PropPie Broker -- B2C Virtual Honest Broker¶
"Talk to it like your wealthiest uncle who actually read every document before answering -- and shows you the page he's quoting from."
1. Product identity¶
| Field | Value |
|---|---|
| Product name | PropPie Broker |
| Internal codename | Broker / VHB (Virtual Honest Broker) |
| Tagline | "The proptech that shows its work." |
| Status | In design (foundation phase) |
| Target launch | Private beta Q3 2026 (Pune residential); public launch Q1 2027 (Maharashtra) |
| Parent company | PropPie |
| Shared data foundation | MahaRERA + IGR + GRs + News/Social, ~140 attributes (data-attributes.md) |
| Sibling products | PropPie Fractional (live at proppie.in); PropPie Analytix (B2B, in design) |
Target users¶
| Persona | Description | Typical budget | Key need |
|---|---|---|---|
| Retail investor | Buying first or second residential property in Maharashtra | 50L -- 2 Cr | Cut through broker noise; understand true cost; verify builder |
| HNI | Diversifying into RE from equities/MFs; wants nuance and tax-aware framing | 2 Cr -- 20 Cr | Yield comparison, capital gains simulation, portfolio concentration risk |
| NRI | Remote due diligence from Singapore, Dubai, US; FEMA-aware repatriation | 1 Cr -- 10 Cr | Title chain confidence, FEMA compliance surface, power-of-attorney implications |
2. Jobs to be done¶
Written as user utterances the bot must handle. Grouped by intent.
Discover and explore¶
- "What's happening in the Hinjewadi residential market right now?"
- "Show me all 2BHK projects in Wakad under 90 Lakh with OC received."
- "Which micromarkets in Pune have the highest transaction velocity this quarter?"
- "Is there anything new from MahaRERA in Kharadi in the last 30 days?"
Compare and evaluate¶
- "Compare these two projects side by side -- Godrej Infinity vs Lodha Belmondo."
- "How does this project's price per sqft compare to the micromarket median?"
- "Show me the ASR gap for this society."
- "What's the cap rate for 2BHK rentals in Lower Parel vs BKC?"
Understand costs¶
- "What's the TOTAL cost of buying this flat -- not just the sticker price?"
- "Break down stamp duty, registration, GST, legal fees, society transfer, MSEDCL deposit -- everything."
- "How much more will I pay if I'm buying this as a second property?"
- "What are the annual holding costs after I move in?"
Verify the builder¶
- "What's this developer's track record? How many projects delivered, how late?"
- "Any MahaRERA complaints against this promoter?"
- "Has this developer's project cost overrun the original estimate?"
- "How does this developer compare to other Pune promoters on delivery timelines?"
Check the title and legal¶
- "Walk me through the title chain for this property."
- "Are there any encumbrances or litigation on this land parcel?"
- "Is the NA conversion done? Is there a CERSAI charge?"
- "What's the CRZ status for this coastal project?"
Simulate and project¶
- "If I buy at 1.18 Cr and hold for 5 years, what does the wealth trajectory look like?"
- "What's the projected IRR at the 20th, 50th, and 80th percentile?"
- "Run a rent-vs-buy comparison assuming 8% home loan rate."
- "What happens to my return if the metro gets delayed by 2 years?"
Understand policy¶
- "Any recent Government Resolutions that affect FSI in this area?"
- "What does the new premium FSI GR mean for prices in Mulund?"
- "Is there a TOD policy near the upcoming metro station?"
NRI-specific¶
- "I'm an NRI in Singapore. What FEMA rules apply if I buy residential in Pune?"
- "Can I repatriate the sale proceeds if I sell in 3 years?"
- "What power-of-attorney considerations should I know about for remote purchase?"
3. Core capabilities¶
What PropPie Broker CAN do. Each capability maps to the information-only stance: compare, explain, simulate, flag, contextualise -- never recommend.
3.1 Source-cited property lookup¶
Return structured, attributed information on any MahaRERA-registered project in Maharashtra. Every factual claim links back to its source document (RERA filing page, IGR deed reference, GR PDF page number).
3.2 Side-by-side comparison¶
Compare up to 4 properties across standardised dimensions: price per sqft (carpet), developer track record, title clarity score, hidden costs, infrastructure proximity, risk flags. Output as a structured comparison table with per-cell citations.
3.3 Hidden costs breakdown¶
Calculate the full acquisition cost beyond the listed price: stamp duty (rate by buyer category and property value), registration fee, GST (under-construction vs ready), legal/conveyancing fees, society transfer charges, MSEDCL/utility deposits, maintenance corpus, brokerage (if applicable). Personalised to the user's residency status and property count.
3.4 Developer track record review¶
Present the developer's MahaRERA portfolio: projects registered, delivered, delayed (with median delay in months and percentile vs comparable promoters), cost overruns, complaints filed and resolved. Output is percentile-based and explainable; never asserts "this developer is bad."
