Real Estate Portal with Geosearch and Mortgage Calculator
We develop turnkey real estate portals that combine map-based search, filters, a mortgage calculator, and CRM integration. Each module demands performance and scalability: a typical project handles up to 100,000 listings at 500 requests per second. With over 10 years of experience and 20+ completed projects, we solve challenges in geosearch, duplicate listings, and feed automation. Below are technical details to help you avoid common pitfalls and save your development budget.
Why PostGIS speeds up geosearch?
The main problem in geosearch is spatial query performance. A typical SQL query with a radius condition on 200,000 objects takes tens of seconds. We use PostGIS with GIST indexes — query speed is 3x higher. On one project with 200,000 listings, API response time for a radius query never exceeded 100 ms.
Comparison of geosearch approaches:
| Method |
Query Time (100k objects) |
Implementation Complexity |
| PostGIS GIST |
50–200 ms |
Medium |
| Elasticsearch geo_shape |
200–500 ms |
High |
| Client-side filtering |
2–5 sec (load all data) |
Low |
PostGIS with GIST indexes is 2–3 times faster than Elasticsearch geo_shape on typical radius queries, according to our load tests. For marker clustering on zoom out we use supercluster (npm) on the client or ST_ClusterKMeans on the server. This reduces rendering load and speeds up map interaction.
PostGIS vs Elasticsearch
Elasticsearch is great for full-text search, but spatial queries are more native and faster in PostGIS due to GIST indexes. Also, PostGIS is a PostgreSQL extension, so it doesn't require separate infrastructure. We use Elasticsearch only for text search on listings or for statistics aggregation.
Steps to configure the YRL feed for listings
- Create an XML file according to YRL schema 2.0 with each object details (address, price, area, coordinates, images).
- Set up a cron job on the server to regenerate and submit the feed daily.
- Use Laravel or similar framework with an XML parsing package and send via HTTP.
- Test feed correctness using Yandex Webmaster tools.
Setup for one standard takes 1–2 days. We also support CIAN XML and API, Realty.mail.ru.
Benefits of proper YRL implementation
Proper YRL ensures listings appear in Yandex search without errors. Automating conversion and submission via cron eliminates manual work and increases update frequency. This directly affects the number of leads from aggregators. Our clients typically save up to $50,000 annually by automating feed generation and reducing manual data entry.
Advanced filters
Search by:
- Type (apartment / house / plot / commercial)
- Room count (studio, 1–4+ rooms)
- Floor (not first / not top / any)
- Price range (in rubles or per m²)
- Area (total, living, kitchen)
- Year built / condition
- Building series (panel / brick / monolithic)
- Metro + radius on foot/by car
- District / microdistrict / residential complex
Each filter combination generates a unique SEO-friendly URL.
How new listing notifications boost conversion
A user saves a search query. When a new listing appears, an email or push notification is sent. Implementation: a cron job runs once per hour matching new listings against saved searches. This increases user return rate by 30–40%.
Mortgage calculator
Interactive calculation: property price, down payment (%), term, interest rate → monthly payment, total overpayment. Integration with bank APIs to display real programs (Sberbank, DOM.RF). Setup takes 2–3 days. Adding this module lowers cost per lead — users instantly assess their options.
SEO for real estate portal
Listing pages with filters should have unique URLs and content: e.g., kupit-kvartiru-moskva/chertanovo-severnoe/3-komnatnye. Programmable content from filters for titles and descriptions. RealEstateListing schema.org markup. Additionally, static pages for popular districts improve indexing and attract organic traffic.
What is included in the work
- Analytics and architecture design
- Frontend on React 18/Next.js 14 with TypeScript and Tailwind
- Backend on Laravel 11 (PHP 8.3) or Node.js (Nest.js)
- Integrations with CRM (AmoCRM, Bitrix24), banks, aggregators
- Testing (unit, integration, load) and deployment to your hosting
- Documentation, access credentials, and admin training
- 1-month warranty support
Development cost: from $15,000 for MVP, full portal from $40,000. Save up to $50,000 annually on feed automation. Contact us for an audit of your project — we'll calculate timelines and cost.
Estimated timelines
| Stage |
Time |
| MVP (catalog, filters, map, inquiry form) |
from 2 to 3 months |
| Full portal (CRM, geosearch, alerts, 3D tours, mortgage) |
from 4 to 6 months |
Get a consultation on your project — we'll assess the scope and propose the optimal solution.
Development of Corporate Portals and Internal Systems
We specialize in developing corporate portals — CRM, ERP, LMS, and Intranet. Each project starts not with landing page layout, but with how business rules will be embedded into the architecture: who sees which data, how 1C and accounting systems sync, how 500 contacts turn into 500,000 without performance degradation. Over 7 years, we have delivered over 40 portals for companies with 50 to 5000 employees. We will evaluate your project in two business days — just contact us.
A public website can be launched without detailed design — iteratively improved based on feedback. With a corporate portal, this approach does not work: the cost of fixing architectural decisions after launch for 200 users is incomparably higher. Therefore, we spend 70% of our time on analysis and prototyping, and write code only after the role matrix and integration scheme are approved.
Three areas where bad decisions are often made: access rights model, performance on large data, and real-time updates.
How to build a role model for 30 departments?
