Skip to content
Meal Monster

Logging Friction Index · protocol draft

A meal log should include the whole meal.

We are measuring what happens between describing food and finishing a usable log: missing foods, correction actions, uncertain portions, and time.

Inputs are available. Results are pending.

The corpus contains three phrasings of each scenario: plain, conversational, and informal. Those variants are related observations, not three independent meals. Informal text simulates spelling noise; it is not recorded speech.

These inputs were authored for testing. They do not contain private customer logs and do not establish exact calorie truth. The label-arithmetic cases use synthetic declared numbers and test arithmetic fidelity only.

Explore a meal scenario

210 scenarios. JavaScript loads the full explorer.

breakfast

Two poached eggs, rye toast, and a pear.

Portions or ingredients remain uncertain; do not assign exact nutrition truth.

breakfast

A bowl of plain yogurt with raspberries and pumpkin seeds.

Portions or ingredients remain uncertain; do not assign exact nutrition truth.

breakfast

Three small pancakes with maple syrup and sliced strawberries.

Portions or ingredients remain uncertain; do not assign exact nutrition truth.

breakfast

An English muffin with an egg, cheddar, and tomato.

Portions or ingredients remain uncertain; do not assign exact nutrition truth.

breakfast

Cream of wheat made with milk, topped with raisins.

Portions or ingredients remain uncertain; do not assign exact nutrition truth.

breakfast

A spinach and mushroom omelet with feta and a peach.

Portions or ingredients remain uncertain; do not assign exact nutrition truth.

What we measure

First-result food coverage
Which expected foods appear before corrections. Errors stay in the denominator.
Correction burden
Repair actions and corrected fields, with navigation and mandatory confirmations counted separately.
Completion and time
The share of attempted meals completed, alongside median completion time and variability. Fast failures do not win.
Uncertainty and exclusions
Whether unclear portions remain explicit and whether ingredients the user excluded are mistakenly added.

Conditions before publishing results

Earlier work and our scope

Logging-effort research already exists. MacroFactor described comparing 20 food loggers in 2022, and MyNetDiary publishes a 127-entry comparison. Our proposed contribution is a reusable corpus emphasizing food omissions, correction effort, ambiguity, and repeated runs. It is a testable contribution, not a claim of being first.

Current machine-readable results contain no measured runs yet. Corpus JSON is available under CC BY 4.0 with its version and checksum.

Compare advertised logging methods.