3.5 Title chain walkthrough¶
Walk the user through the chain of ownership as reconstructed from TSR, 7-12 extracts, and mutation records. Flag gaps, recent mutations, Class II land, missing NA conversion, and CERSAI charges. Explain in plain English what each flag means for the buyer.
3.6 Wealth trajectory simulation¶
Monte Carlo simulation of property value and cash flow over a user-defined horizon (typically 3--10 years). Outputs IRR at the 20th, 50th, and 80th percentiles. Inputs include purchase price, financing structure, rental assumptions, appreciation assumptions (drawn from micromarket historical data), holding costs, and exit assumptions. Every assumption is stated, adjustable, and sourced where possible.
3.7 GR and policy impact explanation¶
Retrieve and explain Government Resolutions relevant to a property's location and segment. Classify each GR as tailwind (e.g., premium FSI reduction, TOD benefit) or headwind (e.g., new CRZ restriction, development freeze). Link to the original GR PDF with page reference.
3.8 Market context¶
Present micromarket-level signals: transaction velocity (90d/180d), median price per sqft, YoY price appreciation, absorption rate, lifecycle stage (emerging/growing/mature/cooling), and sector momentum. All derived from IGR registered transactions, not listing-portal asking prices.
3.9 Risk surface¶
Present the composite risk picture: zone risk index, flood/CRZ/forest flags, title clarity score, developer trust score, sentiment score. Each flag is explainable and traced to its source. The bot surfaces "brutal honesty flags" -- things worth knowing before making a commitment.
3.10 Persona-aware contextualisation¶
Use the investor persona (captured with consent) to contextualise outputs: an NRI sees FEMA implications surfaced automatically; an HNI sees capital gains framing; a first-time buyer sees "explain like I'm 5" defaults. The persona shapes presentation, not conclusions.
4. Hard constraints¶
What PropPie Broker CANNOT do. These are non-negotiable.
| Constraint | Rationale | Enforcement |
|---|---|---|
| No personalised investment advice | SEBI IA Regulations 2013 -- we are not registered and do not intend to register as IA | Prompt-level guardrails + output classifier + human audit loop |
| No "you should buy X" or "top pick" language | Same as above; also a SOUL commitment | Banned-phrase list in output filter (see tone-and-voice.md) |
| No guaranteed returns or definitive projections | Every projection is probabilistic with stated assumptions | Output template enforces range + percentile |
| No PII logging in plain text | DPDP Act 2023 | PII masked at ingestion; encrypted at rest; access audit |
| No pre-RERA or non-compliant project promotion | RERA compliance | Only MahaRERA-registered projects surfaced |
| No defamatory developer assertions | Defamation risk | Developer data is percentile-based, explainable, includes right-to-respond |
| No paid placement or ranking influenced by seller payments | SOUL commitment; business model alignment | Revenue model excludes developer advertising |
| No transaction facilitation (booking, payment) | We are info-only; transaction layer is out of scope until regulatory posture evolves | No payment integration, no booking flow |
5. Refusal patterns¶
Exact examples of how the bot handles advice-seeking questions. These are the scripts the AI must follow.
Pattern 1: Direct "should I buy?" question¶
User: "Should I invest in this Hinjewadi flat?"
Bot: "I can't tell you whether to invest -- that's your call. But I can show you the picture pretty completely:
Price: Listed at 1.18 Cr. Comparable units in this society (n=8, last 12 months) closed between 95L and 1.12 Cr (median 1.04 Cr). Asking is roughly 12% above recent transaction median.
Yield: At 1.18 Cr and current Hinjewadi 2BHK rents (35-40K/month, n=42 L&L registrations), gross yield is 3.5--4% -- typical for Pune residential.
Builder: Eight projects on MahaRERA, six delivered (median delay 5 months), one ongoing, one withdrawn. Two MahaRERA complaints, both around amenities, none resolved.
Want me to (a) build a 5-year wealth scenario with explicit assumptions, (b) compare to 2 other Hinjewadi options, or (c) walk through the title chain?"
Pattern 2: "Which is better?" question¶
User: "Which one is better -- Project A or Project B?"
Bot: "'Better' depends on what matters to you. Here's how they compare on the dimensions I can measure:
[Comparison table with price/sqft, developer percentile, title score, hidden costs, infra proximity, risk flags -- all cited]
I can dig deeper on any of these. Which dimension matters most to you?"
Pattern 3: "Is this a good deal?" question¶
User: "Is 85 Lakh a good price for this flat?"
Bot: "I can show you what comparable buyers paid. In this micromarket, the median price for similar-config units (2BHK, 750-800 sqft carpet) over the last 6 months was 78L (n=14 registered transactions, IGR). The range was 72L to 91L. At 85L, this is above median but within the range. The ASR value for this address is 68L -- so there's a 25% premium to government valuation, which is typical here.
Whether 85L is 'good' for you depends on your alternatives, your financing cost, and your timeline. I can model the scenarios if you'd like."
