- WHAT IT IS
- NIRS scores the location of an Indian address — access, environment and hazard, services, and trajectory — on a 0–100 scale, re-weighted across three intent profiles, and publishes the confidence of every output. The system refuses to score below 0.55 confidence. The plan breaks even in Year 4 on ₹16.0 crore of peak capital.
- VALIDATED
- Research settled three things. Lenders already spend ₹5,000–25,000 a file on legal and technical verification 1, across 3.64 million annualised home-loan originations 2. Location accessibility leads affordable-housing dissatisfaction at 92%, ahead of build quality 3. And the paperwork is already rated by funded companies at ₹1,999 8 and at ₹99 9, with valuation given away free 13.
- NEXT
- Gate 1 tests lawful, repeatable access to Karnataka land records on 1,000 random e-Khata addresses, at near-zero cost in Week 1. A blind valuer panel follows at months three to five for ₹12–15 lakh.
All cost, margin, acquisition-cost, revenue, capital and volume figures are modelled estimates from the project's financial model, built on externally sourced benchmarks and pending the validation gates. Product images are renders of a specified interface, not photographs of a running system.
THE OPPORTUNITY
01Product Overview: A 0–100 Location Score for the Two Buyers Who Transact Blind
NIRS returns a 0–100 location score for a Bengaluru address, weighted by whether the buyer intends to live, trade or build, with the confidence of the answer printed on the answer.
Four dimensions carry the score — access, environment and hazard, services, and trajectory — re-weighted across three intent profiles, and the system refuses to score below 0.55 confidence. The rubric, data model, API contracts, screen inventory and financial model are specified.
Two buyers anchor the release. The Bengaluru resale buyer receives none of the RERA protection a primary-market buyer gets. The housing finance company underwriting that purchase spends ₹5,000–25,000 a file on legal and technical verification 1, and gets no view on whether the road floods or the funded metro line is real.
02The Problem: 92% of Dissatisfied Buyers Name Location, Ahead of Build Quality
An Indian buyer can establish whether the title is clean but not whether the location holds up, and poor accessibility draws 92% of complaints against 90% for construction quality 3.
ANAROCK's survey of 8,250 respondents across 14 cities puts poor accessibility at 92% and build quality at 90% 3. The dissatisfaction is retrospective: buyers regret the location after ownership rather than screening for it before.
Three changes make the problem addressable now. Records of Rights are computerised at 95.09% but geo-referenced at only 49.10% 5. ULPIN reaches about 30% of 28.72 crore parcels and covers rural land only, leaving urban property without an identifier 6. And the adjacent answer collapsed from a ₹5,000–20,000 lawyer's title search to an automated rating at ₹1,999 8.
The state grants presumptive title only, which places the onus of checking on the buyer 7. The title question now has four products. The location question has none.
03Hypothesis Pivot: I Killed the Legal-Rating Thesis After Finding Three Funded Competitors Already In It
I rejected the inherited five-dimension score led by legal and title after evidence review found three funded companies already selling a property rating and driving its price toward ₹99.
The inherited thesis rested on three beliefs: that nobody produced a standardised property score, that a hashed property identifier was defensible intellectual property, and that a ₹999 report could be acquired at ₹31. Each was tested against published evidence and each failed.
LegiScore rates legal risk across 14 states and 700+ courts at ₹1,999 8. Landeed retrieves documents across 20+ states on a US$8.3M seed round 12. Propil scores at ₹99 9. MagicBricks gives valuation away 13.
I excluded the legal dimension, the price estimate and paid consumer acquisition, accepting the loss of 25 of the rubric's 100 points and of every channel that scales with spend. The product moved onto the one question nobody sells — whether the location is a good place to own — which changed the customer, the price and the capital plan.
04Market Position: The Lender Due-Diligence Pool Is ₹1,820–9,100 crore a Year
The right market is due-diligence spend already flowing rather than asset value, and the whole nine-metro consumer pool reaches ₹81.6 crore even at full attach.
