Basara

Real estate market data in the Waterloo Region is available, but it is not accessible. MLS boards publish aggregate numbers at the city level, usually months after the fact, and only to agents who know where to look. For consumers, the picture is even worse: vague headlines about "the market" with no way to see what is happening in their specific neighbourhood.

A local real estate agent wanted to change that. The goal was to build a platform that takes raw MLS data and turns it into neighbourhood-level intelligence that anyone can use. Not a generic listing site, but a purpose-built market data tool that establishes authority through depth and transparency.

Basara is a full-stack market intelligence platform that covers 7 cities and over 50 neighbourhoods in the Waterloo Region. Each neighbourhood page breaks down sold data into freehold and condo segments with average prices, median prices, days on market, sale-to-list ratios, and year-over-year changes.

The data pipeline starts with monthly MLS PDF exports. Custom scripts parse the sold records, normalize them against a neighbourhood taxonomy, and push the data into a relational database through protected API endpoints on a custom backend. Once ingested, AI generates plain-language narratives explaining what the numbers mean for each neighbourhood and market segment.

The frontend is built on a modern framework that uses incremental rebuilding to keep pages fresh without rebuilding the entire site. City pages revalidate daily, the homepage revalidates hourly, and article pages refresh every hour. The result is a fast, always-current site that costs almost nothing to operate.

The biggest challenge was the data pipeline. MLS boards export sold records as PDFs, not APIs. We built custom parsing scripts that extract structured data from these exports, map each record to the correct neighbourhood using board codes, and push normalized summaries into the database through authenticated ingest endpoints.

The backend uses a minimal API pattern with direct SQL queries instead of an ORM. Public endpoints serve market data and articles without authentication. Protected endpoints handle data ingestion and content approval, secured with bearer tokens. The database schema uses idempotent upserts so the pipeline can be re-run safely.

AI-generated narratives were a key differentiator. Raw numbers are useful but context matters. We integrated AI to generate plain-language explanations of what the data means for each neighbourhood. An admin approval workflow ensures nothing publishes without human review.

The SEO strategy was designed for compounding returns. Seven city pages, fifty-plus neighbourhood breakdowns, and a growing article feed create a crawlable surface area that builds Google trust over time. Structured data, dynamic sitemaps, and semantic HTML round out the technical SEO foundation.