Methodology

About Zomefy says where the data comes from and who runs the site. This page is the harder half: what is actually computed, which filings are excluded and on what stated rule, and where the register is too thin to support the figure a reader would like to see. Anyone quoting this site should be able to check it, so the thresholds are given as numbers rather than described as care.

What the source is

One source: the public register of the Maharashtra Real Estate Regulatory Authority, read directly for every project in the Pune district. Nothing is bought from a data vendor, scraped from a listing portal, submitted by a builder or typed in by hand. Each project page names the registration number and links to the authority's own record for that project, so any figure here can be checked against the filing it came from.

Six things are read for each project: the registration record itself, the unit inventory, the promoter's profile, the cost filing, the title documents, and the land address. A project is only published once all of those have been read, so a page never shows a half-collected record.

What a registered rate is

Every rate on this site is one registered sale agreement divided by the carpet area of the flat it was for. It is what a buyer contracted and paid stamp duty on, not an asking price, a brochure price or a guess at what a flat might fetch. Carpet area is RERA carpet, filed in square metres and converted at 10.7639 square feet to the metre, so a rate here is per square foot of carpet and not of any built-up or saleable area.

This has a consequence worth stating plainly: a Zomefy rate will usually look higher than a portal's rate for the same project, because the portal is quoting a larger area for the same flat. It is the same money over a smaller, legally defined floor.

Homes only. Shops, offices and showrooms are registered by the square foot just as flats are, and at very different money, so 22,058 such agreements are left out of every rate on this site. The test is each unit's own residential flag rather than the project's filed type, because the two disagree in both directions: some registrations filed as Mixed are entirely flats, and others entirely shops. Filtering unit by unit keeps the flats in a tower that has shops on its ground floor.

Where a project has no homes in it at all, it gets no rate and no neighbourhood comparison, because a locality's median describes its housing and would be answering a question the page did not ask.

Which filings are set aside

Promoters type the register in themselves and some of it is wrong. One unit in this corpus is filed as 9042 square metres rather than 90.42. Rather than smoothing errors away with a statistical rule nobody can check, anything outside a plainly stated range is set aside and counted, and the count is published.

Two tests. First an absolute one: a carpet area outside 8 to 1,000 square metres, or a rate outside ₹500 to ₹100,000 per square foot, is not a Pune flat and is dropped.

Second, a test against the project's own filing, because an absolute band cannot catch a rate that is plausible for Pune but impossible for its own tower. Within one project, sold from one price list, a rate more than 2.5 times that project's median — or the same multiple below it — is treated as a mis-filed carpet area rather than a real price. A ratio to the median is used deliberately, in place of a standard deviation or an interquartile fence: a reader can check a ratio against the size table printed on the page, and cannot check a variance.

Of the 270,165 rates the Pune filings on this site yield, 734 were set aside because they sit more than two and a half times away from their own project's median. That is about one in 368. Almost all of them are a carpet area filed wrongly, not a flat that genuinely sold at that price.

Agreement dates get the same treatment. A handful are filed as 1970-01-01, which is an empty field that has been through a date parser, and some predate the authority. Nothing dated before January 2017 can be a MahaRERA agreement, so those dates are dropped for the purpose of a trend while the unit keeps its rate — the rate does not depend on knowing when it was signed. That is why a page carrying both a median and a trend prints two different counts rather than one figure for "sales".

When a locality gets a published rate

A locality needs 10 registered agreements before its median is printed as a rate. That is a readability threshold rather than a statistical one: below it the figure swings by hundreds of rupees on the strength of a single flat, and a reader has no way to see that from the number alone. 160 of 197 Pune localities clear it. The rest carry their project pages and say on the page how many agreements they have instead of showing a rate — Aundh's page, for instance, says there is one agreement on file and that this is too few to publish an area rate.

Where a rate is published it comes with the spread around it, because a median alone invites the reader to treat an area as a single price. The figure shown is a median across agreements rather than a mean, so one tower of small units at a high rate cannot drag a whole locality with it.