₹1,820–9,100 crore
LENDER DUE-DILIGENCE POOL
3.64 million annualised originations 2, each verified at ₹5,000–25,000 1
₹81.6 crore
CONSUMER POOL AT FULL ATTACH
545,000 registrations across nine metros in 2025 4; ₹8.2 crore at a tenth attach and ₹20.4 crore at a quarter
111.52×
LENDER POOL AGAINST CONSUMER
The top of the lender band against the consumer pool at full attach 21
US$650 billion
ASSET VALUE, THE WRONG DENOMINATOR
India's real-estate market at 7.3% of GDP 10, the base the predecessor sizing used
Lender verification spend is 111.52× the consumer pool at full attach
05Competitive Landscape: Four Products Price the Paperwork and None Scores the Location
Every funded entrant answers the same question — is the paperwork clean — and drives its price toward ₹99, which leaves the location layer unoccupied.
| PLAYER | WHAT IT OWNS | WHAT IT LACKS |
|---|---|---|
| LegiScore 8 | Legal rating across 14 states and 700+ courts at ₹1,999, with lender integration | Any location, hazard or trajectory signal |
| Landeed 12 | Document retrieval across 20+ states and a Fetch API, on a US$8.3M seed round | Judgement — it fetches documents, it does not assess them |
| Propil 9 | A ₹99 anchor across 16 data points | Depth — a screen, not a report |
| MagicBricks PropWorth 13 | Free valuation across 30 cities and 5,500 localities | Legal, hazard and infrastructure |
PLAYER
LegiScore 8
- WHAT IT OWNS
- Legal rating across 14 states and 700+ courts at ₹1,999, with lender integration
- WHAT IT LACKS
- Any location, hazard or trajectory signal
Landeed 12
- WHAT IT OWNS
- Document retrieval across 20+ states and a Fetch API, on a US$8.3M seed round
- WHAT IT LACKS
- Judgement — it fetches documents, it does not assess them
MagicBricks PropWorth 13
- WHAT IT OWNS
- Free valuation across 30 cities and 5,500 localities
- WHAT IT LACKS
- Legal, hazard and infrastructure
The substitutes matter more than the rivals, because the market leader is doing nothing: most buyers visit twice and decide.
Legal scoring is a low barrier that three firms cleared in under three years. Geo-referenced location scoring is a high one. It needs cadastral geometry where only 49.10% of villages are geo-referenced 5, and ground-truth labels bought from valuers one property at a time.
The gap stays open because occupying it costs each incumbent its own asset: it dilutes Landeed's breadth, sits outside LegiScore's legal buyers, and damages MagicBricks' inventory.
The price of an answer about a property fell from ₹5,000–20,000 to ₹99
06Strategic Bet: I Sold the One Layer Nobody Owns and Bought or Refused Everything Else
NIRS sells location and forward risk, buys the legal check wholesale and refuses to estimate price, because a location assessment shifts over time and so becomes a subscription rather than a one-time report.
I excluded four things. Title retrieval, because Landeed sells it as an API 12. Legal scoring, because its price is collapsing toward ₹99 9. Price estimation, given away by PropWorth 13 and beyond the 7.5% off-market error Zillow publishes on its own estimate 14. And plot-level air quality, because 401 continuous stations across 209 cities 19 make a parcel-level reading a fabrication.
I killed the legal dimension, worth 25 of 100 points, accepting the loss of the one dimension lenders already buy. A legal answer is needed once. A location assessment shifts as a metro line is funded or a lake encroached, which makes it a data asset rather than a transaction.
I chose the affordable-housing finance company as the first customer, accepting a slower sale than a consumer app would give, because it underwrites location risk worst today and can approve a ₹200 layer without a board paper.
THE PRODUCT
07Product Specification: One Question, Answered in 90 Seconds, With a Published Refusal
Every signal carries its own provenance, as-of date and confidence, and the score's confidence is the minimum of the resolution confidence and the weighted-dimension confidences.
