Half of Waverley's Refused DAs Fail the Same Control. Just 3% of Approved Ones Do
We normalised every compliance record in Waverley's panel history. The most common reason for refusal and the most predictive one turns out to be different controls.
Our prior article, The DCP Compliance Patterns That Predict Approval, established that DCP flag count predicts refusal — the more controls an application trips, the worse it does. Useful, but blunt: it says non-compliance is bad without saying which non-compliance matters.
So we went back to the compliance records with a sharper question. Not which controls appear most often on refused applications, but which appear at rates approved applications never approach. Those are different questions, and they have different answers.
Frequency is not prediction
Start with the trap. Landscaping and biodiversity is flagged as non-compliant on 73 of 176 refused applications — 41.5%, one of the most common findings in the entire refusal record.
It is also flagged on 201 of 663 approved applications: 30.3%. The gap is 1.4×. Knowing an application has a landscaping problem tells you almost nothing about how it will be determined, because approved applications have them too, and they get conditioned.
Excavation behaves the same way — 23.3% of refusals against 14.8% of approvals. Front and rear building lines, 33.0% against 16.0%. Common, and close to useless as predictors.
Design Excellence, and the gap that matters
Design Excellence is flagged on 87 of 176 refused applications — 49.4%, close to half of every refusal in the panel's history. It is flagged on 20 of 663 approved applications: 3.0%. A 16× separation, the largest of any control at meaningful volume.
Ranked by that gap rather than raw frequency: general objectives, 27.3% of refusals against 0.5% of approvals (60×). Design Excellence, 49.4% against 3.0% (16×). Streetscape and visual impact, 34.7% against 4.2% (8.2×). Heritage, 13.1% against 2.1% (6.2×). Then the drop — solar access 3.0×, stormwater 2.2×, building lines 2.1×, excavation 1.6×, landscaping 1.4×.
The top of that list is qualitative. The bottom is technical.
Why the subjective controls decide
Applicants tend to assume the dangerous controls are the numerical ones — height, setback, floor space ratio, the things you either comply with or you don't.
The data says the opposite. Measurable controls are negotiable: a number can be conditioned, varied or traded. Design Excellence cannot. It is a judgment about whether the building is good enough, and once a panel has formed that view no clause fixes it. It survives to the determination while the technical objections get resolved along the way.
The measurement problem underneath this
None of the above is visible in the raw data. Waverley's records store control names as free text, and Design Excellence alone appears under 43 distinct labels — "B12 Design Excellence", "Part B12 – Design Excellence", "12. Design Excellence", plus a long tail carrying appended detail like "view corridors, bulk/massing, ESD, overshadowing".
Counted raw, the most predictive control in the dataset fragments into forty-three small ones and vanishes from the top of any frequency ranking. Analysis that doesn't consolidate label variants first doesn't get a slightly imprecise answer. It gets a wrong one.
Clause 4.6 is not the risk applicants think it is
Applications that invoke Clause 4.6 are refused at 6.8% — 30 across 439. Applications that don't are refused at 36.5% — 146 across 400. Formally departing from a development standard is associated with roughly one fifth the refusal rate of not departing from one.
The direction needs stating carefully. Invoking Clause 4.6 does not make an application safer. It is a document that requires a planning consultant to write, so its presence marks professional preparation rather than causing approval — the same selection effect our prior article found from the other side (see The Clause 4.6 Variation Success Rate: When Breaking the Rules Works).
The consultant data supports that reading. Refused applications carry an average of 2.4 consultant reports, approved 4.1, approved with modifications 4.4. The best-supported applications are the ones the panel will negotiate with.
Cost, by contrast, buys nothing. Refusal sits near 20% at every project size: 21.3% under $500,000, 19.5% between $1 million and $5 million, 20.0% above $10 million.
The trend, including the part that reverses it
Refusal rates by determination year have fallen a long way. In 2017 the panel refused 20 of 51 applications — just under 40%. By 2024 it was 8.6%, and 2025 came in at 7.9%.
The obvious reading is a maturing system: applicants learning what the panel expects, pre-lodgement filtering removing unviable proposals, and seven years of published decisions making the standard legible.
2026 complicates it. Seven of the 28 determinations so far this year have been refusals — 25.0%, back to 2021 levels. On 28 applications that is a partial year that could easily revert, and we would not call it a trend. But it is worth flagging, because the comfortable story is that refusal risk has permanently declined and this year does not yet support it.
One correction to a figure we have used before: refused applications are not markedly slower to determine. Median lodgement-to-determination is 148 days for refusals, 140 for approvals and 170 for approvals with modifications. Negotiation takes longer than rejection.
What this means for applicants
Of 171 addresses with a refused application in the panel record, 13 later secured an approval at the same address — 7.6%. Refusal is close to terminal.
The lesson is not to chase compliance with every control. It is to know which ones the panel treats as negotiable and which it does not. A landscaping issue is a condition. A design objection is a refusal.
Numerical compliance is table stakes. The judgment calls decide the outcome.
About UNDA Development
UNDA Development provides precedent intelligence for property developers and property planning consultants — every DA, every panel decision, every compliance record at your fingertips. Our sister product, UNDA Management, provides market intelligence for strata managers. Learn more at unda.management
