Case study · realtor system
Listings in WhatsApp groups, clients in DMs, shortlists built by hand. We built a realtor system for 1,106 listings
RealtyOS is our own system for real estate on Mexico's Riviera Maya coast. Broker messages become clean listings, the system matches them to what each client is looking for, builds shareable shortlists, and every morning a Telegram briefing says who's waiting on a reply. The public side runs as the somaya.mx catalog in three languages.
- 1,106
- listings in the catalog as of Sep 29, 2026
- 3,207
- URLs in the sitemap, 7 coastal areas in three languages
- 111
- leads moved over from WhatsApp Business
- 114
- automated tests covering matching, duplicates and money
How it used to work
Brokers post listings as messages in WhatsApp groups: price, area, commission, photos, all free-form. The same condo shows up in several groups at different prices, some in pesos, some in dollars.
Clients write in DMs and on WhatsApp Business. To send a shortlist, you have to remember who wanted what and dig the listing out of a chat. Follow-ups live in your head, and the client goes with whoever answered first.

What the system had to solve
- 1
One database instead of twenty chats
A broker's message becomes a listing card: price, type, area, commission. Duplicates are caught by similar text and a price within 10%, and pesos never get mixed with dollars.
- 2
Matching without manual searching
Every listing gets a 0-100 score against each client's search. At 85 or higher, the system suggests who to send it to.
- 3
A shortlist the client browses on their own
A link in three languages: the client likes a listing or passes with a reason, and the lead moves forward on its own.
- 4
Mornings where you know who to message
A Telegram briefing: hot clients, today's showings, who's owed a reply, which listings to double-check with the broker.
- 5
A catalog people find in search
A page for every coastal area in English, Spanish and Russian, plus a sitemap and search markup.
What the system is made of
Listings
- Import from WhatsApp groups and a broker database, with AI parsing each message
- Duplicate detection, currencies stored separately
- A daily freshness check: questions for brokers are drafted automatically
- Listing copy in three languages generated from the card
Clients and deals
- A client card with their search
- 0-100 matching and an alert on strong matches
- Showings with a calendar invite and a reminder 2 hours ahead
- A deal pipeline, pesos and dollars kept apart
Every day
- A morning Telegram briefing
- Follow-up drafts, sent by the realtor from their own WhatsApp
- A Monday review of the week: what worked, what didn't
The somaya.mx catalog
- English, Spanish, Russian
- 7 coastal areas, a page for each in every language
- An inquiry form with spam protection and a Telegram alert
What it gives a realtor
Nobody digs through chats for a listing anymore: it's in the database, priced in the right currency, with a note on when the broker last confirmed it's available. New listings find matching clients on their own, and every morning there's a list of who to message.
We started on Aug 28, 2026. All seven stages of the first plan are done, and we're now working on the full deal cycle. Stack: Next.js, Prisma, a background worker and the Claude API with a spending cap.
Honest status
This is our own project. We don't publish client or broker names or deal amounts. We'll share results in deals and catalog inquiries once there's enough data.
Questions, answered
Do I have to stop using WhatsApp?
No. Messages to clients go out from your own WhatsApp, by your hand. The system drafts them and reminds you. We never hook a personal number up to automated bulk messaging, because that's an easy way to get it banned.
Will it work outside the Riviera Maya?
The logic is the same anywhere: listings from chats, matching, shortlists by link, a morning briefing. Areas and the catalog are set up for your city. We start with a free review of how you work with listings today.
What does it cost?
We quote it after the review, using the same packages as our custom CRMs: from a single listings-and-clients feed up to a full system with a public catalog. Package prices are on the custom CRM page.
See also
All projects: websites, CRMs, apps
What we've already built, with one fact you can check for each.
Custom CRM
Packages and prices, from an order feed to a full system with money.
Case study · a system for the business
Four channels, four hundred orders a month, one system. Repeat orders up from 28% to 36% since May
Read the case
Case study · money & operations
The money came in, but the order still says unpaid. Six leaks we found in our own shop
Read the case
Free listings review
Realtor? Show us where your listings and clients live
Broker chats, spreadsheets, notes. We'll review it for free and tell you what to pull into one database first.
- Where listings and clients get lost between chats
- What to pull into one database first
- A rough estimate and timeline
We reply in writing within a couple of hours. No calls needed and no sales script.
Numbers you can check
- ≈400
- orders a month in the flower shop I run
- 36.2%
- of orders in Sept 2026 came from returning customers, up from 28.1% in May
- 10 years
- that site in search, full history on the table
The same case shows a 51% Google drop in May 2026. Most people hide that. I show it and write what I am doing about it.