Methodology
Everything on this site is computed from public review data, re-harvested and re-verified in full on every run (current run: August 2026). No restaurant pays for placement; no score can be bought.
Sources
- Naver Place — visitor review score and volume, blog/café review volume, menus with prices. The dominant Korean review platform; volumes routinely 10–100× the others.
- KakaoMap — star rating and count, blog review count, menus, price level. Korea's second local signal; Naver Place itself no longer shows star ratings (phased out in 2021), so Kakao supplies the numeric local star.
- Google Places — rating and review count. The tourist verdict: the overwhelming majority of visitors to Korea rate restaurants here.
Matching
The Kakao listing is the anchor (name, both address forms, coordinates). The same branch is located on Naver and Google by name similarity plus coordinate distance — matches beyond 400 m are rejected, so a same-named branch across town can't pollute a score. Match confidence is stored per record and shipped in the open dataset.
The scores
- Local score = weighted blend of Naver visitor score (2×) and Kakao rating (1×), 0–5.
- Tourist score = Google rating, 0–5.
- Gap Score = tourist − local, rounded to two decimals. Positive = tourists rate it higher than locals; negative = locals rate it higher.
- Qualification: gap badges require ≥100 reviews on both sides. 225 of 352 listed restaurants currently qualify; the other 127 appear in the district tables marked thin data and get no page of their own.
- Buzz = log₁₀ of total local review volume — how much locals talk about a place, independent of sentiment. District pages sort by it.
Badges
The five bands are half-open and jointly cover the whole range, so every gap score lands in exactly one of them and none can fall between two. Counts are current as of the August 2026 run.
| Badge | Band | Means | Now |
|---|---|---|---|
| Trap alert | gap ≥ +0.45 | tourists rate it far above the local consensus | 5 |
| Tourist-tilted | +0.20 ≤ gap < +0.45 | a smaller gap, same direction | 20 |
| Aligned | −0.20 < gap < +0.20 | both sides broadly agree | 170 |
| Local favorite | −0.45 < gap ≤ −0.20 | locals rate it above the visitor consensus | 28 |
| Hidden gem | gap ≤ −0.45 | locals rate it far above the visitor consensus | 2 |
The two index pages list both tiers on their side of zero — the Tourist Trap Index carries 25 rooms of which 5 clear the Trap Alert threshold, and Hidden Gems carries 30 of which 2 clear the Hidden Gem threshold. Every count on this site is generated from one function over one dataset, and the build refuses to publish if two surfaces disagree.
Caveats we're honest about
- Platform cultures differ: Korean reviewers grade harder than Google's global average. The gap measures disagreement, not objective quality.
- Kakao star ratings skew low on famous places (locals rate the queue, not just the food). That's part of the signal, not a bug — but it's why Naver gets 2× weight.
- Cafés and dessert bars are excluded from v1 — this is a restaurants index.
- TripAdvisor is not yet included; it's on the roadmap as a second tourist signal.
- Menu prices are as listed on Naver/Kakao at harvest time and can lag reality by a few weeks. Rows the platforms publish without a real price — promotional lines such as “free corkage” or “air-conditioning on” — are dropped rather than shown as ₩0.
- The Gap Score is a measure of two crowds disagreeing. It is not a rating of the restaurant, and we never publish it as one: the structured data on this site carries no aggregate rating, because the ratings belong to Google, Naver and Kakao rather than to us.
Verification
Every record carries the date it was last checked and whether its Naver page is live. Closed restaurants stay in the dataset (flagged) rather than silently vanishing — a public closures ledger is coming with the next data drop. Corrections are welcome and logged: see how to dispute a badge.