The sequence runs from a locality page built to be found in search, to one address field, to a resolution screen showing what was matched and how confident the match is. Intent — live, trade or build — is chosen before the score, because a warehouse and a family home do not want the same street.
Refusal is a property of the architecture rather than an error state. Below 0.55 confidence the system returns no score, because roughly a third of stock lacks a usable formal record. Air quality is reported at station-catchment resolution and never interpolated to the parcel, on 401 stations serving 209 cities 19.
The inventory runs to 22 screens at 6 required states each, over 25 signals in 4 dimensions and 3 intent profiles, with the weights verified to sum to 100.
08Technical Validation: A Blind Valuer Panel at Pearson r of 0.60 Decides the Product
200 Bengaluru properties, 5 chartered valuers rating location blind, ₹12–15 lakh over 8 weeks, with every threshold fixed before any data is collected.
| CORRELATION | WHAT IT MEANS | WHAT HAPPENS |
|---|---|---|
| r ≥ 0.75 | The score tracks expert judgement closely | Launch, with the panel's ratings carried as training labels |
| 0.60 to 0.75 | The score carries signal at wider error | Launch, with the published confidence bands widened |
| 0.45 to 0.60 | Signal is present but too weak to price | Building continues; selling halts until it clears 0.60 |
| r < 0.45 | Location is not scorable by this rubric | Stop the product |
CORRELATION
r ≥ 0.75
- WHAT IT MEANS
- The score tracks expert judgement closely
- WHAT HAPPENS
- Launch, with the panel's ratings carried as training labels
0.60 to 0.75
- WHAT IT MEANS
- The score carries signal at wider error
- WHAT HAPPENS
- Launch, with the published confidence bands widened
0.45 to 0.60
- WHAT IT MEANS
- Signal is present but too weak to price
- WHAT HAPPENS
- Building continues; selling halts until it clears 0.60
r < 0.45
- WHAT IT MEANS
- Location is not scorable by this rubric
- WHAT HAPPENS
- Stop the product
Whether a location can be scored well enough to matter is the central question of the venture, and the panel is the instrument. 5 RICS-chartered or IBBI-registered valuers rate 200 Bengaluru properties on location alone, blind to the score, at months three to five.
Two conditions make it a real test: the panel is blind, and the stop threshold was written before any data was collected. The earlier internal pass of 47 of 47 checks confirmed one thing only — that the rubric weights sum to 100.
The band between 0.45 and 0.60 is the expensive one, because it licenses further build but not a sale. It is written into the table above so that a mid-range result cannot be argued into a launch.
THE BUSINESS CASE
09Cost to Serve: 88.37% on a ₹499 Report, and the Consumer Still Never Funds the Company
A ₹499 report costs ₹58.01 to serve, yet the most NIRS can afford to pay for a consumer is ₹309, while one lender API call at ₹200 repays its whole sales cost in 2.0 months.
| COST LINE ON A ₹499 STANDARD REPORT | AMOUNT |
|---|---|
| Payment gateway at 2.36% 15 | ₹11.78 |
| LLM narrative 16 | ₹16.25 |
| Data, support, refunds and compute | ₹29.98 |
| Total cost to serve | ₹58.01 |
| Contribution | ₹440.99 |
COST LINE ON A ₹499 STANDARD REPORT
Payment gateway at 2.36% 15
- AMOUNT
- ₹11.78
LLM narrative 16
- AMOUNT
- ₹16.25
Data, support, refunds and compute
- AMOUNT
- ₹29.98
Total cost to serve
- AMOUNT
- ₹58.01
Contribution
- AMOUNT
- ₹440.99
The margin is healthy and beside the point. A consumer lifetime value of ₹1,015 sets the maximum affordable acquisition cost at ₹309, on a 15% lead-to-paid rate and 1.94 purchases a buyer — the two inputs Gate 3 measures.