Pattern 4: Tax or legal question beyond scope¶
User: "How much capital gains tax will I pay if I sell in 3 years?"
Bot: "That's a question for your CA -- it depends on your full income picture, other gains/losses, and whether you reinvest under Section 54/54F. What I can show you: the holding period is short-term (under 2 years for non-residential, under 2 years for residential as of current rules), so STCG rates would apply at your slab rate. I can plug that into the wealth trajectory model with your stated tax bracket -- want me to run it?"
Pattern 5: Speculative question with no data¶
User: "Will this area appreciate because of the new airport?"
Bot: "I can't predict that with certainty. Here's what I do know:
The Purandar airport has been approved at the planning stage (GR reference: UD-2024/CR-xxx/UD-16, dated [date]). It does not have a confirmed completion date. Historical pattern: when Navi Mumbai airport was approved in 2017, properties within 10km saw a 12-18% price run-up in 12 months, but prices then stagnated for 4 years as the project slipped.
This area is currently 15km from the proposed site. I have [n] transactions in the last 6 months -- not yet showing an airport premium.
Shall I show the price trend over time and let you judge the trajectory?"
6. Key user flows¶
Text-based screen descriptions. Not mockups -- these describe what the user sees and does at each step.
6.1 First-time user onboarding (persona capture)¶
Goal: Capture enough about the user to personalise information framing. Not required to use the product -- but improves output relevance.
Step 1 -- Welcome screen
"Welcome to PropPie Broker. I'm your property research assistant -- I show you what the data says about any property in Maharashtra, and I cite every claim. I don't recommend, I don't sell, I don't take commissions.
To give you more relevant context, I can ask you 4 quick questions. You can skip any or all of them. Your data is stored only to personalise your experience and is never shared. [Privacy policy link]
Ready? Or jump straight to a property search."
Step 2 -- Persona questions (skippable)
- "What's your primary goal? (a) Buying a home to live in, (b) Investment for rental yield, (c) Investment for capital growth, (d) Tax planning, (e) Just exploring"
- "What's your budget range? (a) Under 50L, (b) 50L--1 Cr, (c) 1--2 Cr, (d) 2--5 Cr, (e) 5 Cr+, (f) Skip"
- "How long do you plan to hold? (a) Under 3 years, (b) 3--5 years, (c) 5--10 years, (d) 10+ years, (e) Not sure"
- "Are you a resident Indian or NRI? (a) Resident, (b) NRI, (c) Skip"
Step 3 -- Consent
"Got it. I'll use this to frame the information I show you -- for example, if you're an NRI, I'll surface FEMA implications automatically. You can update or delete this anytime from Settings. [Explicit consent checkbox]"
Step 4 -- First query prompt
"Ask me anything about a property, a micromarket, a developer, or a policy in Maharashtra. Or try one of these: - 'What's happening in Wakad right now?' - 'Compare two projects I'm considering' - 'Break down the full cost of a 1 Cr flat in Pune'"
Data stored: persona.investment_goal, persona.budget_range, persona.horizon_years, persona.residency_status -- all with consent flag.
6.2 Property lookup and exploration¶
Trigger: User types a project name, MahaRERA number, or address.
Bot action: Entity resolution against the canonical project store. If ambiguous (multiple matches), present options with MahaRERA numbers and addresses.
Output -- Property card:
[Project Name] | MahaRERA: P52100XXXXXX Developer: [Promoter Name] | Developer Trust Score: 72/100 (68th percentile vs Pune promoters) Location: [Full address] | Micromarket: Hinjewadi Phase 2 Type: Residential | Status: Under Construction | Expected completion: Dec 2027
Quick numbers (all cited -- tap any number to see source): - Price range: 78L -- 1.12 Cr (median 94L, n=23, IGR last 6 months) - Carpet area: 650 -- 1,100 sqft (1BHK to 3BHK) - Price per sqft (carpet): 7,200 -- 8,400 - Micromarket median: 7,800/sqft - Gross yield (current rents): 3.8% (L&L data, n=42) - Title clarity score: 85/100 - Infra proximity score: 78/100 (metro Phase 2 under construction, 1.2km)
What would you like to explore? (a) Full hidden costs breakdown (b) Developer track record (c) Title chain walkthrough (d) 5-year wealth trajectory simulation (e) Compare to similar projects (f) Risk flags and "things worth knowing"
6.3 Comparison mode (side-by-side)¶
Trigger: User asks to compare two or more projects (up to 4).
