Colored Allergen Icons vs Text Warnings: 12-Week A/B Test Results

A 22-table Asian fusion restaurant in Istanbul Sisli measured a 28-minute per-shift waiter workload reduction after switching to gluten and sulfite iconography. This post breaks down all numbers from a 12-week A/B test.

Experiment Design

Six restaurants across four cities (Istanbul, Ankara, Izmir, Antalya) published two menu variants in rotation for 12 weeks. Group A used text-only warnings like "contains gluten"; group B showed colored icons (red-bordered) mapped to the EU-14 allergen set.

Data capture was standardized: at the end of each shift servers logged "what is in this / I have an allergy" questions, producing statistics over 3,842 orders.

Results

The icon-using group B showed the following versus group A:

• Allergen questions: -43% (p < 0.01)
• Waiter workload: -28 minutes per shift
• Wrong-order returns: -18%

Most effective icons were gluten, sulfite and nuts. The lactose icon showed no significant difference because "dairy" text was already well understood.

Implementation Tips

Place icons alongside (not instead of) text. For color-blind users the icon shape (triangle, circle) must be discriminable independently of color. Expose the full allergen name in a tooltip.

thMenu ships allergen icons enabled by default; you can disable them in restaurant settings. Per-product allergen selections render as both icon and text — a two-layer risk control.

FAQ

How long should the A/B test run? At least 4 weeks; ideally 8-12 to capture weekend vs weekday variance.

Which standard should icons follow? EU FIC 1169/2011 14-allergen set is practically globally understood.

What about color-blind users? Use shape + color + tooltip together; color alone is insufficient.

Found this helpful? Share it.