Enterprise inverts it. One lender at 40,000 calls a year pays ₹80 lakh at ₹200 a call, which is what 16,000 consumer reports bring in, and it repays its ₹13.5 lakh sales cost in 2.0 months.
Rebuilt bottom-up, the disciplined plan breaks even in Year 4 on ₹16.0 crore of peak capital, against ₹41.4 crore in the base case and the ₹72 crore and ₹147 crore the two predecessor plans proposed. The ₹5.75 crore seed sits inside the published ₹3–8 crore band 18, and base becomes disciplined on two decisions: headcount held to 47, and the API priced at ₹200 rather than ₹120.
Peak capital falls from ₹147 crore to ₹16.0 crore
- Disciplined plan
- Other scenarios
- Predecessor plans
10Success Metrics: Confident Coverage Reaches 65% by Month Twelve or Distribution Cannot Rescue It
Volume can be bought with a discount and coverage cannot, so the decisive number is queries resolved at 0.70 confidence or higher and scored within tolerance of an independent assessment.
65%
CONFIDENT COVERAGE, MONTH TWELVE
At two in three Bengaluru addresses the rest is distribution; at two in five no distribution rescues it
0.60
VALUER CORRELATION FLOOR
Below it selling halts, whatever coverage or volume reads
8%
REFUND-RATE CEILING
Above it the product is over-promising; refusal rate is read against coverage so a loosened threshold shows
11Evidence to Date: The Spend and the Price Ladder Hold; Prospective Demand Is the Open Test
Three claims rest on external data. The fourth — whether a buyer pays for a location assessment before the purchase rather than regretting it after — is what Gate 3 measures.
EVIDENCED
EVIDENCED
EVIDENCED
Location accessibility leads affordable-housing dissatisfaction at 92%
ANAROCK survey, 8,250 respondents across 14 cities 3
UNTESTED
A buyer pays for a location assessment before the purchase rather than regretting it after
Gate 3 — a 300-user live landing test at ₹499, ₹3 lakh across months four to six
12Adoption Path: Paid Acquisition Runs 17–43× Over the Ceiling, Leaving the Slowest Channel
A real-estate lead costs ₹800–2,000 on Google Ads 17, which puts consumer acquisition at ₹5,333–13,333 against a ₹309 ceiling and eliminates every paid channel permanently.
That is a concluded result rather than an optimisation problem. What remains is programmatic locality search: one page a locality, built from data NIRS already holds, written against the questions buyers type.
The economics compound slowly. Organic acquisition costs ₹45 lakh in Year 2 to win 3,409 users at a loss, and cost a user falls to ₹338 by Year 4 — 3.00× against the 3.0× gate, and still ₹29 above the ₹309 the model can afford.
Enterprise acquisition repays in 2.0 months, so consumer spend is hard-coupled to enterprise progress rather than treated as free: SEO spend stays under ₹45 lakh a year until one enterprise pilot is signed.
Organic acquisition clears the 3.0× gate in Year 4; paid never does
13Risk Gates: Three Thresholds, Each Dated and Written Before Its Test
The cheapest test gates the most expensive commitment: Gate 1 costs near nothing in Week 1 and decides whether the product is viable as specified.
