A planning consultant receives a call from a developer client. They have an option on a 0.4-hectare urban infill site in the London Borough of Islington. They need to know the maximum development potential before they exchange.
Under a traditional workflow, the consultant spends three to five days manually calculating the zoning parameters, drawing a rough massing in two dimensions, checking each constraint, and running a viability model in a spreadsheet. The client needs the answer in 48 hours.
Under the algorithmic workflow, the consultant spends 90 minutes — and produces a more rigorous output.
This is not a hypothetical. It is the worked example from the Computational Feasibility guide — and the 90-minute timeframe is not a marketing claim. It is the measured time from sitting down with the Local Plan to having a draft feasibility report ready for MRTPI review.
Why the Four Steps Must Be Completed in Sequence
The planning feasibility workflow has four steps for a reason. Each step establishes the conditions for the next. Skipping steps — or completing them out of order — is the most common source of error in AI-assisted planning consultancy work.
The temptation is to jump to Step 3 (viability modelling in ChatGPT) because that is where the commercially useful numbers emerge. But the viability model produces meaningless output if the geometric input is wrong. And the geometric input is wrong if the regulatory parameters were not verified correctly in Step 1.
The sequence:
- Step 1: Regulatory parameter ingestion — source every value from the Local Plan with policy reference numbers
- Step 2: Generative envelope creation — Finch 3D generates the Maximum Feasible Envelope; consultant reviews against officer interpretation
- Step 3: Real-time feasibility tuning — viability scenarios in ChatGPT; CIL and S.106 modelled across multiple yield assumptions
- Step 4: Reporting output — feasibility report drafted, reviewed by MRTPI-qualified consultant, issued to client
Step 1: Regulatory Parameter Ingestion
This is the step that determines everything else. A parameter entered incorrectly into Finch 3D produces an envelope that is wrong — and an incorrect envelope fed into a ChatGPT viability model produces misleading commercial advice.
The regulatory parameter set must be sourced directly from the applicable Local Plan and SPDs, with policy reference numbers noted for every value. The date of the Local Plan must also be noted — a policy that was saved from an older NPPF iteration may apply different tests to one adopted under NPPF 2023.
The minimum parameter set for a UK urban residential scheme:
- Floor Area Ratio (FAR): Expressed as GIA divided by site area. Note which area definition the LPA uses — GEA, GIA, or NIA — and confirm before inputting. Different LPAs use different definitions in different contexts. This is a common source of calculation errors.
- Height limits: Measured in metres above ground level or above ordnance datum (AOD). Confirm the measurement datum with the LPA — finished ground level versus existing ground level can change the effective height by a storey on sloped sites. Some LPAs specify both; where both apply, the lower governs.
- Plot coverage: Expressed as a percentage of gross site area. Note whether the LPA includes party walls in the coverage calculation or excludes them. This can change the effective coverage by 3–5 percentage points on terraced or attached configurations.
- Setback requirements: Minimum horizontal distance from each site boundary to the building face. Confirm the measurement point — edge of boundary treatment, centre of boundary, or face of external wall. Standard UK residential rear setback is 9–12m to preserve the 45-degree daylight cone from adjacent habitable windows.
- Angular plane: The 45-degree rule is measured from the centre of the nearest ground-floor habitable room window at 1.7m above ground level. The 25-degree vertical angle is measured from the centre of a neighbour's main window to the top of the proposed building at the midpoint of the wall. Both derive from BS EN 17037:2018 (the current mandatory UK standard, which superseded BRE 209 in 2022). The applicable standard must be confirmed with the LPA planning officer — different boroughs apply these rules differently.
- CIL rate: From the LPA's published Charging Schedule. The rate must be confirmed for the correct year — CIL is charged at the index applicable to the year of planning consent, not the year of this report. A consent granted in 2028 will be charged at a higher rate than today's figure due to index-linking.
- S.106 heads of terms: Negotiated at pre-application stage. The LPA's standard requirements must be established before viability modelling. In London, affordable housing policy (London Plan) requires 50% on-site on schemes of ten or more units. This is not a negotiating position — it is the policy requirement. The viability case for reducing it must be established in the feasibility report.
- Permitted Development Rights: Class MA (commercial to residential) allows up to 1,500 square metres GIA without a full planning application. An Article 4 direction removes these rights. PDR availability must be confirmed with the LPA before any acquisition strategy relies on it.
Step 2: Generative Envelope Creation
With the parameter set verified and entered into Finch 3D's rule engine, the algorithm generates the Maximum Feasible Envelope — the largest volumetric massing that simultaneously satisfies all input parameters.
The planning consultant then reviews the 3D model against the planning officer's likely interpretation. This is where professional knowledge is applied to algorithmic output — and where the consultant earns their fee.
Two things must always be checked at this stage:
- The angular plane: Finch 3D applies the 45-degree rule as a numerical setback parameter. But the 45-degree cone from a neighbour's window is three-dimensional — it extends in both horizontal and vertical planes simultaneously. A setback that is compliant on plan may still breach the vertical component of the cone. The consultant must verify that the algorithm's simple setback parameter actually captures the full three-dimensional constraint.
- The height reading: Confirm whether the LPA measures height from finished ground level or existing ground level. On a sloped site, this can mean the difference between a 12-storey building and a 14-storey building at the same AOD cap. The algorithm uses whatever datum the consultant inputs. If the datum is wrong, the envelope is wrong.
