Five filing signals that precede a stalled project

No project announces that it is about to stall. But almost every project that stalls has already told the register, quarters earlier, in five specific ways. India's stalled-housing problem is enormous, the national numbers run to several lakh units, and the government's SWAMIH rescue fund alone carried 1,01,443 homes in its completion portfolio with 63,200 delivered by January 2026 (SWAMIH programme data). Every one of those projects filed, or conspicuously stopped filing, on a public register on the way down. This is the watchlist.

The logic: why stalls are legible in advance#

A residential project stalls when its funding equation breaks: construction cost remaining exceeds committed capital plus realistic future collections. That equation deteriorates in stages, and each stage produces a register artifact. Sales slow first (collections fall), then withdrawals strain against the certificate gates, then the timeline slips formally (extension), then the promoter goes quiet with the regulator, then disputes surface. The register does not predict the stall; it documents the approach to it, quarterly, while the borrower's interest payments are still current. The lender's edge is simply reading the documentation in order.

The five signals, with thresholds#

Signal 1: filing silence. QPRs are due within 20 days of quarter end (MahaRERA Orders 18/2021 and 33/2022). One late filing is administrative; two consecutive lapses put the project in enforcement territory, and the company it keeps there is instructive: 8,212 projects drew section 7 show-cause notices in the May 2026 sweep alone. Threshold: any missed deadline triggers a query; two consecutive trigger a site visit.

Signal 2: the sales freeze. Booked units flat across three or more quarters while the project is actively marketing. The register makes this precise because inventory tables are restated quarterly. A freeze with construction continuing means the promoter is building on debt alone; a freeze with construction slowing is the equation already broken. Threshold: bookings moving less than 2 percent of inventory per quarter for three quarters, in a catchment where peers are selling.

Signal 3: serial extensions. Section 6 allows a one-year extension for reasons beyond the promoter's control; MahaRERA processes further relief with majority allottee consent. One extension, documented, can be honest. The signal is seriality and spread: multiple extensions on one project, or extensions arriving across the promoter's portfolio within a few quarters. Threshold: a second extension anywhere in the group is an underwriting event, not a formality.

Signal 4: the progress-versus-time gap. Percentage of work done against percentage of registered timeline elapsed, both filed. MahaRERA's own grading matrix, designed in 2023 and shelved before any grade was published, used exactly this ratio as a core input, which tells you the regulator considered it diagnostic. Threshold: work done lagging time elapsed by more than 20 percentage points, or a work-done line that moves less than 3 points a quarter through what should be the superstructure phase.

Signal 5: litigation plus enforcement. Litigation disclosures, complaint accumulation, recovery warrants, and at the hard end a lapsed or revoked registration. By November 2025 MahaRERA's recovery-warrant ledger stood at Rs 792 crore ordered with Rs 268.87 crore recovered (MahaRERA data via Business Standard, December 2025); names recur on that ledger. Threshold: any warrant, or litigation touching the project land, moves the account to the watch category regardless of payment status.

The fifth signal is the one desks most often discount, on the reasoning that an order is not a loss until it is enforced. The enforcement record argues the other way: what the ledger shows is not that warrants get paid, but that they mostly do not.

Roughly a third of ordered money reaches the buyer who won it. For a lender that ratio is not a consumer-protection statistic, it is a statement about the recoverable value sitting in the same promoter's other projects, and about how much of the buyer claim stack ahead of you is likely to stay unpaid and therefore litigious. What that stack does once the promoter reaches the NCLT, seen from the buyers' side of the table, is in the homebuyer's insolvency playbook.

Worked example: scoring a borrower quarterly#

Illustrative numbers. A Rs 35 crore exposure to a 480-unit Thane project, reviewed on filings each quarter. Score two points per signal: zero if absent, one if emerging, two if firm.

Q1 review: bookings moved 11 to 13 percent (signal 2 emerging, 1 point), all else clean. Score 1. Noted. Q2: bookings 13 to 13.5 percent (firm, 2), QPR filed six days late (emerging, 1), work done 34 percent against 45 percent of time elapsed (emerging, 1). Score 4. Borrower call, written explanations requested. Q3: bookings flat (2), QPR missed entirely (2), gap now 24 points (2), first extension application filed citing approvals (1). Score 7 of 10. The account is still fully current on interest. On filings alone it is now a watch-category name with a documented four-quarter deterioration path, and the desk is negotiating protections while the borrower still needs goodwill, not after.

