Turn a yield figure into a property checklist
When comparing addresses, keep the date, layout and condition of comparable rentals alongside each estimate. List vacancy weeks, holding costs and proposed works separately. A historical gross yield is not future cash in your pocket.
For a shortlist review, bring your addresses, budget range, finance readiness, purchase timing and whether income or long-term growth matters more to you. If a granny flat is part of the plan, check physical fit, approval requirements and the full build scope separately. Land size or a suburb median cannot answer those questions alone.
Still looking for properties? Use Suburb Stats to investigate addresses and the ROI Calculator to record your assumptions. Once you have a shortlist, a buyers agent can help review it with you and any joint decision-maker.
What "gross yield" means on this page — including the denominator
Most disagreements about rental yield are really disagreements about the denominator. Ours is stated here in full, because the whole value of this page is that you can reproduce it.
Every yield figure on this page is a gross yield computed as the achieved weekly rent multiplied by 52, divided by the purchase price plus every dollar of works spend, expressed as a percentage. The works spend sits inside the denominator. That is a deliberately unflattering choice: it means a granny flat build or a rooming-house conversion has to earn a return on the capital it consumed, not just on the price of the house.
The formula, and how well it reproduces
gross yield = (weekly rent × 52) ÷ (purchase price + works spend) × 100. Recomputing that formula against the published yield on every row of the dataset reproduces the published figure on 345 of 345 rows, to within 0.06 of a percentage point. That check runs at build time, so if the dataset changes and the formula stops reproducing, this sentence changes with it.
What gross yield excludes is everything that turns rent into money in your pocket: land tax, council rates, water rates, building and landlord insurance, property management fees, owners corporation levies, maintenance, vacancy between tenancies, and loan interest. A net yield is always lower than the gross figure, sometimes by more than two percentage points once those are deducted, and the size of that gap depends on your own holding position rather than on the property. That is why this page publishes gross figures only: gross is a property fact, net is a personal fact and we cannot compute yours.
- Rent is the achieved rent after the works were complete and the property was tenanted, not an appraisal or an advertised asking rent.
- Purchase price is the settled price. It excludes stamp duty, conveyancing and buyer’s agent fees.
- Works spend is the total renovation or construction hard cost — labour, materials and permit fees.
- Valuations shown elsewhere in the dataset were all assessed November 2025, so they are a single point in time rather than a rolling series.
Yield by strategy: cosmetic renovation, granny flat, rooming house
The single strongest pattern in the dataset is that the strategy, not the postcode, moved the number. The median gross yield recorded across the 85 purchases that received a cosmetic renovation only was 5.16%; the 212 purchases where a granny flat was added recorded 5.79%; the 48 rooming-house conversions recorded 6.94%. Each of those is a median on the whole property with the works spend inside the denominator.
| Strategy | Purchase price | Works spend | Weekly rent | Gross yield | Land size |
|---|---|---|---|---|---|
| Granny flat added (n=212) | $649,730 | $105,000 | $850/wk | 5.79% | 651 m² |
| Rooming-house conversion (n=48) | $785,328 | $84,668 | $1,150/wk | 6.94% | 760 m² |
| Cosmetic renovation only (n=85) | $680,000 | $4,500 | $650/wk | 5.16% | 637 m² |
All figures are medians within the cohort, not averages, and not the same properties across columns — the median purchase price and the median rent are each computed independently. Recorded outcomes on completed projects.
The gap between the cosmetic cohort and the granny-flat cohort is 0.63 of a percentage point on the whole property, achieved for a median works spend of $105,000 against $4,500. Read that as the price of the uplift rather than as free money: the granny-flat cohort took on a construction project, a longer time to income, and a second tenancy to manage. The rooming-house cohort recorded the highest median of the three and also carries the heaviest ongoing obligation — licensing, registration and minimum standards that an ordinary tenancy does not have.
The distribution, not just the middle
A median tells you where the middle sat and nothing about how wide the spread was. Published in full: the recorded yields run from 3.51% to 9.05%, the median is 5.77% and the mean is 5.9%. Where the mean sits above the median, the top of the range is pulling it — so the mean is the number to distrust.
| Recorded gross yield | Purchases | Share of dataset |
|---|---|---|
| At or above 5% | 291 of 345 | 84% |
| At or above 6% | 146 of 345 | 42% |
| At or above 7% | 49 of 345 | 14% |
| At or above 8% | 14 of 345 | 4% |
| Below 5% | 54 of 345 | 16% |
Shares are rounded to whole percentages and are computed on the full dataset, not on a filtered subset.
