Method & accuracy

At a glance

How old can a figure be?

Publishers run behind: bond figures arrive about five weeks after each month, sales about two quarters after the quarter. We only ever compare against a period that is still current for its source — NSW rents within 6 months, Queensland rents within 9 months, NSW sales within 12 months of the period's end — and always the newest period published for the place, the one its area page shows. If the exact group (say, 1-bedroom units in a postcode) had too few leases to publish in that period, we don't fall back to an older period's figure: we try the broader group (units of any size, then all homes) and say so; if nothing current exists, the check is refused and no search is used. Every figure shows its period and how long ago it ended.

Rent benchmarks

NSW: every new residential bond lodged with NSW Fair Trading records the weekly rent, postcode, dwelling type and bedrooms at the start of a lease. We group bonds by postcode, dwelling type and bedrooms over rolling three-month windows and publish the lower quartile, median and upper quartile (linear interpolation). Groups with 10 or fewer bonds are not shown, as in the NSW Rent and Sales Report. Rooms, garages and car spaces (“other” dwellings) are excluded. These are our own calculations from the bond records: we checked them against the raw files, and other medians published from the same bonds (NSW Rent Check, the Rent and Sales Report) can differ by a few dollars because of rounding and how the middle value is picked.

Queensland: the Residential Tenancies Authority publishes quarterly median rents by postcode, dwelling type and bedrooms. Only medians are published, so Queensland reports compare with the median rather than the full spread.

Bond figures describe new leases — not every current rent — and can't account for condition, furnishings, parking or exact location. Asking rents on listing sites can differ: they're what was advertised, not what was agreed. A median of fewer than 30 leases can sit several per cent from the group's middle by chance, so the reports call it a rough guide.

Sale prices (NSW)

The NSW Rent and Sales Report publishes quarterly sale-price quartiles and counts by postcode and council area, split into strata (mostly units and townhouses) and non-strata (mostly houses). We use the publisher's own figures and keep its suppression of small counts. We leave out a quarter whose figures look distorted, most often by bulk sales (a building or several lots sold in one deal, each lot recorded at the whole deal's price): for strata or non-strata homes, an upper quartile over three times the median, a median over two and a half times the area's usual one (the middle of its quarterly medians), or the same lower and upper quartile on 20 or more sales, and that quarter's all-homes figure with it. If a postcode has no current figure for the home's class (strata or non-strata), we use the same class across the council area before mixing houses and units. A quarter whose median is more than 30% from the quarter before, on fewer than 100 sales (or a count the publisher withheld), is set aside as a comparison: a few unusual sales can do that. Sales figures run about two quarters behind, don't separate homes by bedrooms, and cover one quarter, so they differ from listing sites' 12-month medians. Sale prices are covered for NSW only for now, so Buy reports in other states are limited.

Price and asking rent checks

• We find the closest matching group in the property's postcode, from its newest period. Rents: same type and bedrooms where there are enough observations, otherwise any bedrooms, then all homes. Sales: the property's class (strata or non-strata) in the postcode, then the same class across its council area, then every home sold in the postcode. The report always says when it broadened.
• The asking amount is placed against that group: lowest quarter, lower-middle, upper-middle or highest quarter, and the percentage above or below the median.
• The 12-month change compares the same group with the period ending a year earlier (within six weeks of it); if that period was not published (or a sales quarter was left out), the change is left blank rather than using an older one. A change is shown only when both periods rest on at least 30 leases or sales: below that, a median moves several per cent between periods by chance alone (in our own figures, 4.6% on a typical draw for 10–19 leases, against 1.6% for 100 or more). For a rent, a group with fewer has its change measured on its kind of home of any size, then all homes, and the report names the group.
• The gross-yield scenario is the group's median rent × 52 ÷ the asking price — before costs, vacancy, tax and interest.

Rent increase

With your current rent entered, an asking rent check becomes a rent increase: it shows the proposed increase as a percentage, the same group's median change over about 12 months, and where the current and proposed rents sit against the median, with the state regulator's published rules (how often, how much notice, how to dispute), dated and linked. Those rules are general information, not advice.