Access rights model. "A manager sees only their own clients, a department head sees the entire department, a director sees the whole company, but financial data is visible only to the CFO and above." This is not three roles — it's a matrix of roles, permissions, organizational units, and record ownership. Implementing it with if ($user->role === 'manager') in controllers will make the code unmaintainable after six months.
The correct approach: Spatie Laravel Permission for basic role model + Policy classes for object-level permission (can('view', $deal) checks not only the role but also ownership). For complex hierarchical structures — ABAC (Attribute-Based Access Control) instead of RBAC.
Performance on large data. CRM with 500,000 contacts, filtering by 10 fields, sorting by activity — a naive implementation yields 15-second queries. Composite indexes, denormalization of aggregates (last_activity_at on the record itself instead of MAX over related table), Elasticsearch for full-text search on contacts.
Real-time updates. Multiple employees working on the same document or task. Without WebSocket — constant setInterval with polling every 5 seconds, extra server load, update delays. Laravel Broadcasting + Pusher/Soketi or a custom WebSocket server on Node.js — for notifications and real-time changes.
CRM Systems
Typical set: contacts, companies, deals, activities, sales funnel, reports. Technically straightforward. The complexity lies in the details.
Pipeline with custom stages. Every company wants its own funnel. Stages must be configurable without deployment. Table pipeline_stages with position, color, is_final, probability — and drag-and-drop for reordering on UI (React DnD or dnd-kit).
Change history. Who and when changed a deal status, reassigned a responsible person, added a note. Audit log via Observer or spatie/laravel-activitylog. On UI — timeline with filtering by activity type.
Email integration. IMAP/SMTP for connecting corporate mailbox, automatic linking of incoming emails to contacts by email address. This works reliably only with proper handling of bounces, spam, auto-replies — filtering is required.
Why is ERP not about code but about data?
ERP is when CRM, warehouse, production, accounting, and HR are unified into a single system. Full ERP from scratch is rare (usually integrating with existing systems), but modular systems for specific businesses are common.
Key principle: financial operations must be immutable. Not UPDATE orders SET status = 'cancelled' — but creating a new record order_cancellations with a reference to the original order. This is the immutable ledger principle, which simplifies auditing and reconciliation.
Integration with 1C is almost always part of an ERP project. Two-way synchronization: from 1C to portal (directories, balances, prices) and from portal to 1C (orders, documents). RabbitMQ as an event bus between systems is more reliable than direct HTTP interaction — if 1C is unavailable, messages wait in the queue.
How are LMS structured: learning platforms?
Learning Management System — courses, modules, lessons, tests, certificates, user progress.
Video content is the most demanding part of an LMS. Storing video on your own server and serving via Nginx is a bad idea: expensive, slow, no adaptive bitrate. Correct approach: upload to S3/Cloudflare R2, transcode via AWS Elemental MediaConvert or Mux, HLS playlist for adaptive streaming via Video.js or Plyr.
Viewing progress — periodic sending of watch_position from the frontend (every 10–30 seconds), storage in Redis with periodic synchronization to PostgreSQL. Do not save every second to the database — it will kill performance.
SCORM compatibility — if integration with corporate training materials is needed. Separate module, there are ready libraries (scorm-again).
Intranet and HR Portals
Corporate intranet: news, documents, organizational structure, HR processes (vacations, requests, KPIs).
Organizational structure in the database is a hierarchical structure. Adjacency list (parent_id on each record) is simple to implement but slow for recursive queries. Nested Sets or Closure Table are faster for reading hierarchies, more complex for changes. In PostgreSQL — recursive CTEs (WITH RECURSIVE) with adjacency list — a balance between simplicity and performance.
Document and request approval — workflow engine. Simple linear approvals (employee → manager → HR → accountant) can be done without a special engine. Non-linear (parallel branches, conditional transitions, delegation) — consider ready solutions: Temporal.io for workflow orchestration or a custom state machine based on the state-machine pattern.
What is included in the work
When ordering a corporate portal development, you receive:
- Architectural documentation (ER diagrams, integration scheme, role matrix)
- Full code in a Git repository with CI/CD
- Access to infrastructure (hosting, databases, storage)
- Training for administrators and key users (2–3 sessions)
- Warranty support for 3 months after launch
Our design principles rely on official Laravel documentation on authorization (Policies) and recommendations for working with queues.
Technical Stack for Portals
| Layer |
Tools |
| Backend |
Laravel + PostgreSQL |
| Frontend |
React + TypeScript (Inertia.js or separate SPA) |
| Real-time |
Laravel Echo + Soketi / Pusher |
| Search |
Meilisearch (quick start) or Elasticsearch (volume) |
| Queues |
Laravel Queue + Redis |
| Files |
S3-compatible (MinIO self-hosted or AWS S3) |
| Monitoring |
Sentry + Telescope (dev) |
Timeline Estimates
| Portal Type |
Timeline |
| CRM (basic) |
10–16 weeks |
| LMS (courses + video + tests) |
14–22 weeks |
| HR Portal (vacations, KPIs, org structure) |
12–20 weeks |
| Corporate ERP (modular) |
24–52 weeks |
The cost is calculated individually after a detailed analysis of requirements and role model. To get a preliminary estimate, contact us — we will analyze your task and offer an optimal turnkey solution.