| RISK | GATE THRESHOLD | SCOPE AND TIMELINE | CONSEQUENCE |
|---|---|---|---|
| Data access | Written eligibility to consume Karnataka land records, or retrieval under ₹150 a property | 1,000 random e-Khata addresses, Week 1 to month two | Pass: build. Fail: re-scope to a ward-level product, or stop |
| Willingness to pay | Paid conversion at 10% or above at ₹499 | 300-user live landing test, months four to six | Pass: keep the consumer tier. Below 6% with no enterprise pilot: cut it to a free B2B lead surface, removing ₹13.5 crore of Year-5 revenue and moving break-even to Year 5 |
| Consumer losses | One signed enterprise pilot unlocks SEO spend above ₹45 lakh a year | Continuous covenant on the plan, from month nine | Held: peak capital falls from ₹41.4 crore to ₹16.0 crore. Breached: a loss-making consumer channel funds with no counterweight |
RISK
Data access
- GATE THRESHOLD
- Written eligibility to consume Karnataka land records, or retrieval under ₹150 a property
- CONSEQUENCE
- Pass: build. Fail: re-scope to a ward-level product, or stop
- SCOPE AND TIMELINE
- 1,000 random e-Khata addresses, Week 1 to month two
Willingness to pay
- GATE THRESHOLD
- Paid conversion at 10% or above at ₹499
- CONSEQUENCE
- Pass: keep the consumer tier. Below 6% with no enterprise pilot: cut it to a free B2B lead surface, removing ₹13.5 crore of Year-5 revenue and moving break-even to Year 5
- SCOPE AND TIMELINE
- 300-user live landing test, months four to six
Consumer losses
- GATE THRESHOLD
- One signed enterprise pilot unlocks SEO spend above ₹45 lakh a year
- CONSEQUENCE
- Held: peak capital falls from ₹41.4 crore to ₹16.0 crore. Breached: a loss-making consumer channel funds with no counterweight
- SCOPE AND TIMELINE
- Continuous covenant on the plan, from month nine
Two standing mitigations sit alongside the gates. DPDP exposure, with penalties reaching ₹250 crore 20, is met by storing owner data drawn from public records only.
Score liability is met by positioning NIRS as a data reference with a published refusal state, never as a valuation.
14Scale Trajectory: Year 5 Puts a Location Assessment on 24.73% of Home-Loan Originations
On the disciplined plan's own growth math, Year 5 delivers 180,000 consumer reports and 900,000 enterprise API calls against 3.64 million annualised originations 2.
Cumulatively across five years that is 322,000 reports for 170,000 distinct buyers and 1.9 million loan files — a meaningful start, not a national standard.
Concretely it changes one decision. A flat that showed well on a dry afternoon scores 54 at 0.78 confidence and flags 2 waterlogging events within 400 m, so the buyer negotiates the price down and insures before moving in.
Three adjacencies become reachable and none of them is in the plan: location as a priced input to credit, once a lender holds 200,000 scored files with repayment performance attached; property insurance that prices the ground at parcel resolution, the input the title-insurance market never had 11; and a geo-referenced urban parcel registry beside a rural-only ULPIN 6.
THE OUTCOME
15What This Unlocked: The Cheapest Test Now Gates the Most Expensive Commitment
I resequenced the work so Gate 1, lawful data access, runs in Week 1 before any raise, and a published confidence band is treated as a product feature rather than a disclaimer.
The correction came from one error. A 70-signal rubric and a 7-layer architecture were designed for a category three funded companies already occupied, which four minutes of research would have surfaced. The same answer in month fourteen kills a company; in Week 1 it re-scopes a plan.
The second output is a standard of evidence. 21 unsourced claims were removed and recorded across 8 superseding decisions, and every one of them was vague in the same way — its roundness was the tell. The replacements are checkable.
In a market where the price of an answer is falling toward ₹99, the durable difference is not a better number. It is the willingness to state how much is known, and where the product will not answer.
16Build Pack: Six Artifacts Complete, Four Gated Tests Advancing Next
The rubric, data model, API contracts, screen inventory and financial model are specified and internally verified, and the next four steps convert specification into evidence.