Output -- Comparison table:
| Dimension | Project A | Project B | Source |
|---|---|---|---|
| Price/sqft (carpet) | 7,800 | 8,200 | IGR, last 6 mo |
| Developer trust score | 72/100 (68th pctl) | 81/100 (79th pctl) | MahaRERA portfolio |
| Median delay (months) | 6 | 3 | MahaRERA completion dates |
| Title clarity score | 85/100 | 92/100 | TSR + 7-12 |
| Hidden costs (total %) | 11.2% of base price | 10.8% of base price | Calculator |
| Infra proximity | 78/100 | 65/100 | GIS + metro/SEZ |
| Gross yield | 3.8% | 3.2% | L&L + sale data |
| Micromarket stage | Growing | Mature | IGR time-series |
| Risk flags | 1 (flood zone adj.) | 0 | MRSAC + BMC |
| MahaRERA complaints | 2 | 0 | MahaRERA portal |
"Both projects are within your stated budget. The numbers above are all fact-based -- tap any cell to see its source. I haven't ranked them because 'better' depends on what you're optimising for.
Want me to run a side-by-side wealth trajectory simulation?"
6.4 Hidden costs breakdown¶
Trigger: User asks about total cost or hidden charges.
Output -- Cost breakdown table:
| Cost component | Amount | Basis | Source |
|---|---|---|---|
| Base price (as negotiated) | 95,00,000 | User input | -- |
| Stamp duty (5% for male, 4% for female, 6% for second property) | 4,75,000 | 5% (male, first property) | Maharashtra Stamp Act |
| Registration fee | 30,000 | 1% capped at 30,000 | Registration Act |
| GST (5% under-construction, 0% if OC received) | 4,75,000 | 5% (under-construction, no ITC) | GST Act |
| Legal / conveyancing fees | 15,000 -- 25,000 | Market rate, Pune | Estimate |
| Society transfer charges (resale only) | 0 | New booking | -- |
| MSEDCL / utility deposits | 10,000 -- 15,000 | Standard for 2BHK | MSEDCL tariff |
| Maintenance corpus (one-time) | 1,50,000 -- 2,00,000 | Per developer declaration, MahaRERA | MahaRERA filing |
| Parking (if separate) | 3,00,000 | Per developer price list | Price list |
| Total acquisition cost | 1,09,55,000 -- 1,10,20,000 | 15.3 -- 16.0% above base price |
Annual holding costs (post-possession): - Maintenance: 3,500 -- 4,000/month (declared in RERA filing) - Property tax: ~8,000 -- 12,000/year (PMC rates for this zone) - Insurance: optional, ~2,000 -- 3,000/year
"These numbers are for a resident Indian male buying a first property under-construction. If your situation is different (NRI, female buyer, second property, ready-possession), tell me and I'll recalculate."
6.5 Title chain walkthrough¶
Trigger: User asks about the title or ownership history.
Output -- Title narrative + timeline:
"Here's the title chain I've reconstructed from the TSR filed with MahaRERA and the 7-12 extract:
1. Original landowner: [Name], holding survey number [X], recorded in 7-12 extract dated [date]. 2. Sale to developer's land entity: [Company name] acquired via registered sale deed (IGR doc no. [X], dated [date]) for [amount]. 3. NA conversion: Non-agricultural use order granted by [authority], order no. [X], dated [date]. 4. Development agreement: Between [landowner entity] and [promoter], registered [date]. 5. BPA and CC: Building plan approved [date], commencement certificate [date]. 6. Current status: Under construction, expected completion [date].
Title clarity score: 85/100
Flags: - There is a mutation entry from 2018 with an unusually short gap between sale and re-sale (3 months). This could be benign (entity restructuring) or worth verifying with your lawyer. - No CERSAI charges found as of [date]. - No litigation records found in available court databases.
I'd suggest your lawyer specifically verify the 2018 mutation. I can generate a summary PDF of this chain with source links for them."
6.6 Wealth trajectory simulation¶
Trigger: User asks about returns, IRR, or "what happens if I hold for X years."
Step 1 -- Assumption capture:
"Let me run a wealth trajectory. I'll use these defaults (change any): - Purchase price: 1.18 Cr - Down payment: 20% (23.6L) - Home loan: 94.4L at 8.5% for 20 years (current median prime rate) - Annual appreciation assumption: based on this micromarket's 5-year IGR trend (6.2% median, range 2--11%) - Rental yield: 3.8% gross (current L&L data) - Rental escalation: 5%/year (IGR L&L median for this segment) - Holding costs: 15.6% of base (calculated above) - Exit costs: 1% brokerage + capital gains (per your tax bracket) - Horizon: 5 years
Anything you'd like to adjust?"
Step 2 -- Simulation output:
"Here's the 5-year wealth trajectory at three probability levels (1,000 Monte Carlo runs using this micromarket's historical volatility):
Metric Pessimistic (20th pctl) Base case (50th pctl) Optimistic (80th pctl) Property value at exit 1.38 Cr 1.59 Cr 1.84 Cr Total rental income (net) 18.2L 19.8L 21.4L Total EMI paid 47.6L 47.6L 47.6L Net wealth change -5.2L +12.4L +31.6L Pre-tax IRR on equity 2.1% 9.8% 16.2% Key assumptions that matter most: appreciation rate (drives 60% of variance), rental vacancy (assumed 1 month/year), and interest rate (assumed flat -- RBI could move).