Why locality pages carry no price trend

Because the register cannot support one honestly at that size. A quarterly median over an area does not measure price; it measures which projects happened to be registering that quarter, and in Pune those differ by a factor of three. Taking Wakad's agreements quarter by quarter produces a median of about ₹4,650 for one quarter against roughly ₹10,300 for the locality as a whole and ₹12,400 a year later — a market that halved and then tripled, none of which happened. One cheap tower was registering.

So locality pages get a median and a spread, and the trend is published once, for the city, where there is enough depth to compute it.

How the Pune price index is built

The registered-price index never compares one quarter's prices with another's. Each step is measured only within projects that registered sales in both quarters: every such project contributes the ratio of its own median rate between the two, the median of those ratios is the step, and the steps are chained into a level. A project entering or leaving the market therefore moves the sales mix without moving the index, which is the whole point. It is the reasoning behind a repeat-sales house price index, applied to projects rather than to individual flats, because this register re-prices projects continuously and almost never re-sells the same flat.

A step is only published when at least 8 projects sold in both quarters. Below that, one or two projects are deciding what the whole city did, and because the steps compound, that error never washes out. Quarters that cannot be linked are dropped from the front and back of the series rather than interpolated, so the chart is sometimes shorter than the data looks — a level with nothing behind it is worse than a shorter series.

What a project with no sales of its own is shown

Most projects on this register have not registered a sale yet. Rather than leave those pages blank, they are given a comparable: the median of what buyers paid for the same configuration in the same locality. That is only quoted when the locality-and-configuration pair has at least 8 registered agreements behind it, drawn from at least 2 different projects. The second condition matters as much as the first — eight agreements from one project is that project's own rate wearing the locality's name, and a reader following the link would find the identical figure.

A comparable is always labelled as one. It is what the neighbours paid, not what this builder is asking, and the page says so.

Which projects get a page

Not every registration earns one. Three are excluded outright: a lapsed registration is one the authority has stopped standing behind, a deregistered one has been withdrawn, and one in abeyance is frozen mid-dispute. All three describe the paperwork rather than the building.

The fourth test does most of the work: a filing with no unit count is not published. Units are what the size table, the sold figures, the absorption rate and the per-square-foot rate are all computed from, so without them a page is a name, a promoter and two dates. 5378 projects meet the bar today and 7483 are held back, 4936 of them for having filed no unit count.

How current this is

The register is re-read on a schedule and the site is rebuilt daily, which matters because status here is arithmetic against today: a project goes overdue because the calendar moved, not because a filing changed.

The date in the footer is the date the register was last read, and it is deliberately the oldest of the reads behind the page rather than the newest. If the unit detail was refreshed this morning but the project sweep last ran in March, March is the honest answer, and March is what it says.

The figures as files

Everything above is on the pages one locality at a time, which is not much use to anyone checking the arithmetic or citing it. The same numbers are here as CSV, rebuilt nightly with the site:

• Pune locality rates — the median for every locality that clears the threshold above, with the middle-half spread, the number of agreements behind it, the dates they span, and the delivery record of the projects in the area.
• Pune registered-price index — the quarterly series, with the matched-project count and agreement count behind each step.

Use them for anything, including commercially. The registrations underneath are public record and the arithmetic is mine; all I ask is a link back so a reader can see how a figure was made. If you are writing something up and want a cut of the data that is not here, ask on the contact page.

What this cannot tell you

The site is only ever as good as the filings behind it. A promoter who files late, thinly or wrongly appears that way here, because reproducing the record faithfully means reproducing its gaps. A project with no registered sales has no rate of its own. An absence of complaints is an absence of evidence and not evidence of quality. And a registered price is a historical fact about one agreement, not a valuation of a flat on sale today.

None of this is advice about a specific purchase. Before you transact, verify anything material directly at the MahaRERA register. If a figure here does not match the filing, tell me the registration number and what you are seeing and I will fix it — details are on the contact page.