Duplex build cost from 17 measured projects

Figures

Duplex — net construction cost for both dwellings, before builder's margin and GST

Deciles: p10 A$1,543,460, p90 A$3,641,960.

Per dwelling, total divided by two

This is arithmetic, not measurement. The two dwellings in a duplex are rarely identical, and none of these estimates was split by dwelling in the job record. Read the per-dwelling column as an average across the pair, not as the cost of either half.

By state

The VIC row is three projects. It is listed for completeness and is not a state benchmark.

Split between structure and finishes

Two of the seventeen projects carry no lock-up split in the job record and are excluded from that row only.

Cost per square metre

Basis: gross floor area as recorded in the delivered bill of quantities for each job, divided into the net construction total, before builder's margin and before GST, screened as set out below. Multiply by roughly 1.27 for a margin-inclusive, GST-inclusive rate.

Floor areas behind the duplex rate

The per-dwelling row is arithmetic, as in the totals table above. Nothing in the record splits the area between the two dwellings.

The duplex rate sits above the single-dwelling rate. The duplex median of A$4,667/m² is 16% higher than the A$4,013/m² median measured across 40 single dwellings on the same dataset over the same period. This is the clearest evidence in these tables that a duplex is not cheaper per square metre than a house. It is the same conclusion the totals reach, reached independently of floor area.

How the rate population was screened. Of the 106 completed jobs, 81 have a delivered bill of quantities carrying a floor area. Three screens were applied in order — a floor of 60 m² on the recorded area, an implied rate inside A$1,500 – A$10,000/m², and agreement within 10% between the rate stored in the pack and the rate computed from the construction total. Nineteen rows failed one of those and were removed, leaving 62 residential projects, 14 of them duplexes.

The median does not depend on which rate definition is used. Across the 75 rows carrying both a stored rate and a computable one, the two medians are A$4,045 and A$4,001, 1.1% apart. Fifty of the 75 agree to within half a percent.

Basis and screening

• Source: EstiFlow production job records, filtered to status COMPLETED with a construction total above zero, then to the duplex classification. That returns 17 projects.
• Completion dates: 30 April 2026 to 20 July 2026, except one project with no completion date recorded, discussed below.
• Delivery evidence: all 17 have deliverable files sealed against the job record with content hashes.
• Currency: Australian dollars, net of builder's margin and net of GST. See the basis note at the top of this article.
• No rows removed. The duplex population contains no duplicate sites and no scope-class mismatches. Two rows carry gaps that are disclosed rather than screened:
• One project at A$2,022,084 has a blank site label, no completion date, and three sealed files rather than the twelve typical of a full pack. It shipped and is marked customer-visible. It happens to sit at the median. Recomputing the median without it gives A$2,057,361, a difference of 1.7%, so its presence does not move the published figure materially. It is named here so the reader can see the median rests on a row whose site cannot be identified from the record.
• One project at A$2,983,375 has no lock-up split recorded and is excluded from the lock-up row only.

What moves the number

A duplex is not two houses. The median duplex total of A$2,022,084 is 2.14 times the median single-dwelling total of A$943,639 measured on the same dataset over the same period. That is a premium, not a discount. Duplexes in this population are concentrated in established metropolitan suburbs where a detached house on the same street would also cost more, and they carry work a detached house does not — party wall fire and acoustic separation, two of every service connection, dual driveway and crossover, and a subdivision-ready layout.

The party wall is a cost item, not a saving. Separating wall construction to the fire and acoustic performance required between attached dwellings, and the discontinuous construction that goes with it, is priced work in every project here. The intuition that a shared wall halves a wall's cost does not hold once the performance requirement is priced.

Services duplicate almost completely. Two electrical mains, two water and sewer connections, two gas or electric hot water systems, two switchboards, two NBN lead-ins, and where the design requires it two separate stormwater discharge paths. Authority connection charges follow the same doubling and are excluded from these totals in any case.

The top of the range is more than five times the bottom. Four projects sit above A$3.4M against a lowest of A$980,515. The job records do not carry slope, soil classification or excavation extent as queryable fields, so no attribution for that spread is offered here. What can be said is that duplex development in this population is concentrated on infill lots in established suburbs, where site constraint is the normal condition rather than the exception.

Site works, driveways and external works scale with the frontage, not the floor area. A wide corner lot duplex with two crossovers absorbs more external works cost than a narrow battle-axe arrangement of the same internal area.

The bottom of the range is a real project, not an error. The lowest total, A$980,515, is a regional NSW duplex. It sits in the same band as a mid-range single dwelling, which is the clearest evidence in this table that geography moves a duplex total at least as far as the dwelling count does.

What's excluded

These are construction cost totals. They do not include:

• Land acquisition, stamp duty and finance costs.
• Subdivision and Torrens or strata titling costs. A duplex is frequently built to be subdivided. Survey, plan of subdivision, lodgement, and the legal work to create separate titles are outside the construction total.
• Statutory and authority fees. These are enumerated as line items in the delivered packs and are never priced into a total. For a NSW project the set enumerated is the NSW Home Building Compensation Fund contribution where applicable, the Long Service Levy where applicable, and council contributions where conditioned. For VIC it is Victorian Building Authority and building permit context, plus local council contributions where conditioned. Section 7.11 and 7.12 contributions, water authority headworks charges and electricity connection charges all fall in this class. They are listed and excluded from the totals above because the amounts are set by the authority and by the conditions of consent rather than by the builder, and because they are commonly doubled on a two-dwelling development.
• GST and the builder's margin. Every figure above is before both. This is a change from an earlier version of this article, which described the totals as merely excluding GST and was incomplete. Multiply by roughly 1.27 for a margin-inclusive, GST-inclusive figure.
• Professional fees. Architectural, engineering, certification, surveying and town planning fees are outside the construction total.
• Demolition of an existing dwelling, unless it was shown in the documents supplied. Most duplex sites carry an existing house. Where the drawings supplied included demolition it is priced, where they did not it is absent.
• Loose furniture, appliances not fixed, and landscaping beyond what the approved drawings show.

About this dataset

EstiFlow is not a registered quantity surveying practice, and the documents behind these figures are not certified quantity surveying reports. They are cost estimates, bills of quantities and estimating packs produced from the architectural drawings and supporting documents supplied for each project. Each figure traces to a job record with sealed, content-hashed deliverable files.

Seventeen projects is a small population. The median is stable against the one weakly-evidenced row, but the quartiles rest on four observations each. Treat the range as the useful output and the median as an indicator rather than a settled benchmark.

Figures dated 2026-09-01.