That last clause is the entire economics of early warning: options are cheap while the promoter is solvent and expensive after. At the statutory delay rate of SBI's highest MCLR plus 2 percent, currently 10.90 percent per annum, every quarter past the registered date also accrues buyer claims that rank effectively ahead of recovery value.

Calibrating the thresholds to the market#

Thresholds are only as good as the baseline they sit on, and the baseline is observable, as a cross-section if not yet as a history. Across Active MahaRERA registrations from 2024 onward that had been on the register for at least six months by 21 September 2026, the median project had booked 1.9 percent of its units per quarter since registration (the middle half: 0 to 4.7 percent) and advanced 4.6 points of construction per quarter (1.8 to 7.4); the QPR reading guide explains how both are read. The 2 percent sales default above therefore sits close to the median project, which is exactly why it has to be set against the catchment before it means anything: a "flat" quarter says something different in a catchment clearing at 5 percent than in one clearing at 1, and the same is true across cities in the same state at the same time: H1 2026 had Mumbai registrations at a 13-year high while Pune's unsold stock rose 19 percent (IGR data via Knight Frank; Knight Frank India Real Estate, H1 2026). Calibrate each borrower's thresholds to their catchment's filed medians, not to a statewide constant, and re-baseline annually. The signal is always the borrower's divergence from their own market, because a whole catchment slowing together is a different problem, portfolio concentration, with a different playbook. The supply-side version of the same discipline, which developers use to time launches, is in the filings-as-intelligence opener; a lender reads the identical curves with the sign flipped.

One more calibration note: signal 1 needs a state adjustment. Maharashtra's enforcement makes filing silence meaningful because the deadline is policed; in states where late filing is routine, the silence threshold must lengthen or the desk drowns in false positives. The framework travels; the constants do not.

The pattern is not hypothetical. The great stalled portfolios now being unwound in public, the Amrapali group whose registrations the Supreme Court cancelled in July 2019 with completion handed to NBCC, the Jaypee Infratech book resolved through the NCLT with 20,000 homes still being delivered under the Suraksha plan approved in March 2023, deteriorated exactly this way: sales stopping, money moving, timelines slipping, disclosure thinning. Those collapses predate the mature register, which is precisely the point: the disclosure regime was built from their post-mortems, and today the same trajectory prints quarterly in public. We replay those cases against the filings record in a companion piece in this series.

Kavya Menon's desk (illustrative, as ever) runs the 10-point score on every real estate name at quarter close, from the filings, before reading the borrower's own pack. The discipline matters more than the sophistication: the score exists so that the third consecutive flat quarter cannot be explained away one quarter at a time.

What the five signals cannot see#

A watchlist is only usable if the desk knows where it goes blind, and this one has four blind spots worth writing into the credit note rather than discovering during a review.

Coverage is not universal, and signal 2 is the one that suffers. The building-level inventory table that makes a sales freeze precise is filed by only about a third of published Maharashtra projects, 18,600 of 55,631 on our August 2026 read of the register, and a project-level unit count appears on 38.8 percent. Filing has improved sharply in recent registrations, so a 2024-onward borrower is usually measurable and a 2019 one often is not. Where the table is absent, signal 2 degrades from a number to an inference, and the desk should say so in writing instead of scoring a zero that reads like an all-clear.

Bookings are counts, not values. RERA files no sale price, so a promoter discounting 15 percent to keep the booking line moving is indistinguishable on the register from one selling at list. Collections can therefore deteriorate while signal 2 stays green. This is the single most important reason the register supplements the borrower's receipts rather than replacing them.

Progress percentages are certified, not audited. The architect and engineer who certify work done are appointed by the promoter. That is a real check, and it is not an independent one; the lender's own engineer still earns their fee.

Everything lags up to a quarter. A project can break in the first month of a quarter and print healthy until the QPR lands eighty days later. The register tells you sooner than the borrower does, which is the whole claim being made here, but it does not tell you immediately.

Running this across a book is an aggregation problem, and it is the one ReraGenie is built for: every registration's current filing, extension and litigation records, promoter footprints matched across registrations, and a record of each change to each filing since July 2026, which the portals overwrite. The Rs 2,999 project analysis sets a single exposure's filing against its micro-market, its operators and a risk register; the area market report, a flat Rs 2,999 per pincode, lays out a catchment's registrations, slips and disputes, which is the portfolio version. The district pages show registration status and possession slips at market level, free.

The one-line summary#

Stalls are funded quarters in advance and filed quarters in advance: watch silence, frozen sales, serial extensions, the progress-time gap and enforcement flags, score them together, and act on the cluster while acting is still cheap.