The corrected Maximum Feasible Envelope from the Islington example was approximately 9,600 square metres GIA — not the 12,000 square metres that the raw algorithmic output suggested, because the algorithm's simple rear setback parameter did not fully capture the three-dimensional 45-degree cone from the adjacent Victorian terraced housing.
Step 3: Real-Time Feasibility Tuning
With the MFE established, four viability scenarios are run in ChatGPT. Each scenario calculates CIL, S.106, gross development value, net development value, and residual land value. The client sees the viable outcome at each scenario before committing to site acquisition.
- Scenario A — Baseline MFE at full affordable housing: This is the worst-case commercial position. Full policy-compliant envelope at 50% affordable housing (London Plan). CIL and S.106 calculated at full policy ask. This scenario establishes the commercial floor — the minimum viable land value the developer can justify.
- Scenario B — Reduced footprint with increased landscaping: Footprint reduced by 10% to provide additional public open space. Some LPAs offer viability credit for public benefit. This scenario tests whether the planning gain is worth the commercial cost.
- Scenario C — Increased height with BRE/BS EN 17037 re-check: Height increased by two storeys. Before modelling this scenario, the angular plane must be re-checked — if the increased height breaches the 45-degree cone, this scenario fails on daylight/sunlight grounds and must be abandoned. If it passes, the additional yield is tested for viability against incremental construction cost.
- Scenario D — Affordable housing optimisation: Tenure split negotiated at 35% with a Viability Review Mechanism. This is the standard negotiating route in London. The NDV improvement under Scenario D versus Scenario A is the consultant's primary lever in pre-application negotiations with the LPA.
The yield gap — the difference between the MFE (what can be built within the zoning rules) and the MVY (what can be built, sold, and financially sustained after all costs) — is disclosed prominently to the client. In the Islington example, the MFE was 9,600 square metres GIA, but the effective viable private yield after affordable housing deduction was approximately 1,950–2,300 square metres GIA. This gap is the commercial evidence base for the viability negotiation.
Step 4: Reporting Output
The feasibility report is the final deliverable. Its structure follows a standard professional format:
- Executive Summary (maximum one page): Maximum yield, headline viability position, key risks, top-line recommendation.
- Planning Policy Context: Applicable Local Plan, SPDs, NPPF, and London Plan (where applicable). Every parameter cited with its policy reference number.
- Maximum Development Potential (MFE): Finch 3D envelope description, parameter set used, zoning constraints summary.
- Key Planning Constraints: Daylight/sunlight, heritage, TPOs, CIL zone, PDR availability. For each constraint, the specific assessment methodology and applicable standard.
- Viability Position (MFE vs. MVY): Sensitivity analysis table from Step 3. Yield gap disclosed prominently. The distinction between maximum feasible and maximum viable clearly labelled.
- Recommendation: Proceed/do not proceed/negotiate. Design strategy advice. Acquisition price guidance based on MVY, not MFE.
- Disclaimer: Professional limitations statement, AI tool disclosure, MRTPI Code of Professional Conduct compliance statement.
Every number from Finch 3D must be independently verified against Local Plan policy before inclusion in the report. The algorithm finds the geometric ceiling; the consultant applies the policy interpretation.
What This Workflow Cannot Do
Finch 3D generates the Maximum Feasible Envelope. It does not interpret local planning policy nuance. It does not calculate CIL liability. It does not model affordable housing viability. It does not assess heritage impacts, TPOs, ground conditions, or site access constraints. These are the professional judgements that only the MRTPI-qualified planning consultant can make.
All viability figures from ChatGPT must be independently verified by a qualified quantity surveyor or RICS-registered valuer before use in financial close, acquisition pricing, or planning appeal documentation.
The formal disclaimer that must appear at the foot of every feasibility report produced under this methodology:
> "This report has been prepared by [Consultant Name], MRTPI, of [Firm Name], for the sole use of [Client Name] in connection with the site at [Site Address]. It is not a planning application. It is not a viability assessment. It is not a RICS valuation. It is not legal or financial advice. All yield figures require independent verification against the current adopted Local Plan before use in any acquisition, financing, or planning decision."
Related Resources
The complete four-step feasibility-to-advice workflow — including the regulatory parameter verification checklist, the three-yield distinction framework, the CIL and S.106 calculation methodology, and the formal feasibility report disclaimer — is documented in full in Computational Feasibility: Algorithmic Zoning Envelope Generation. That guide includes the worked Islington example with all four viability scenarios and the yield gap calculation.
If you are working at the earlier architectural design stage and need to test the shortlisted massing directions against environmental performance targets, the Parametric Iteration in Early-Phase Design guide covers the parallel workflow using Autodesk Forma — from site parameterisation through to client presentation, with the fails criteria framework and LOD accuracy table.
Stop Guessing, Start Systemising
The planning consultant who uses algorithmic tools without understanding the distinction between MFE, MME, and MVY will produce fast, accurate geometric output and misleading commercial advice. The consultant who understands the distinction — and applies it in the right sequence — will outperform both the manual-only competitor and the tool-only user.
The Computational Feasibility guide has the complete four-step workflow, the nine-point zoning parameter verification checklist, the CIL/S.106 calculation methodology, and the formal feasibility report disclaimer. Ninety minutes of structured analysis instead of three to five days of manual calculation.
Grab it here
Practical AI for Professionals