How we know this

Where the facts come from

Five sources, and every fact on the site carries the one it came from.

• Public business listings. Name, location, category, hours, rating counts, and the facilities a business or park authority publishes itself. Bathrooms, parking and step-free access come from here.
• What visitors wrote. We read the public reviews for every playground and turn repeated reports into structured facts: shade, equipment for little kids, water play. We store how many people mentioned each thing and the date of the most recent mention, and we show you both. We never republish the review text or visitor photos. Those belong to the people who wrote and took them.
• Open map data. Playground outlines, benches, restrooms, picnic tables and parking from OpenStreetMap, used under the Open Database License. This is what puts the numbered markers on each site plan, and what lets us say how far the restroom is from the swings.
• Public-domain aerial photography. USGS NAIP imagery, with the survey year printed under every image. Tree cover in an unedited aerial is our second check on shade.
• Visit patterns. The hourly busy-ness index Google publishes for each place, which is what the quiet times pages rank on.

Why drawings instead of photographs

We would rather show you a photograph of every playground. We looked hard at what is available and legally reusable at this scale, and the honest answer is that there is nothing. The open archives hold almost nothing pointed at a play area, and the commercial sources either forbid storing their images or forbid republishing them on a site like this one.

So each playground page opens with a drawn site plan instead. We are specific about how it is made, because every other fact on this site carries its source and this one should too: an image model redraws the public-domain aerial photograph of that exact spot, keeping the outline, the paths and the green space at a fixed scale. The numbered facility markers are not drawn by the model; they are placed from OpenStreetMap coordinates. Nothing is invented from a text description, and the frame is chosen from mapped playground outlines rather than by hand.

Under the plan sits the unedited aerial photograph, so you can check the drawing against the ground yourself. Where we could not establish which part of a park the playground occupies, the caption says so and makes no layout claim — and where we could not establish it at all, there is no plan. That is why 713 pages have no drawing. If you send us a picture of a playground you visited, we will use it.

What we hold back

An attribute we cannot get right does not get published, not even hedged. 5 of them are switched off as we write this.

• Fencing is off. A fence matters most to the parent of a child who bolts, which is exactly why we will not publish a number we do not trust. Our hand checks kept finding the same mistake: a fence around a dog run, a construction site or a stairway being counted as an enclosed play area. It stays off until it passes.
• Step-free access is a different problem. Public listings report a wheelchair accessible entrance at 97% of the 6,362 playgrounds we track. A filter that returns nearly everything tells you nothing about which one to pick, and nothing at all about whether the equipment itself can be used. Families who need accessible play deserve better than that, so we do not offer it as a filter.
• Silence is not a no. A playground where nobody mentioned shade is not listed as having no shade. It is listed as a playground we do not know that about.
• Safety-relevant facts are always attributed. Water play and fencing are never stated flatly. They appear as what visitors reported, with how many said it and when.

How we check it

Before anything is published, a random sample of claims is pulled and checked by hand, against the reviews they came from and against the aerial image. Accuracy is measured for each attribute separately. If any attribute comes in under 90%, the build stops and nothing ships. Then we fix the classification and measure again. We do not adjust the number.