What this does NOT account for: your full tax situation (consult your CA), emotional value of ownership, opportunity cost vs your specific alternatives.
Want me to stress-test a scenario -- e.g., what if metro gets delayed 2 years, or what if rates go up 100bps?"
6.7 Developer track record review¶
Trigger: User asks about a developer or promoter.
Output -- Developer card:
[Developer Name] | PAN: XXXXX (masked) | CIN: [if corporate]
MahaRERA portfolio: - Registered projects: 14 (since 2017) - Delivered: 8 (median delay: 6 months) - Under construction: 4 - Lapsed / withdrawn: 2
Delivery performance (vs comparable Pune promoters with 10+ projects): - Median delay: 6 months (65th percentile -- 35% of comparable promoters deliver faster) - Longest delay: 18 months (Project Y, MahaRERA P521XXXXX) - Most punctual: Project Z, delivered 1 month ahead of schedule
Cost discipline: - Average cost overrun: 12% above original MahaRERA Form B estimate - Industry median: 15% -- this developer is slightly better than average
Complaints: - Total MahaRERA complaints: 7 across portfolio - Resolved: 4 | Pending: 2 | Dismissed: 1 - Common themes: amenity shortfall (3), delayed possession (2), carpet area discrepancy (1), other (1)
Note: This summary uses MahaRERA public filings only. It does not capture projects registered before RERA (pre-2017) or projects in other states. The 65th-percentile placement means this developer delivers later than 65% of comparable promoters -- not a judgement, a measurement.
Want me to compare this developer to others you're considering?"
6.8 GR / policy impact query¶
Trigger: User asks about a Government Resolution or policy change.
Output -- GR impact card:
GR reference: UD-2025/CR-xxx/UD-16, dated [date] Subject: Reduction in premium FSI charges for residential projects in PMC limits Source: Maharashtra Urban Development Department ([link to GR PDF, page X])
What it does: Reduces the premium FSI charge from 50% of ASR to 35% of ASR for residential projects in PMC limits, effective [date], valid for 2 years.
Impact on this project: - This project is within PMC limits and has not yet consumed full permissible FSI. - If the developer adds floors under this GR, the cost of additional FSI drops by ~30%. - Historically, premium FSI reductions have been associated with 5-8% price reductions by developers in competitive micromarkets (based on IGR price trends after similar GRs in 2021 and 2023).
Classification: Tailwind (for buyer -- potential price softening; for developer -- lower input cost)
Caveat: GR validity is 2 years. Developer may or may not pass savings to buyer. Trend data is observational, not predictive.
Shall I check what other active GRs affect this location?"
7. Data dependencies¶
Which attributes from the ~140 attribute catalogue (data-attributes.md) power each flow.
7.1 Attribute mapping by flow¶
| Flow | Primary attribute categories | Key attributes |
|---|---|---|
| Property lookup | 01 Project Identity, 02 Location, 03 Area, 07 Financial | proj.maharera_number, proj.name, loc.*, area.carpet_sqft, fin.price_per_sqft_carpet, fin.asr_gap_pct |
| Comparison mode | 01, 02, 03, 04, 05, 07, 10, 12, 13 | All of the above + dev.trust_score, legal.title_clarity_score, risk.zone_risk_index, ai.comparable_set |
| Hidden costs | 07 Financial, 14 Persona | fin.stamp_duty_paid, fin.sale_price, persona.residency_status, ai.hidden_costs_breakdown |
| Title chain | 05 Legal, 06 Approvals | legal.chain_of_title, legal.encumbrances, legal.cersai_charges, legal.litigation_records, legal.title_clarity_score |
| Wealth trajectory | 07, 08, 10, 14, 15 Macro | fin.price_per_sqft_carpet, lease.rent_amount_monthly, mkt.price_appreciation_yoy_pct, macro.home_loan_rate_median, ai.wealth_trajectory_paths, ai.projected_irr_p50_p20_p80 |
| Developer review | 04 Parties, 01 Project Identity | party.promoter_name, party.promoter_pan, dev.trust_score, proj.promised_completion_dates, proj.actual_completion_date, fin.cost_overrun_pct, legal.complaints_maharera |
| GR / policy impact | 11 Policy & Infra | policy.applicable_grs, policy.tailwind_flags, policy.headwind_flags, ai.gr_impact_summary |
| Market context | 10 Market Signals | mkt.transaction_velocity_90d, mkt.median_price_per_sqft_90d, mkt.price_appreciation_yoy_pct, mkt.absorption_rate, mkt.micromarket_lifecycle_stage |
| Risk surface | 12 Risk, 05 Legal | risk.zone_risk_index, risk.flood_zone_flag, risk.crz_flag, legal.title_clarity_score, ai.brutal_honesty_flags |
| Persona contextualisation | 14 Persona | persona.* (all 8 attributes) |
7.2 Critical-path dependencies¶
These attributes MUST be production-ready before the corresponding flow can launch:
| Phase | Required attributes | Blocking flow |
|---|---|---|
| Private beta (Q3 2026) | Cat 01, 02, 03, 04, 07 (raw), 10 (90d velocity + median price) | Property lookup, basic comparison |
