How we rank restaurants

Short answer

Our rankings are not a single star rating: they combine dish-level review signals (how often a dish is praised), current prices, overall platform ratings and freshness. A confidence-weighted (Bayesian) score keeps a one-review "5.0" from beating a well-reviewed "4.5".

Data sources

Four delivery platforms + Google — each source stored separately.

Restaurant, menu and price data is scanned daily from Yemeksepeti, Getir, Trendyol GO and Migros Yemek. The same restaurant's listings are merged into one profile via name + branch + location matching. For famous dine-in places without delivery, photos and reviews come from Google and are shown as a separate badge.

Flavor score (Bayesian)

Few perfect reviews should not beat many good ones.

For a ranking like "best pide in Kadıköy", each restaurant has a dish-level average (R) over n reviews. The score is n / (n + M) · R + M / (n + M) · C: with few reviews it leans on the restaurant's overall flavor prior (C); with many, its own average dominates. No evidence for a dish means the restaurant does not enter that ranking at all.

Prices and freshness

Price = lowest current portion across platforms.

Menu prices are refreshed every morning; the price you see is the lowest current one across the four apps, shown with its last-seen date. Prices can change on the platforms at any time — the price at checkout is the one that counts.

Data sufficiency

We do not publish thin pages.

Every programmatic page is scored on restaurant count, priced menus, dish-mention reviews and independent sources before it is allowed into search indexes. Pages below the threshold stay out. Fewer, trustworthy answers — not more, empty ones.

Independence and limits

We say whose ratings they are.

Rankings cannot be bought; any sponsored visibility will be clearly labeled. We never mark third-party platform ratings as our own in structured data, and we do not copy review text — we use signals derived from reviews.