Methodology

Scrape active listings

Each day at 6am, we scrape upcoming Melbourne auction listings from Domain.com.au. We capture the advertised price guide (from to range), auction date, agent, agency, and the URL to the Statement of Information PDF.

Parse the Statement of Information

We download and parse the PDF Statement of Information for each listing. The SOI contains the agent's indicative selling price (ISP) -- their formal estimate submitted to Consumer Affairs Victoria. We extract this figure as the benchmark.

Check the sold price

After auction day, we re-check Domain.com.au for the recorded sale price. This runs daily at 10am and 12pm. Once a sold price is confirmed, we lock the record and calculate the variance.

Calculate variance

Variance is calculated as:

variance % = ((sold_price - price_guide_midpoint) / price_guide_midpoint) x 100

We use the midpoint of the advertised price range as the guide. A positive variance means the property sold above the guide; negative means below.

Classify the result

We assign each variance to a category based on absolute distance from the midpoint:

Agent accuracy scores

We roll up variance data per agent and agency to produce accuracy scores -- average variance, percentage of auctions above/below guide, and trend over time. These scores update nightly.