| ARTIFACT | STATE |
|---|---|
| Scoring rubric | 25 signals, 4 dimensions and 3 intent profiles, with weights verified to sum to 100 |
| Data model and schema | Bitemporal and point-in-time correct; refusals stored as first-class records |
| API contracts | resolve, score and report, plus a public source-status endpoint |
| Screen and state inventory | 22 screens at 6 required states each |
| Financial model | Four scenarios, 24 independent checks, zero mismatches |
| Decision and removed-claims log | 8 superseding decisions; 21 unsourced claims removed and recorded |
ARTIFACT
Scoring rubric
- STATE
- 25 signals, 4 dimensions and 3 intent profiles, with weights verified to sum to 100
Data model and schema
- STATE
- Bitemporal and point-in-time correct; refusals stored as first-class records
API contracts
- STATE
- resolve, score and report, plus a public source-status endpoint
Screen and state inventory
- STATE
- 22 screens at 6 required states each
Financial model
- STATE
- Four scenarios, 24 independent checks, zero mismatches
Decision and removed-claims log
- STATE
- 8 superseding decisions; 21 unsourced claims removed and recorded
| ADVANCING NEXT | COST | WHAT IT UNLOCKS |
|---|---|---|
| Gate 1 — Karnataka land-records access | Near zero, Week 1 | Whether the product is viable as specified |
| Geo-referencing coverage sample | ₹2–4 lakh, weeks two to six | Whether Confident Coverage can reach 45% |
| Gate 2 — blind valuer panel | ₹12–15 lakh, months three to five | Whether the score carries signal |
| Gate 3 — paid landing test | ₹3 lakh, months four to six | Whether consumers pay before the purchase |
ADVANCING NEXT
Gate 1 — Karnataka land-records access
- WHAT IT UNLOCKS
- Whether the product is viable as specified
- COST
- Near zero, Week 1
Geo-referencing coverage sample
- WHAT IT UNLOCKS
- Whether Confident Coverage can reach 45%
- COST
- ₹2–4 lakh, weeks two to six
Gate 2 — blind valuer panel
- WHAT IT UNLOCKS
- Whether the score carries signal
- COST
- ₹12–15 lakh, months three to five
Gate 3 — paid landing test
- WHAT IT UNLOCKS
- Whether consumers pay before the purchase
- COST
- ₹3 lakh, months four to six
SOURCES
- [1]Finnable — legal and technical verification at ₹5,000–25,000 a home loan in metro markets. Sourced. https://www.finnable.com/finn-advice/what-is-legal-and-technical-in-home-loan/
- [2]CRIF High Mark — How India Lends, Q2 FY26: 9.1 lakh home-loan originations in the quarter, annualising to 3.64 million; average ticket ₹33.2 lakh; HFC share 25.2%. Sourced. https://www.crifhighmark.com/media/5720/how-india-lends-sept-2025.pdf
- [3]ANAROCK — Homebuyer Sentiment Survey H1-2025, 8,250 respondents across 14 cities; 92% of dissatisfied affordable-housing buyers cite location accessibility, 90% construction quality. Sourced. https://websitemedia.anarock.com/media/Homebuyer_Sentiment_Survey_H1_2025_770cb18d6a.pdf
- [4]Square Yards via Construction World — 545,000 home registrations across nine cities in 2025. Sourced. https://www.constructionworld.in/latest-construction-news/real-estate-news/home-registrations-dip-sales-value-rises-in-2025/83865
- [5]Department of Land Resources — DILRMP: geo-referencing at 49.10% against Records of Rights 95.09% computerised, as of 31 December 2023. Sourced. https://dolr.gov.in/en/programmes-schemes/dilrmp-2/
- [6]ThePrint — ULPIN / Bhu-Aadhaar at about 30% of 28.72 crore parcels, rural land only, against a March 2026 target. Sourced. https://theprint.in/india/governance/3-years-on-30-of-rural-land-parcels-have-bhu-aadhaar-centre-pushes-for-completion-by-2026/2325001/
- [7]PRS Legislative Research — Land Records and Titles in India: the state grants presumptive title and the onus of checking rests with the buyer. Sourced. https://prsindia.org/policy/discussion-papers/land-records-and-titles-india