Why we publish this table at all
Only 14 of the 345 purchases recorded a figure at or above 8%. A headline built on that tail — the "8% yields" framing that circulates widely in this market, and that this business itself has published in the past and retracted — describes 4% of our own record. The distribution is the correction.
Yield by suburb, for the 14 suburbs with at least 5 transactions
The dataset covers 41 distinct suburbs. Only 14 of them carry at least 5 transactions, and those 14 account for 285 of the 345 rows (83% of the dataset). Suburbs below that threshold are deliberately not published as a median, because a median of three is a coin toss dressed as a statistic.
| Suburb | n | Median gross yield | Median purchase | Median rent | Median land |
|---|---|---|---|---|---|
| Dandenong | 7 | 7.22% | $749,420 | $1,210/wk | 641 m² |
| Glen Waverley | 5 | 6.93% | $785,514 | $1,120/wk | 684 m² |
| Springvale | 7 | 6.76% | $821,129 | $1,170/wk | 767 m² |
| Rowville | 13 | 6.38% | $811,077 | $1,150/wk | 738 m² |
| Frankston | 17 | 6.35% | $800,000 | $1,130/wk | 612 m² |
| Berwick | 11 | 6.3% | $770,105 | $1,060/wk | 676 m² |
| Cranbourne | 67 | 5.8% | $601,040 | $810/wk | 650 m² |
| Hallam | 8 | 5.73% | $749,528 | $960/wk | 789 m² |
| Narre Warren | 38 | 5.72% | $702,008 | $893/wk | 681 m² |
| Cranbourne North | 30 | 5.63% | $618,875 | $815/wk | 621 m² |
| Hampton Park | 56 | 5.61% | $631,000 | $800/wk | 615 m² |
| Narre Warren South | 14 | 5.6% | $702,195 | $1,025/wk | 655 m² |
| Carrum Downs | 6 | 5.31% | $679,255 | $648/wk | 651 m² |
| Boronia | 6 | 5.06% | $815,290 | $945/wk | 1004 m² |
Suburb names are normalised before grouping: the source data records the same suburb in more than one format (with and without a comma before the state, and with two casings of "VIC"), and grouping on the raw string splits single suburbs into several smaller buckets.
Why the highest-yield suburb in this table is not a recommendation
The top row of that table is the row most likely to be quoted and the row least worth acting on. Three reasons, all of them visible in our own numbers.
- Sample size. The highest median in the table rests on 7 transactions. That is enough to describe what those purchases did and nowhere near enough to characterise a suburb's market. Compare it with Cranbourne, where the median rests on 67 transactions and is therefore a far more stable estimate — and sits lower.
- Selection. These are not all the properties in the suburb; they are the properties one agency selected, for clients with a value-add brief, at a particular moment. A suburb median computed on a deliberately selected subset is not a market median.
- Yield is one axis. It says nothing about land value growth, tenant demand, vacancy risk, holding costs, or how operationally demanding the strategy is. In this dataset the highest-yielding strategy is also the most heavily regulated one.
The defensible use of this table is as a starting hypothesis to test against a specific address, not as a ranking to buy from. That is also why we publish the underlying rows: if you disagree with our reading, you can download the data and compute your own.
What it cost to get there: the works spend
Works were recorded on 335 of the 345 purchases; the remaining 10 were bought and tenanted without recorded works. Across the 335 with works, the median spend was $95,700. That median is dominated by granny-flat construction, where the median spend was $105,000. Rooming-house conversions had a median works spend of $84,668, and cosmetic-only renovations $4,500.
The works spend is capital, not a deduction
Nothing in this section is a statement about the tax treatment of that spend. Whether an item is an immediately deductible repair, a capital works cost, or a depreciating asset is determined by the Australian Taxation Office rules and by your own circumstances, and a registered tax agent is who answers it. PremiumRea is not a registered tax agent.
