Check the filing, not the brochure.

The guides do not. They publish no project lookup, no 0-100 composite and no buy or pause labels - the readiness tool counts your own checklist ticks (readiness X/10, documents X/N) and models an EMI from your loan inputs, so judgement stays with you and the portal record. One clearly-marked experimental prototype is different: the Gujarat explorer reads public GujRERA filings and computes a paperwork score from them. It is a mechanical read of the filing record, not a rating of the project.

Seven: Maharashtra, Gujarat, Karnataka, Telangana, Tamil Nadu, Odisha and Rajasthan. Each links straight to its official portal and project-search surface. Maharashtra is first because its portal carries the deepest public record.

Filings change with extensions, orders and uploads. Mirroring them would go stale and mislead. The portal is the authority; this hub is the walkthrough for what to read there.

A standard reducing-balance EMI from the loan amount, rate and tenure you enter - the same maths as the home-loan EMI calculator. Illustrative, not an eligibility quote, and it never touches a project number.

Four things read straight off the public filing record, out of 100 points. Delivery standing is 40 and falls in steps the further the declared completion date has slipped - on time or delivered scores full marks, while a project five years late scores zero on it. Filing discipline is 25, for quarterly progress returns submitted against those on record. Promoter track record is 20, for projects the same promoter has completed on the portal's own record. The last 15 are a risk check on complaints and extension requests, where entries cost points rather than their absence earning them. It does not measure construction quality, resale value, or whether you should buy, and it is not a LetsInvest verification.

Because it is the only criterion that reliably separates the projects. On the real corpus, 326 filings are past their declared completion date and not declared 100% built, and a scoring model that gives lateness twenty of a hundred points cannot tell a project that is three months late from one that is eight years late. Delivery now carries forty and degrades in steps - 24 points under two years late, twelve between two and five, zero beyond five - so a stalled project cannot hide behind a clean filing trail. Two earlier versions of this score got this wrong: complaints and extensions together once carried thirty points purely for being absent, which told a reader nothing.

Complete record at 80% or more of the points available, Some gaps at 60-79%, Material gaps below 60%. The cut-offs come from how all 1,664 filings in this snapshot actually spread, not from a fixed grade - roughly a third reach Complete record, so it describes this filing rather than praising it. They are fixed numbers, so a project's band only changes if its own filing changes, never because the database grew. If a project has no quarterly returns on record it is scored out of 75 rather than 100, because absent data is not a filing failure.

Ours, and we publish it. Four criteria read off the public GujRERA record, weighted against what actually varies across real filings rather than by assertion. We tried it the obvious way first - equal-ish weights across filings, delivery, complaints, extensions and promoter - and it did not work on the data: delivery was worth only 20 of 100 and was pass or fail, so a project three months late looked identical to one eight years late, while thirty points went to the roughly three-quarters of projects that simply had nothing on record against them. Nothing in that award required the reader to have done anything. So delivery now carries forty and degrades in steps the further the declared date has slipped, and a clean complaint and extension record is worth fifteen rather than thirty. Every project page prints the full breakdown and the arithmetic behind it, so you can check the reasoning rather than take it on trust.