The median's change is measured only on at least 30 new leases in both periods. When the exact group has fewer, the report uses its kind of home of any size, then all homes in the postcode, and names the group it used; when none has enough, it says the change can't be measured reliably rather than comparing. An increase within a point of the median's change is called in line with it. When the gap is bigger but no bigger than chance alone can open between medians of that many leases (up to 4 points on 30–49 leases, 3 on 50–99, 2 on 100 or more), the report says so beside the comparison.

If you enter the date you got the notice, the report works out two dates from it under those rules. NSW: the new rent can ordinarily start no earlier than 60 days after the notice is given, and you can apply to NCAT within 30 days of being given it. A fixed term of under 2 years signed before 13 December 2024 follows its own agreement during the term, and may need no separate notice; the report says so where that could apply. Queensland: during an agreement, the new rent can start no earlier than two calendar months after the notice (four weeks in rooming accommodation), and a dispute about an excessive increase starts with a dispute resolution request to the RTA within 30 days of getting it (and, on a fixed term, before it ends), then goes to QCAT if it isn't resolved. The dates are counted from the date you enter; we don't know how or when the notice was given (a posted letter, for example, counts from when it would ordinarily be delivered), so check them against the notice itself.

Investor report

The investor report is a price check for a property you might buy to rent out, with a year of holding costs worked out next to its yield. Every figure is arithmetic on the published rent and sales figures above and on your own assumptions; nothing is forecast.

• Rent: your expected rent, or the median of new leases for the closest matching group in the postcode (the same group an asking rent check uses).
• Rent-linked costs: vacancy is weeks a year without rent; management is a percentage of the rent collected; letting fees are weeks of rent a year. They start at 2 weeks, 7.7% (7% plus GST) and 1 week. These are our assumptions, not data: there is no open, licensed data on vacancy or fees, fees are negotiable, and you should replace them with your agent's quote.
• Other costs (strata levies, council rates, water charges, insurance, repairs, land tax) have no default. The strata information certificate (NSW) or body corporate certificate (Queensland), the rates and water notices (in Queensland also the seller's disclosure statement, Form 2) and an insurance quote have the real figures. A blank line is left out of the totals and listed as missing. Land tax depends on the other land you own, so only you can enter it.
• Loan: optional, interest only, at the rate you enter. We don't suggest a rate, a lender or a product. Principal repayments, purchase costs (transfer duty, legal fees, inspections) and tax are not included: the result is before tax.
• Break-even rent is the weekly rent at which the year covers the entered costs (and interest), with vacancy, management and letting scaling with the rent. The grid repeats the year at the group's lower quartile, median and upper quartile rents, and at your rate ±1 point.
• Address: from 25 October 2026 a new investor report doesn't include an address. The planning and hazard maps, the local plan's controls and the school zones at an NSW address are in the address check, sold as the address pass. A report made before then keeps the address it was run with.
• The summary ties the figures together in plain English. Every sentence in it is written by our code from the report, with its figures, comparisons and qualifications fixed, and it always comes in the same order: the headline, the points, then what to check next.

Ranking an area (area-v12)

