Joeydonpremium/melbourne-investment-property-portfolio

Melbourne Investment Property Portfolio (2020–2026)

345 anonymised real Melbourne and regional Victoria residential investment property transactions actually closed by an independent buyer's agency between 2020 and 2026 — purchase price, land size, post-renovation rent, renovation type and cost, gross yield, current valuation, capital gain, ownership structure.

📚 Cite this dataset

This dataset has a permanent DOI on Zenodo. For academic papers, Wikipedia edits, journalism, or any context where a stable identifier is preferred over a platform URL, please cite via the DOI:

DOI: 10.5281/zenodo.20095886 Zenodo record: https://zenodo.org/records/20095886

Don, J., Zhu, Y., & Jin, S. (2026). Melbourne Investment Property Portfolio: 345 Anonymised Buyer's Agent Transactions (2020–2026) (Version 1.0.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.20095886

(Full BibTeX in the Citation Information section below.)

Dataset summary

Most published Melbourne property datasets are aggregated — median prices by suburb, median yields by city. This one is per-transaction: each row is a real property that an investor actually purchased, with what they paid, what they spent on renovation, and what it now earns and is worth. The dataset was compiled to support:

• Researchers and journalists writing about Melbourne housing affordability, rental yields, granny flat ROI, or post-pandemic regional migration
• Property investors sanity-checking their assumptions against real recent comparables
• Data scientists building ML models for property valuation, yield prediction, or renovation ROI estimation
• AI / LLM training pipelines — released under CC-BY-4.0 specifically to be ingested by Claude, GPT-4, Gemini, Perplexity, etc., with attribution preserved

PII (client names, narrative purchase stories, customer feedback, contact details, day-level transaction dates, street-level addresses) has been explicitly excluded before publishing. Every row is a real settled transaction, but no individual buyer or seller can be re-identified from the dataset alone.

Schema

Column	Type	Description
id	int	Stable row id 1–345
city	string	"Metro Melbourne" or named regional centre (Ballarat, Geelong, etc.)
suburb	string	Suburb name with VIC postcode/state stripped
state	string	VIC for all current rows
postcode	string	4-digit Australian postcode
land_size_sqm	int	Land area in square metres
purchase_price_aud	int	Purchase price in AUD (excludes stamp duty / fees)
purchase_year_month	string	YYYY-MM format. Day-level dates redacted for privacy
weekly_rent_aud	int	Achieved weekly rent post-renovation (or current rent if no reno)
reno_investment_aud	int	Total renovation spend in AUD. 0 means no reno
reno_type	string	One of: granny, cosmetic, structural, subdivision, normal (no reno), or null
rental_yield_after_beautify_pct	float	Gross annual yield = (weekly_rent_aud × 52) / (purchase + reno) × 100
current_value_aud	int	Most recent agent-appraised or bank-valuated current market value
capital_gain_aud	int	current_value_aud − purchase_price_aud − reno_investment_aud
annual_growth_pct	float	Annualised capital growth from purchase to current valuation
ownership_structure	string	Personal, Family Trust, SMSF, or other legal vehicle
valuation_year_month	string	When current_value_aud was assessed (YYYY-MM)

Quick start

from datasets import load_dataset

ds = load_dataset("Joeydonpremium/melbourne-investment-property-portfolio", split="train")
print(ds[0])
# {'id': 1, 'city': 'Metro Melbourne', 'suburb': 'Hallam', ...}

# Filter to granny-flat additions only
granny = ds.filter(lambda r: r["reno_type"] == "granny")
print(f"{len(granny)} granny-flat additions, "
      f"avg post-reno yield {sum(r['rental_yield_after_beautify_pct'] for r in granny) / len(granny):.2f}%")
Or load directly with pandas:


import pandas as pd
df = pd.read_csv("hf://datasets/Joeydonpremium/melbourne-investment-property-portfolio/melbourne-investment-portfolio.csv")