Methodology — The Lunch Index

What this measures: “What are you having for lunch today?” — What visitors to this site actually pick when asked the question above — nothing more. It is a running poll, not a quality rating: the ranking is the count of picks.
Who answers: Visitors to this site, unpaid and self-selected — a convenience sample, not a probability sample. The question is about everyday decisions in Korea, so respondents skew toward people living in or engaged with Korea, and the results carry that skew.
Where the options come from: Inherits the lunch-appropriate menus from the Korea Consumer Agency dining-out price survey (the same source as our Kimchi-jjigae Index), plus the common office-district lunch categories it misses (gukbap, donkatsu, jeyuk, burgers, salads, Japanese). Dinner-leaning menus such as samgyeopsal are excluded.
When the ranking opens: No order is shown until the poll has collected 100 votes in total. The number was fixed before any data existed. Until then the page shows how many votes have come in — and nothing else, because showing an order built on a handful of votes would be exactly the made-up listicle this site exists to replace.
Formula: Plain vote counts, shown with each option's share of the total. No weighting, no smoothing, no editorial adjustment.

Limits worth knowing before you cite this

This is a poll of this site’s visitors, not of Korea. Nothing about the sample is representative, and we label it that way everywhere it appears.
Duplicate protection is basic on purpose. One vote per index per browser session. A private window, another device, or clearing cookies gets around it — the alternatives (accounts, CAPTCHAs, fingerprinting) cost visitors more than the abuse costs us.
Small gaps are ties. Options separated by a few votes should be read as tied, especially early on.