How LocalSelections chooses and ranks nearby places

When a wheel lands on "Italian" and LocalSelections shows you a few real restaurants near you, those results do not come from one list. They are assembled from several sources, filtered for quality and ranked with a small set of rules. This guide explains those rules in plain terms, including where they fall short.

The main source: Overture Maps

Most results come from a place database we keep ourselves, built from the Overture Maps Foundation's open places dataset. Overture combines records contributed by several organisations, and every place carries a confidence score between 0 and 1 that reflects how sure Overture is that the place exists and its details are right. The contributing sources and their licences are listed on our legal and attribution page. A smaller set of places gathered from Geoapify supplements it.

How a nearby search works

A search takes the location you share and looks within about five miles (eight kilometres) of it, aiming for five good candidates in the category you chose. The steps are:

1. Query our own data first, closest first.
2. Apply a confidence floor. Records with an Overture confidence below 0.5 are not served as recommendations.
3. Drop what we know is closed. Places Overture marks as permanently closed, and places our review process has confirmed as closed or wrong, are excluded.
4. Remove duplicates. One business can appear twice with slightly different spellings, so near-identical entries are merged.
5. Fill any gap from a third-party map service. If our own data cannot supply enough results, the search can fall back to Google Places. Those lookups are capped by a daily limit and cached.

Distance is always re-checked on our side: a map provider's "near here" is only a hint, so anything beyond the cut-off is discarded.

What affects the order

Closeness matters most, but a few other signals shape what you see:

• Confidence, for food. When there are plenty of candidates, restaurant searches lean towards higher-confidence places. It is a preference, not an exclusion.
• Source reliability. Records come from several contributing sources. A source that proves unreliable, meaning its records are often closed or wrong, is excluded from what we serve. That can mean slightly fewer results in some categories.
• What other people pick. The app records which places people were shown as their result and whether they liked it. Once at least three different people have picked a place, a line such as "Trending this week" can appear and the signal can contribute to how candidates are ordered. Below three, nothing is shown, because "picked by one person" says very little.
• Sponsored listings. A business that has claimed its listing can have an active boost, which moves it up among results that already qualified. A boost changes the order, not whether a place is eligible, and boosted results carry a Sponsored label.

What "Report closed / wrong" does

Place cards from our own data have a "Report closed / wrong" button, and you do not need an account to use it. A report is only a flag; it does not remove anything on its own, so one mistaken or malicious report cannot hide a good restaurant.

When three different people report the same place, it goes to a review queue where an administrator checks it. If the review confirms the problem, the place is excluded from searches and from its public page, along with exact duplicates (same name and same address). If not, the place stays. Places that came from Google are not handled this way, because that data is Google's.

Why a place can still be wrong

Open data lags reality. A restaurant can close the week after a snapshot, change its name or move. We refresh our copy from newer Overture releases from time to time, but there is always a gap. So:

• A place can stay listed after it closes, until a refresh or a report removes it.
• The Overture records we hold carry no opening hours, so for those places we cannot tell you whether somewhere is open right now. Check before you travel.
• The confidence score measures how sure Overture is that a record is correct, not whether the food is good.

How this connects to the rest of the app

The wheel and the other games pick the category; this process picks the places inside it. To compare a few places on your own priorities, import them into a weighted decision matrix. For a group, the live group vote produces the category first. To look at the data itself, the places browser lists what we hold by state and city.