Propti — From WhatsApp Chaos to Instant Matches
Overview
Propti is a fully offline Android app I designed and built to solve one of the most painful daily problems for Indonesian property agents: finding the right property for the right buyer, fast.
Before Propti, agents managed hundreds of listings and buyer requests scattered across WhatsApp chats, screenshots, phone notes, and WhatsApp groups they created with only themselves as the member — just to have somewhere to save things. A single matching query took 15–30 minutes of manual scrolling. Deals got missed. Clients got frustrated.
Propti turned that 30-minute search into a 3-second match.
The Problem
Who It's For
Indonesian property agents are heavy WhatsApp users. They receive listing details, buyer requests, and transaction updates almost entirely through chat. Most are solo agents or part of small agencies. They are not tech-averse, but they work in the field and need tools that are fast, mobile, and work offline.
The Real Pain
Through conversations with agents early in the process, three pain points emerged clearly:
- Lost listings. Property details shared in WhatsApp groups get buried within hours. Agents had to re-scroll hundreds of messages, send "ada yang punya listing X ga?" (does anyone have a listing like X?) again and again.
- Missed matches. An agent might have the perfect listing for a buyer's request — but wouldn't know, because the connection was never made. Deals slipped through.
- Tool friction. The alternatives — spreadsheets, notes apps, cloud CRMs — all required internet, desktop setup, or manual re-entry that didn't fit the agents' mobile-first, in-the-field workflow.
The core insight: the matching problem was not a data problem. It was a retrieval and connection problem. Agents had the data. They just couldn't access and link it in time.
Goals & Success Criteria
The product had one north-star metric: time from buyer request to first matching property surfaced.
Supporting goals:
- Reduce listing input time from 5+ minutes (manual) to under 3 seconds (paste & parse)
- Surface matches automatically with a score agents could trust (≥70% threshold)
- Work fully offline — no backend dependency for core features
- Enable one-tap WhatsApp share when a match was found
Non-goals were equally important to define early: Propti is not a marketplace. Buyers do not use the app. There is no public listing feed. This keeps the product focused and avoids complexity that would delay the core value.
Discovery & Research
Shadowing the Workflow
I spent time with agents watching how they actually work: checking WhatsApp, searching through gallery folders for property photos, copy-pasting specs into reply messages. The workflow was entirely text-and-chat-based.
Key finding: the single most valuable moment in an agent's day is when a buyer asks "ada rumah budget X, area Y?" (any house with budget X in area Y?) — and the agent can reply in under 1 minute with a formatted listing, photos included. That moment builds trust and closes deals. Everything else is administration.
Existing Tool Audit
- WhatsApp itself: the de facto CRM. Infinite scroll, no search by attribute, no structure.
- Excel / Google Sheets: used by some; painful on mobile, requires internet, no matching logic.
- Cloud property CRMs: desktop-first, require ongoing internet, steep learning curve.
None of these gave agents a mobile-first, fully-offline, matching-native tool.
Design Decisions
1. Paste & Parse as the Primary Input Method
Instead of asking agents to fill out a form field-by-field, I built a text parser that accepts a WhatsApp message dump and auto-extracts all structured fields: price, bedrooms, bathrooms, land area, building area, certificate type, location, and more.
This lowered the input friction from ~5 minutes per listing to under 3 seconds. Agents could paste 10 listings from a WhatsApp group and review them in a batch — Propti parsed all of them in one go.
Tradeoff: the parser has to handle very messy, inconsistent text. A rule-based parser was chosen over ML to keep it fully offline and deterministic. It covers ~85% of real-world formats with graceful fallback to manual fields for the rest.
2. Score-Based Matching with a Trust Threshold
The matching engine scores every property against every open buyer request across multiple criteria: price range, bedrooms, bathrooms, land and building dimensions, property type, certificate preference, condition, furnishing, facing direction, and GPS proximity.
The 70% threshold was set based on testing with real data: below 70%, matches felt random and eroded trust. Above 70%, agents consistently said "ya ini memang cocok" (yes, this is actually a good match).