| Public launch (Q1 2027) | All of the above + Cat 05, 06, 08, 11, 12, 13 (hidden costs, title, developer, risk) | Full flow set |
| Persona + simulation | Cat 14 (persona), Cat 15 (macro), ai.wealth_trajectory_paths |
Wealth simulation, persona-aware |
7.3 Derived attributes with highest Broker dependency¶
These derived attributes from derived-attributes-spec.md are essential to the Broker experience:
| Derived attribute | Why it matters for Broker |
|---|---|
dev.trust_score |
Anchors developer review; must be percentile-based and defamation-guarded |
legal.title_clarity_score |
Anchors title walkthrough; user-facing confidence signal |
ai.hidden_costs_breakdown |
Powers the hidden costs flow; must personalise by persona |
ai.wealth_trajectory_paths |
Powers wealth simulation; must be Monte Carlo with stated assumptions |
ai.comparable_set |
Powers comparison flow; similarity matching across projects |
ai.brutal_honesty_flags |
The "Honest Broker" differentiator; surfaces what others would hide |
ai.persona_fit_score |
Personalises ranking/presentation order (info-only, not advice) |
risk.zone_risk_index |
Composite risk signal; user needs to see this before committing |
8. Channel strategy¶
| Channel | Priority | Notes |
|---|---|---|
| Web (responsive) | P0 -- launch channel | Full conversational UI; citation rendering; comparison tables; simulation charts. Desktop and mobile-responsive. |
| P1 -- fast follow | India's default messaging app. Rich-text formatting limitations require careful output adaptation. Quick lookups, alerts, follow-ups. Cannot render complex tables -- fall back to summary + "view full report on web" link. | |
| Mobile app (native) | P2 -- later | Only if web + WhatsApp retention justifies native investment. Not before Q2 2027. |
| Voice | P3 -- exploratory | Marathi/Hindi voice input for users uncomfortable with typing. Adds ASR complexity. Not before bilingual text is solid. |
OPEN QUESTION: WhatsApp Business API vs WhatsApp Cloud API? WhatsApp has message-template approval requirements that constrain proactive outreach. Needs scoping with Vishal on infra implications.
OPEN QUESTION: Do we build a standalone Broker web app or embed it within the existing proppie.in domain? Arguments for standalone: separate brand positioning for B2C vs fractional. Arguments for embedded: shared auth, cross-sell, single domain authority. Decision needed before Q3 2026 beta.
9. Multilingual strategy¶
| Language | Priority | Scope | Notes |
|---|---|---|---|
| English | P0 | Full capability | Launch language |
| Hindi | P1 | Conversational + summaries | Large addressable base; many Maharashtra buyers are Hindi-primary |
| Marathi | P1 | Conversational + summaries + GR source display | Critical for GR content (originals are in Marathi); rural Maharashtra users |
Approach: - User selects preferred language at onboarding (changeable anytime). - Conversational responses generated in the selected language. - Source citations remain in their original language (Marathi GRs stay Marathi; RERA filings stay English) with optional translation toggle. - Numbers and units follow Indian conventions in all languages. - Bot can accept mixed-language input (code-switching is common in Maharashtra -- "yeh flat ka carpet area kitna hai?").
OPEN QUESTION: Timeline for Hindi/Marathi support. The vision doc targets Q3 2027 for bilingual Broker. Can we ship English-first beta in Q3 2026 and add Hindi/Marathi incrementally? Need to assess LLM quality for Marathi real-estate domain terminology.
10. Monetisation model¶
Status: Not finalised. Three options under evaluation.
| Model | Description | Pros | Cons |
|---|---|---|---|
| Freemium | Free: basic property lookup, market overview. Paid: comparisons, simulations, developer deep-dives, title chain, alerts | Low barrier to trial; large free tier builds trust and habit | Free tier must be useful enough to hook but limited enough to convert; risk of "good enough for free" |
| Subscription | Monthly/annual subscription for full access. Tiered by usage or persona (Retail / HNI / NRI) | Predictable revenue; aligns incentives (we make money when user finds value, not when seller pays) | Cold-start problem; users may not pay before experiencing value |
| Per-report | Pay per deep-dive report (title chain, simulation, developer review). Basic lookup free | Lowest commitment; aligns payment to value delivered | Revenue lumpy; user friction at every paywall |
Likely direction: Freemium with a generous free tier (property lookup + basic comparison + market overview) and paid tier for simulations, developer deep-dives, title chain, alerts, and PDF export. Subscription preferred over per-report for retention. NRI tier may command premium pricing (higher willingness to pay, higher value delivered).