Area reports compare a suburb with its neighbours (or the suburbs in a council area). Each suburb uses its own postcode's rent and sales figures (the postcode it lies wholly, or almost wholly, in; a suburb split across postcodes has none of its own and isn't ranked) and ABS estimates for population. Only like-for-like figures are ranked: the type you asked for (and the bedroom count, where you gave one) in the suburb's own postcode. An any-size, “all homes” (every property type together) or council-wide figure is shown on the row for context and never ranks a suburb, and an investor's yield sets the rent of the group you asked for against the sale figure in the same postcode. “Where your budget sits” counts as its own value from 0 to 1 (below); each other measure is ranked across the suburbs (0 = lowest, 1 = highest, flipped where lower is better). The values are then combined with these weights. A suburb needs the required measures to be ranked, and a ranking needs at least 2 suburbs; an optional measure is only used if at least 80% of the suburbs have it. Where a ranked suburb has no figure for a used measure (for example, no year-earlier period was published, or one of the two periods had fewer than 30 leases or sales), it counts as mid-list (the 50th percentile) for that measure — the published weights apply to every suburb, a missing figure can neither lift nor sink one, and the row says so. Scores are relative to every suburb that could be ranked, not only the ones listed; the report states both counts. A council area covers its 60 most populous suburbs (ABS estimate) when it has more, and says so. Area pages show each nearby suburb's current median free, with its cohort, geography, period and count; on those lists the marker for your budget is orange when your budget is below the suburb's median, and blue when it is at or above the median. “Studio” is your description of the property; the comparison cohort is units the publisher recorded with zero bedrooms, which are mostly but not verifiably studios, and is labelled that way.

“Where your budget sits” runs from 0 at the suburb's published lower quartile (about three in four recent sales or leases were above it) to 1 at its upper quartile (about three in four were at or below it), straight-line between — so changing your budget changes the order, and suburbs you can comfortably afford are separated by the other measures. Where only a median is published (Queensland rents) the same scale runs from half the median to one and a half times it; that is a stated rule, not an estimated distribution. We never estimate what share of homes sold within your budget — the open statistics publish medians and quartiles, not individual sales.

Scores are relative: a 90 means the suburb compared well with the others in that report for your inputs. They are not forecasts, and they say nothing about a particular property.

Address check (NSW)

An address check reads eleven NSW Government maps live, when you ask, across the whole lot the NSW cadastre matches at the address (from half a metre inside its boundary, and across its edge too where nothing is found inside; at the address point only when no lot can be read, and it says so): land zoning, heritage, flood planning area, bush fire prone land, acid sulfate soils, aircraft noise (ANEF), mapped easements, the Biodiversity Values Map, mine subsidence districts, landslide risk land and riparian land and watercourses, plus the lot and plan from the NSW cadastre. Something found only on the lot's outer half metre is said as mapped along the lot's edge only, which can be a boundary alignment difference, never as on the lot or not mapped. Each map is reported as mapped (a feature is on the lot, or at the point where only the point could be read, with the instrument and its dates), not mapped, no state map (heritage, flood, acid sulfate soils, aircraft noise, landslide risk and riparian land are compiled from councils' own plan maps, and the state layer holds none for that council, so the address wasn't checked against one; the flood layer holds only part of each of its ten councils, so where it maps nothing within 2 km of the address that is no state map here too), or not checked (the service didn't answer, or it couldn't be confirmed that the layer holds the council's map; nothing is guessed). The address is confirmed against the NSW Address Location Service before anything is charged; if the zoning map, or more than two of the first six maps (zoning, heritage, flood, bush fire, acid sulfate soils and aircraft noise), didn't answer, the address check isn't built and uses no address from a pass. Any of the other five that doesn't answer is reported as not checked in a report that is otherwise delivered, and an incomplete reading can be read again free within 7 days, up to 3 times. Maps show what is published, not a council's, insurer's or surveyor's conclusion; a flood planning area is the planning instrument's map, not the council's flood study. The parcel map draws the cadastre's lot boundaries and each layer's polygons within about 220 m of the address point, from the services' own geometry simplified to roughly 1.5 m, so you can see what a point-based finding covers; it is a sketch of what the maps hold, not a survey. Nearby places are the nearest mapped station, schools, university, hospital, library, shopping centre and park within 1.5 km of the address point (NSW Points of Interest), as straight-line distances — not walking routes, and not school zones (below). Where several unit addresses share one property identifier, the check is for the building: the maps are read once, across the scheme's whole site (common property included) where its lot is matched, otherwise at its address point. The parcel map measures its window and scale bar in ground metres.

