and what you want to improve. We can shape the next step together.