What we will NOT do: Monetise via developer advertising, paid placement, or selling user data. This is locked in the SOUL.
OPEN QUESTION: Pricing. No data yet on willingness to pay for Indian proptech B2C subscriptions. Comparable signals: PropEquity charges 1.5--2.5L/year for B2B; Housing.com premium is ~999/month; Zerodha (different asset class) proved Indians will pay for transparent financial services. Need user research before setting price.
11. Success metrics¶
Metrics are grouped by what they measure. Vanity metrics are excluded.
Trust metrics (most important)¶
| Metric | Target | Measurement |
|---|---|---|
| Source-citation rate | 100% of factual claims | Automated: every factual assertion in bot output has a citation link |
| Hallucination incident rate | <1 per 10,000 queries | Human audit sample + automated consistency checks |
| "This answer was honest" user rating | >85% positive | Post-interaction micro-survey (optional, non-intrusive) |
| Regulatory incidents | 0 ever | Internal compliance review + external counsel audit |
Engagement metrics¶
| Metric | Target (beta) | Target (public launch +6 mo) |
|---|---|---|
| Monthly active users (MAU) | 100 | 10,000 |
| Queries per session | >3 | >4 |
| Return rate (7-day) | >40% | >50% |
| Comparison flows initiated per user/month | >1 | >2 |
| Simulation flows completed per user/month | >0.5 | >1 |
Conversion and revenue metrics¶
| Metric | Target |
|---|---|
| Free-to-paid conversion rate | >5% (benchmark: B2C SaaS India) |
| Paid subscriber churn (monthly) | <8% |
| Revenue per paid user per month | TBD (depends on pricing model) |
| NPS | >50 |
Data quality metrics (upstream dependency)¶
| Metric | Target |
|---|---|
| Attribute freshness SLA compliance | >95% of attributes within declared cadence |
| Extraction confidence (median across critical attributes) | >0.85 |
| Entity resolution accuracy (promoter-level) | >95% |
12. Phasing¶
Phase 1 -- MVP (Private beta, Q3 2026)¶
Scope: Pune residential only. English only. Web only. Invited users (100).
| Capability | Included | Notes |
|---|---|---|
| Property lookup (by name / MahaRERA number) | Yes | Core entity resolution must work |
| Basic property card (price, area, developer, location) | Yes | Subset of full card |
| Side-by-side comparison (2 projects) | Yes | Limited dimensions |
| Hidden costs breakdown | Yes | Static rate assumptions; no persona personalisation yet |
| Market context (micromarket overview) | Yes | 90d velocity + median price only |
| Developer track record (basic) | Yes | Project count, delay stats, complaints |
| Title chain walkthrough | No | Requires legal data pipeline maturity |
| Wealth trajectory simulation | No | Requires macro + simulation engine |
| GR / policy impact | No | Requires GR classifier in production |
| Risk surface (composite) | No | Requires enrichment feeds |
| Persona capture | Yes | Stored but minimally used in Phase 1 |
| Refusal patterns | Yes | Non-negotiable from day 1 |
| Citation on every claim | Yes | Non-negotiable from day 1 |
Success criteria for Phase 1: - 100% citation rate on factual claims (no exceptions) - Zero hallucination incidents - >70% of beta users say "this is more useful than what I had before" (exit survey) - Entity resolution works for >90% of Pune residential projects in MahaRERA - Latency: <5 seconds for property card; <15 seconds for comparison
Phase 2 -- Public launch (Q1 2027)¶
Scope: All of Maharashtra. English + Hindi. Web + WhatsApp.
| Capability | Included | Notes |
|---|---|---|
| All Phase 1 capabilities | Yes | Expanded to Maharashtra-wide |
| Title chain walkthrough | Yes | Legal data pipeline must be ready |
| Wealth trajectory simulation | Yes | Monte Carlo with stated assumptions |
| Developer deep-dive (full) | Yes | Percentile-based, defamation-guarded |
| GR / policy impact | Yes | GR classifier + retrieval in production |
| Risk surface (composite) | Yes | Zone risk index, flood/CRZ flags |
| Comparison (up to 4 projects) | Yes | Full dimension set |
| Persona-aware contextualisation | Yes | NRI flows, HNI framing, first-time-buyer defaults |
| Hidden costs (persona-personalised) | Yes | Residency, gender, property count |
| WhatsApp channel | Yes | Summary mode + web link for full reports |
| Hindi language support | Yes | Conversational + summaries |
| Freemium paywall | Yes | Free tier + paid tier live |
Phase 2 additions under consideration (not committed): - Marathi language support (depends on LLM quality assessment) - Alerts / monitoring ("tell me when a new project launches in this micromarket") - PDF export of full property report (paid feature) - Fractional cross-sell integration (surface PropPie Fractional assets where relevant)
13. Risks and mitigations¶
Risks specific to PropPie Broker (product-level). Company-wide risks are in vision.md.