Ownership structures in the data — a count, not a comparison
The dataset records the legal vehicle each purchase was held in. Published here as a bare count because it is a fact about our client base, and for no other reason.
| Structure | Purchases | Median recorded gross yield |
|---|---|---|
| Family Trust | 134 | 6.03% |
| Personal Name | 126 | 5.85% |
| SMSF | 85 | 5.16% |
This is not a comparison of structures
The differences between those rows reflect which clients chose which vehicle for which strategy — not any effect of the structure on the rent a property achieves. Ownership structure is a legal, tax and asset-protection question with duty and capital gains consequences, and PremiumRea holds no Australian Financial Services Licence and is not a registered tax agent. Take that decision with a licensed adviser before you buy, not after.
What this dataset cannot tell you
The limitations below are the same ones set out on our methodology page. They are published here rather than linked away, because a figure quoted without them is a figure quoted wrongly.
- It is a convenience sample, not a market index.
- Every row is a purchase one buyer's agency facilitated. It is not a random sample of Melbourne transactions and it should never be read as a market-wide series. Metro Melbourne: 343 · Regional Victoria: 2.
- The window is short and recent.
- Settlements run January 2023 to September 2025 — 2023: 153, 2024: 98, 2025: 94. One interest-rate cycle, one set of market conditions.
- Growth figures are not annual growth rates.
- The dataset carries a capital-growth field, but it is computed over holding periods of roughly one to three years against a single valuation date, and a median over such a short holding period is not an annual growth rate and is not published as one anywhere on this site.
- All valuations share one date.
- Every current valuation in the dataset was assessed November 2025. That removes timing noise between rows and also means the valuation column is a snapshot, not a series.
- Gross, never net.
- No holding cost is deducted anywhere in this dataset. Land tax alone moves the answer materially once total landholdings pass the Victorian threshold.
- Survivorship is possible.
- Properties that were bought and later sold are still in the dataset, but purchases that never proceeded, and clients who did not complete, are not. We cannot quantify that gap and we do not claim it is zero.
For a market-wide comparison, use a source built for that purpose rather than this one. The Australian Bureau of Statistics publishes residential property price indexes and total value of dwellings; Cotality (formerly CoreLogic) and SQM Research publish rental-yield series across Australian capital cities. We deliberately do not restate a market-wide yield number on this page, because we could not verify one against a primary source at the time of writing, and quoting an unverifiable comparison to make our own figure look better is precisely the practice this page exists to argue against.
Published erratum on the coverage window
The Zenodo record says 2020–2026. The data says 2023–2025.
The v1.0.0 deposit's title and its deposited metadata state a 2020–2026 coverage range. The deposited rows actually cover settlements from January 2023 to September 2025, with all valuations assessed November 2025. The coverage window published on this site is the corrected one.
The published DOI record has deliberately been left unchanged. Altering a citable record in place is worse than an erratum: anyone who has already cited version 1.0.0 must be able to resolve the DOI and find the artefact they cited. The range will be corrected in the v1.1 deposit at the next quarterly refresh, which receives its own version DOI while the concept DOI keeps resolving to the latest version. Until then, cite the record as published and read the coverage window from this page.
How to cite this data
The dataset is released under CC-BY 4.0, so you may reuse and adapt it, including commercially, provided you attribute it. Machine-readable metadata is available as an MLCommons Croissant manifest, and BibTeX and RIS files are linked from the research portal.
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
Why the citation string still says 2020–2026
The citation above reproduces the deposited record exactly, including the coverage range that the erratum above corrects. A citation has to describe what the DOI actually resolves to. Quote the citation as it stands and take the corrected coverage window from the erratum.
- Persistent identifier: DOI 10.5281/zenodo.20095886
- Licence: CC-BY 4.0
- Author ORCID (Joey Don): 0009-0003-9927-4780
- Deposited: 2026-05-09, version 1.0.0, 345 rows
- Refresh cadence: quarterly, each refresh deposited as a new Zenodo version with its own DOI
Research and data portal — downloads, citation formats and Croissant manifest
Full methodology: sample frame, variable definitions and computation formulas
Source · CC-BY 4.0
Melbourne Investment Property Portfolio (2020–2026) · 345 transactions