How rent data is calculated

What qualifies as a report

A report is one renter marking their flat on the map and entering the monthly rent, the size (1RK, 1, 2, 3, or 4+ BHK), and a few optional details like deposit, parking, and water source. No account, no sign-in, and no personal identity is attached. The location is rounded to about 110m before it is stored, so the point shows the block, never the door.

A rent has to be at least plausible to be stored at all: anything below ₹2,000 or above ₹5,00,000 a month is rejected outright as a typo or a joke. Values inside those limits but outside the normal range for their size are kept but held aside - see “unusual reports” below.

Minimum sample size for a median

An area needs at least 5 reports of the same size before any median is shown for that size. Below that floor, the individual pins are still on the map, but no summary number is published - a “median” off one or two rents is noise wearing a number’s clothes. The same floor applies to the deposit, parking, and water-source summaries.

How the median is computed

The headline figure is a corroboration-weighted median: a rent counts for more as independent nearby reports (within ±12%) agree with it. A tight, agreed-upon cluster carries the number; an isolated high or low barely moves it, which blunts both honest mistakes and planted pins. Alongside it we show the raw 25th–75th percentile spread - the range the middle half of renters pay - so a genuinely wide or two-sided market stays visible instead of hiding inside a single number.

What low / medium / high confidence means

A median off 5 reports deserves less trust than one off 40, so each is labelled by how many reports it rests on:

• Low - 5–9 reports. A rough signal; treat it as a starting point, not a settled number.
• Medium - 10–24 reports. A reasonable read on the area for that size.
• High - 25 or more reports. Well sampled; the median is stable.

How stale reports are labelled

Rents age. A rent agreed a year or more ago says little about what a landlord asks today, so any report older than one year is marked “over a year old”. Stale reports are not deleted - they are real rents someone paid - but where a median leans on them, the page shows how many of the reports behind it are that old, so you can weigh the number accordingly.

How unusual reports are handled

A report that lands outside the normal range for its size, or more than 2.5× above or below the local median, is accepted onto the map but flagged and held out of the median until reviewed, shown in amber as “unusual”. Keeping it visible but excluded is deliberate: hiding it would erase a genuine outlier, while counting it would let a few extreme entries drag the number.

From there, a report is removed if 2 or more people report it as wrong, or if it stays extreme once the surrounding area has enough clean data to judge it against. A flag that no one resolves and that new data never vindicates is set aside after 30 days as “unconfirmed” - still shown in amber for transparency, still kept out of every median - rather than sitting in a queue forever.

Limitations

• The data is renter-reported and not independently verified. Treat every figure as an estimate.
• Locations are rounded to about 110m, so a pin marks a block, not a specific building.
• Thinly sampled areas can move as more reports arrive; the confidence label tells you which those are.
• Rents on the marketplace are owner asking prices and are separate from this renter-reported map.
• Always confirm the current rent directly before you sign anything.

Think a number is wrong, or want a report removed? Here is how to report it. More on the project is on the About page.