| # | Risk | Likelihood | Impact | Mitigation |
|---|---|---|---|---|
| 1 | SEBI reclassifies conversational AI as Investment Adviser activity | Low-Medium | Critical | Our info-only stance is designed to stay inside the line. If SEBI moves the line, we trim outputs to comply. Legal counsel monitors SEBI circulars quarterly. Worst case: register as IA (expensive, slow, but survivable). |
| 2 | AI hallucination in production | Medium | High | Multi-layer defence: RAG with deterministic retrieval, output classifier for unsourced claims, citation-mandatory templates, human audit sample (1% of queries), incident response SOP. |
| 3 | Defamation claim from a developer | Low-Medium | High | All developer data is percentile-based, cited, and includes the right-to-respond mechanism. Legal review of all developer-facing output templates. D&O insurance. |
| 4 | MahaRERA / IGR portal changes break data pipeline | Medium | High | Vishal's pipeline design must include scraper resilience (format versioning, fallback extraction, alert-on-break). Multiple extraction strategies per source. This is Vishal's domain -- Broker spec defines the contract, not the how. |
| 5 | User expects advice, gets frustrated by refusals | High | Medium | UX challenge: make the information so rich and well-framed that the user doesn't feel refused -- they feel empowered. The refusal patterns above are designed to be warm, not bureaucratic. Beta feedback will calibrate this. |
| 6 | Low willingness to pay for B2C proptech | Medium | High | Freemium de-risks: build habit on free tier, convert on depth. NRI segment likely has higher WTP. Per-report option as fallback. Worst case: Broker becomes a top-of-funnel for Fractional and Analytix. |
| 7 | LLM cost per query makes unit economics negative | Medium | Medium | Aggressive caching of property cards and derived attributes (most queries hit cached data, not live LLM). LLM only for conversational wrapping and novel queries. Monitor cost-per-query weekly. Model-swap flexibility (no vendor lock-in). |
| 8 | Competitor launches similar product faster | Medium | Medium | Speed is not the moat -- depth + citation + compliance are. A competitor can launch a chatbot; they cannot replicate the data fusion, the citation discipline, or the business-model alignment quickly. See moat-and-differentiation.md. |
| 9 | DPDP Act compliance incident (PII exposure) | Low | Critical | PII never stored in plain text. Persona data is minimal and consent-gated. No Aadhaar, no PAN stored from users. Audit trail on all data access. See SOUL.md and the compliance skill. |
| 10 | WhatsApp policy changes restrict bot interactions | Low-Medium | Medium | WhatsApp is P1, not P0. Web is the primary channel. WhatsApp restrictions would limit reach but not kill the product. Monitor Meta's policy updates. |
14. Cross-references¶
| Document | Relevance to Broker |
|---|---|
| SOUL.md | The identity document. Every Broker output must be testable against it. |
| tone-and-voice.md | Writing style, refusal patterns, banned phrases, probability framing. |
| vision.md | Strategic milestones, success metrics, positioning. |
| moat-and-differentiation.md | Why the Broker is defensible; the five moat layers. |
| data-attributes.md | The ~140 attribute catalogue that powers every Broker flow. |
| derived-attributes-spec.md | Math specs for every derived score the Broker displays. |
| data-quality-framework.md | Quality rules that determine when data is good enough to show a user. |
| pipeline-spec-for-vishal.md | The contract between this product spec and Vishal's pipeline. |
| AGENTS.md | Operating manual for AI agents working in this repo. |
Open questions register¶
Consolidated list of open questions surfaced in this spec. These should be tracked in open-questions.md as they are resolved.
| # | Question | Owner | Blocking? |
|---|---|---|---|
| OQ-1 | Standalone Broker web app vs embedded in proppie.in? | Aishvarya + Vishal | Yes (before beta) |
| OQ-2 | WhatsApp Business API vs Cloud API? | Vishal | No (WhatsApp is P1) |
| OQ-3 | Pricing model and price points for paid tier? | Aishvarya | No (before public launch) |
| OQ-4 | Hindi/Marathi LLM quality for RE domain terminology -- good enough? | Aishvarya + dev team | No (before Q3 2027) |
| OQ-5 | Voice input feasibility and timeline? | Vishal | No (P3) |
| OQ-6 | Fractional cross-sell integration -- when and how? | Aishvarya | No (Phase 2+) |
| OQ-7 | Legal review of refusal patterns -- external counsel sign-off? | Aishvarya | Yes (before beta) |
| OQ-8 | Cost-per-query budget and LLM model selection? | Vishal | Yes (before beta) |
| OQ-9 | Alert / monitoring feature scope and notification channel? | Aishvarya | No (Phase 2) |
| OQ-10 | User research on willingness to pay (qual interviews)? | Aishvarya | No (before pricing decision) |
This document is the source of truth for PropPie Broker product scope. It should be updated as decisions are made and open questions are resolved. Every change should be reflected in MEMORY.md.
Last updated: foundation session, May 2026.