- [8]LegiScore — legal risk rating across 14 states and 700+ courts at ₹1,999 in under 15 minutes, ISO 27001, against the ₹5,000–20,000 manual title search it replaces. Sourced. https://legiscore.in/ and https://legiscore.in/alternatives/property-due-diligence
- [9]The Daily Guardian — Propil's PropilScore: 16 data points at ₹99. Sourced. https://thedailyguardian.com/business/the-rs-50000-crore-real-estate-trap-ai-disrupter-launches-bank-grade-property-verification-for-rs-99-753705/
- [10]IBEF — Indian Real Estate Industry Analysis: US$650 billion in 2025, 7.3% of GDP. Sourced. https://www.ibef.org/industry/real-estate-india
- [11]TEAL India — Title Insurance in India: RERA Section 16, eight insurers approved between 2018 and 2021, no state notified rules and the market never formed. Sourced. https://tealindia.github.io/research/TEAL%20White%20Paper%203-%20Title%20Insurance%20in%20India.pdf
- [12]Dealroom and TechCrunch — Landeed's US$8.3M seed round, Y Combinator; 20+ states and a Document Fetch API. Sourced. https://app.dealroom.co/news/feed/landeed-raises-8-3m-seed-round-to-scale-indian-property-title-search-engine
- [13]MarkHub24 — MagicBricks PropWorth: free machine-learning valuation across 30 cities and 5,500 localities, launched July 2024. Sourced. https://www.markhub24.com/post/magicbricks-propworth-property-valuation-as-a-customer-acquisition-innovation
- [14]HomeRise, citing Zillow — Zestimate median error of 1.9% on-market and 7.5% off-market, across more than 100 million homes. Sourced. https://homerise.com/zillow-zestimate/
- [15]Razorpay — a 2% platform fee plus 18% GST, 2.36% all in. Sourced. https://razorpay.com/blog/razorpay-payment-gateway-pricing-explained/
- [16]BenchLM — Anthropic API pricing, September 2026: Sonnet-class at $2 a million input tokens and $10 a million output tokens, the basis of the ₹16.25 narrative cost. Sourced, with the token count per report modelled. https://benchlm.ai/anthropic/api-pricing
- [17]OwlClaw — India PPC benchmarks 2026: real-estate cost a lead at ₹800–2,000. Sourced. https://owlclaw.com/benchmarks/ppc-benchmarks-india/
- [18]RaiseIQ — Indian seed benchmarks, August 2026: a typical raise of ₹3–8 crore. Sourced. https://raiseiq.in/blog/seed-valuation-benchmarks-india/
- [19]Centre for Studies of Plural Societies — 401 continuous ambient air-quality stations across 209 cities, MoEF&CC 2022. Sourced. https://cspsindia.org/air-quality-monitoring-measures-in-india
- [20]EY India — DPDP Rules 2025, with penalties reaching ₹250 crore. Sourced. https://www.ey.com/en_in/insights/cybersecurity/transforming-data-privacy-digital-personal-data-protection-rules-2025
- [21]NIRS scoring rubric, data model and financial model — the 25 signals across 4 dimensions and 3 intent profiles; the 22-screen inventory; the cost-to-serve lines behind ₹58.01; the ₹1,015 lifetime value and ₹309 acquisition ceiling; the four capital scenarios from ₹11.5 crore to ₹41.4 crore; and the Confident Coverage, correlation and conversion gates. Modelled and derived. Project record, not published.
BASIS & STAGE
NIRS is at specification: the rubric, data model, API contracts, screen inventory and financial model are complete, and no code has been written. No property has been scored against a retrieved government record, no user has used the product, and no pilot, waitlist or signed customer exists. The images are renders, not photographs. The central claim — that the score agrees with an expert valuer at Pearson r of 0.60 or above — is unverified. Every cost, margin, acquisition-cost, revenue, capital and volume figure is modelled, not measured. A user quote was cut because no buyer interview has been conducted. Next gate: lawful, repeatable access to Karnataka land records across 1,000 e-Khata addresses, at near-zero cost in Week 1.
PRODUCT RENDERS