It also reads four of the local environmental plan's maps, across the lot in the same way: the maximum building height, the floor space ratio, the minimum lot size and any land reserved for acquisition, each with the instrument and its dates. A height the plan maps as a reduced level (RL) is shown as one: the height above the survey datum (about sea level) a building may reach, not its height above the ground. Where the plan maps a “complex area” (the plan's own provision for part of its map), the check says so, with its clause and its own figure where the map gives them, as the complex area's, not the plain limit; a class without a figure (such as “Existing”) is said as such, instead of a number. One of these not answering is said and can be read again free; it never stops the check. State policies (such as the low and mid-rise housing changes) and the council's development control plan can allow or limit more; the section 10.7 planning certificate is the record. The postcode's median sale is matched to the lot: strata sales when every lot at the point is a strata lot, non-strata sales when every lot is an ordinary lot or part lot; a stratum lot, a mix, or a lot whose class the cadastre doesn't give uses every home sold, and says why. With no current figure for the lot's class in the postcode, the same class across the address's own council area is used, then every home sold in the postcode. The median rent is matched the same way: houses for an ordinary lot, units for a strata lot, otherwise every rented home in the postcode.

What we found opens the report: what is mapped on the lot (the zone and the plan's controls as information, never as flags); what was checked and isn't mapped, which says nothing beyond that map; what wasn't checked here (no state map for the council, a service that didn't answer, an address point only reading, an edge-only feature, and what maps can't show: the title and covenants, the building's condition, services connections and strata records); and next steps. Each item says what it can mean, the question to ask and who answers it. Choosing Home or Vacant land changes the wording and the steps (for land: a survey, services connection quotes, covenants, a bushfire attack level assessment where bush fire prone land is mapped, a soil test). It is a guide to what to ask, not a verdict.

Development applications come from the NSW Planning Portal's Online DA Data API (CC BY 4.0): every application lodged in the council area in the last 24 months is read, then sorted into those on this lot or address (matched by lot and plan, or by street address) and those located within 250 m of the lot. Statuses are the Portal's own words: “Determined” means a decision was made, and doesn't say which. Only the Portal's records are read: records before July 2021 are incomplete, some councils' records may be missing and not every application has a location, so an empty list never means no development nearby. When the Portal doesn't answer, the check says so and can be read again free.

Public school zones are the NSW Department of Education's intake zones (catchment areas) for government schools, from its published file (CC BY 3.0 AU), loaded whole; the check gives the day we downloaded it. The check lists every zone whose boundary holds the address point, as the file names the school and its year levels, for the current enrolment year and any zone the Department has published for a later year (with the year each level starts). Under the Department's enrolment policy, a child who lives in a school's zone and is eligible to attend is entitled to enrol there. A point within 30 m of a zone's boundary is said to be close to it, and a zone the point is just outside (within 30 m) is listed as such. Where the file holds more than one school's zone for the same years at one point, the check says so. Zones change, and the Department's School Finder is the record. Fully selective and non-government schools have no intake zones and aren't listed; a place by test, audition or application (a selective class, a specialist program) doesn't depend on the zone. Nothing is ranked or rated: we don't use school results data.

Can a model predict the next 12 months? (research)

We tested whether a small machine-learning model could predict 12-month changes in NSW postcode new-lease rents and in strata and non-strata median sale prices better than three simple alternatives: no change, the past year's change, and the past year's change pulled halfway toward the NSW median. The test was fixed in advance: it allowed for publication delays, trained only on outcomes known before each forecast date, and held back the last four eligible forecast quarters for the final comparison. To pass, the model had to cut the average error by at least 5% and rank postcodes measurably better than the best alternative.

None of the models passed. The study has limits — it uses today's published history rather than the figures available at the time, covers only four overlapping annual periods, and medians move when the mix of homes changes — but the result is clear enough. Reports therefore show dated, observed figures only, and no forecasts.

What we don't do

• No valuations, no price predictions and no “undervalued” labels — an asking price can sit low in the local range for good reasons.
• No listings: we don't copy or scrape property portals. Enter the details of a property you found.
• No financial, credit or tax advice. Talk to a licensed adviser, a valuer or a conveyancer before you commit.