3. One-Tap WhatsApp Share
When a match is found, the agent shouldn't have to re-type the listing details. Propti generates a pre-formatted WhatsApp message with the property title, specs, price, and photos — ready to send in one tap.
This closes the loop from "match found" to "message sent" in under 10 seconds.
4. Fully Offline Architecture
All core data lives on-device in an Isar database. No backend is required for listing management, matching, contacts, transactions, backups, or exports. This was a hard requirement — agents frequently work in areas with poor connectivity.
The only exception is subscription management via Google Play. Everything else — including ZIP backup and Excel export/import — is fully offline, just local file operations on the device.
5. New Match Badge
The dashboard shows strong matches (≥70%) by default, with a "New" badge on any that were added since the last visit — so agents immediately know what's worth acting on.
6. Potential Earnings on the Dashboard
Most CRM tools show you what you've done. Propti also shows what you could earn — a "potensi komisi" (potential commission) number calculated from your open matched listings and buyer requests, using your own configurable sell and rent commission rates.
The calculation covers all matched listings and buyer requests — agents can also toggle between the ≥70% strong matches view and "semua" (all matches, any score) to see how the earnings estimate changes. The total shows as two numbers — from listings, and from buyer requests.
The design insight was that agents often don't realize how much potential value is sitting in their app waiting to be acted on. Showing the number makes the matches feel urgent, not just informational. It answers "why should I keep adding listings?" in rupiah.
Key Features Shipped
- Paste & Parse: auto-parse property/request text from WhatsApp into structured fields
- Matching Engine: score-based buyer-property matching (70%+ threshold, two-sided)
- Potential Earnings: live commission estimate from matched listings and buyer requests
- Dashboard: at-a-glance matches, active rentals, upcoming birthdays, potential earnings
- Properties & Requests: full CRUD with photo/video gallery, map view, Excel export
- Contacts: CRM with phone import, WhatsApp quick-dial, birthday tracking
- Transactions: commission tracking, rental period management, file attachments
- Backup & Import: ZIP export/import, Excel import/export, full media included
- Referral Program: referral codes with tiered rewards, "5 referrals = 1 free month"
- Multilingual: English and Indonesian
- WhatsApp Integration: deep-links with pre-filled messages, E.164 number normalization
Metrics & Outcomes
The activation signal is clear: paste → parse → see a match. When an agent completes this flow in their first session, retention is high.
Key results from early users:
- Listing input time: from ~5 minutes manual → under 3 seconds with paste & parse
- Match discovery time: from 15–30 minutes of manual WhatsApp search → under 5 seconds
- WhatsApp share time: from 2–5 minutes of manual copy-paste → under 2 seconds
- Pricing: Rp 99.000/month (launch) with monthly, yearly, and lifetime options; referral program drives organic growth
Lessons Learned
1. YAGNI applies to product too. Early temptation was to build a full cloud sync, multi-agent shared database, and buyer-facing portal. Staying fully offline and single-agent-focused shipped a working v1 months faster. Features for hypothetical future needs are features that don't ship.
2. Both sides of the workflow drive adoption AND retention. Easier input (paste & parse) wasn't just the unlock that got agents in the door — it's also why they keep coming back, because adding data never feels like work. And the output side (getting a match, sending a perfect WhatsApp message in one tap) isn't just a delight moment — it's also what makes agents adopt it in the first place, because they can see the payoff immediately. The two aren't sequential. They reinforce each other.
3. Match trust is a binary. If the matching algorithm surfaces noise, agents stop looking at matches entirely. The 70% threshold wasn't arbitrary — it was calibrated to maintain agent trust. A match that wastes an agent's time is worse than no match.
4. Fully Offline is a feature, not a constraint. Positioning Propti as "works without internet" resonated strongly in a market where field agents deal with spotty connectivity daily. Privacy ("data sepenuhnya milik Anda") also resonated — agents don't want their listings on someone else's server.
Tech Stack
- Flutter (Android-first, Material Design 3)
- Isar (offline embedded database, stream-based real-time updates)
- OpenStreetMap / Nominatim (location search)
- Google Play Billing (subscription management)
Try Propti
Propti is available on Google Play. A 7-day free trial is included.
