Plan Advice
← Back to Plan Advice

Our methodology

Why our assessments are accurate.

Most planning tools tell you what the rules say. Plan Advice tells you how your council actually applies them. Here is the logic — and the data — behind every assessment we produce.

The foundation

Four things most alternatives don’t have

A corpus of real council decisions

We have assembled and indexed thousands of historical development application determinations from councils across NSW — including the assessment reasoning behind approvals and refusals. When we assess your proposal, we look at how your council has actually decided comparable applications in your zone, not just what the legislation permits in theory.

A deterministic compliance engine

Core numeric controls — building height, floor space ratio, minimum lot size — are checked mathematically against the actual standards that apply to your land under your council’s Local Environmental Plan. These checks are exact pass/fail calculations, not AI opinion, so the foundation of every assessment is verifiable.

Live, authoritative government data

Zoning, lot boundaries, heritage listings, and planning overlays are pulled in real time from official NSW Government spatial services — the same datasets councils rely on. Nothing is cached from an out-of-date snapshot, and the statutory instruments in our library (LEPs, SEPPs, DCPs) are kept current.

Precedent-matched AI reasoning

Our AI doesn’t answer from general knowledge. For every assessment it retrieves the most comparable local decisions and the exact planning provisions that apply, then reasons over that evidence — citing its sources. Where it flags a risk, it can point to the decision or control that justifies it.

Under the hood

How every assessment is assembled

Five stages run every time you request an analysis. The exact weighting and retrieval logic is our secret sauce — but the shape of the pipeline is no mystery.

  1. 1

    Resolve your property

    Your address is matched to its lot, zone, and the exact planning instruments that govern it.

  2. 2

    Assemble the evidence

    Current controls, overlays and constraints are collected from live government sources, alongside comparable decisions from your council.

  3. 3

    Run the numbers

    Height, FSR and lot-size compliance are computed deterministically against your LEP’s development standards.

  4. 4

    Reason over precedent

    The AI weighs your proposal against how similar applications were actually determined — surfacing likely objections and conditions.

  5. 5

    Produce a cited report

    Findings are compiled into a plain-English report where material claims are traceable to a source.

The honest comparison

How we stack up against the alternatives

Generic AI chatbots

  • Answer from general training data — no local decision history
  • Prone to inventing clause numbers and controls
  • No numeric compliance checking

DIY council research

  • Days lost in LEP clauses, DCP chapters and mapping portals
  • Easy to miss an overlay or misread a development standard
  • Free — but expensive in time and risk

Planning consultants

  • Expert judgement and formal advice
  • Typically $1,500+ and days-to-weeks of turnaround
  • Rarely backed by systematic analysis of past determinations

Plan Advice

  • Grounded in real decisions from your council
  • Deterministic checks on the numbers that matter
  • Minutes, not weeks — with cited, plain-English findings

Where we draw the line: Plan Advice is preliminary planning guidance, not formal planning or legal advice, and no tool can bind a consent authority. What we can do is make sure that when you engage council — or a consultant — you arrive better informed than almost anyone else in the room.

See it work on your property.

The free analysis takes minutes — no signup, no card.

Under the hood

How every Plan Advice report is built

Every paid Plan Advice report is grounded in three independent evidence layers so the answer you receive is verifiable, cite-able and reproducible. This page explains each layer, how they combine, and where an AI language model contributes vs where deterministic code produces the number.

Layer 1 — The planning-instrument grounding

When you enter an address we resolve it to a council, then load every planning instrument that applies to that lot: the Local Environmental Plan (LEP), the Development Control Plan (DCP), the applicable State Environmental Planning Policies (SEPPs), and the NSW Standard Instrument Order zones. Instruments are held as structured objects (permitted-with-consent lists, prohibited-uses lists, Schedule 1 additional-uses lists, cross-zone references, definitions) not raw PDF text — so citations to a specific clause are addressable and Google-search-verifiable.

Layer 2 — Deterministic compliance checks

Numeric development standards (maximum height, floor space ratio, minimum setbacks, minimum landscaped area, minimum lot size, garage width limits, private open space, solar access) are checked by plain code, not by an AI. The compliance section of every report is deterministic: given the same dimensions and the same LEP/DCP, you get the same answer every time. Any breach is flagged with the exact clause reference and the numeric margin, which is what your planner or certifier will ultimately need to see.

Layer 3 — Precedent grounding from 413,770 real DAs

Every determined DA lodged through NSW ePlanning since 2010 is in the precedent corpus. Refusals are structured into a taxonomy of about 40 refusal-reason categories (bulk & scale, streetscape, view loss, tree removal, heritage impact, etc.) with the source Notice of Determination linked. The likely-outcome section of a paid report is grounded in the base rate for your specific council, zone and dwelling type, plus the nearest comparable determinations by geography.

Where an LLM contributes and where it doesn’t

A large language model is used for two things only: (a) drafting narrative sections (the SEE outline, anticipated planner objections and their responses, the plain-English summary) in a citation-first style, and (b) mapping free-text refusal descriptions to the structured refusal-reason taxonomy. Numeric standards, permissibility calls, heritage overlay lookups, zone resolution, and clause references are never LLM-generated — they come from deterministic queries against the planning-instrument database and the NSW Spatial Services ArcGIS layers.

Refresh cadence and data sources

LEP and DCP corpora are refreshed monthly from the NSW legislation register and each council’s publications page. The precedent corpus is refreshed weekly from NSW ePlanning’s open-data feed. Zoning, heritage, bushfire, flood and land-reservation overlays are fetched live from the NSW Spatial Services ArcGIS endpoints at the moment you look up an address — so you’re never looking at a stale zone code. All source URLs are cited in every paid report.

Our accuracy guarantee

Because the process is auditable end-to-end, we back every paid report with an accuracy and quality guarantee. If a report contains a factual inaccuracy or quality issue, tell us within 30 days and we’ll correct it at no cost or refund the report price in full. Separately, if the determination on your DA (or your CDC) contradicts the outlook we published, we refund the report with no time limit — we track the outcome of every purchased report against council’s later determination so we can publicly report our historical accuracy.