GLOBAL ARCHITECTURE DISCOVERYAn independent architecture index
GAD
0中文

PRACTICAL WORKFLOW

Compare site options without losing the assumptions

A repeatable early-feasibility review that separates automated exploration from the decision to proceed.

1. Write the comparison contract

Record the site boundary source, survey date, area definitions, target program and known exclusions. Mark unverified planning inputs as assumptions. Choose one decision, such as the building arrangement to take into a consultant review.

2. Generate within a bounded question

Use a tool suited to the question: TestFit for filtered site solves, Forma for the connection between massing and layouts, or Snaptrude for early brief-to-block studies. These are alternative entry points; this guide does not assert a tested connection between them.

3. Normalize the alternatives

Export or transcribe each option into the same area and capacity table. Preserve rejected results, filter overruns and manual edits. If an area definition cannot be aligned, do not compare its headline number.

4. Review the failure case

Ask a qualified reviewer to inspect the constraints most likely to invalidate the preferred option. Record the source of each decision and the work needed before the next stage. A tool-generated maximum is an exploration result, not an entitlement.

Worked review scenario

Illustrative pilot: use one familiar site and compare three arrangements with the same program. The useful outcome is a defensible reason to keep one, not a promise to find the highest yield. No area or financial outcome is asserted here.

When the pilot is ready to accept

  • Every important number can be traced to a model or explicit assumption.

  • The preferred option survives manual checks of the constraints used to select it.

  • Setup, correction and review effort are recorded alongside any time saved.

Related tools

Brief → spatial blocks

Snaptrude AI

An editable early-design workflow beginning with a brief and site, with a published distinction between initial features and roadmap ambitions.

Available · vendor documented
Inputs, limits & evaluation

GAD / REFERENCE LIBRARY

Research through architectural questions

116 source-linked records

Explore the collection

FOR ARCHITECTURE TECHNOLOGY BRANDS

Put a useful product in context.

Discuss a clearly labelled AI-column sponsorship or bilingual product feature. Editorial assessments remain independent.

Partnership formats

Continue exploring architecture: Projects · Practices · GAD calculators & practical tools · Editorial standards

GAD / FURTHER READING

From fast wind feedback to a defensible site comparison

A two-pass Forma wind workflow: hold assumptions steady, compare massing changes, verify shortlisted options and record the unresolved risks.

This documentation-based workflow is not a hands-on product test. The rapid-wind announcement dates to 6 June 2023; this is a study method, not new-launch news or planning-compliance certification. The Chinese counterpart is an abridged summary.

A wind map becomes useful when it changes a clearly defined design decision. For a courtyard scheme, that decision might be which massing arrangement deserves further investigation, or where to reserve space for entrances and outdoor seating. A faster map helps the team ask more questions; the comparison still needs a controlled method.

Autodesk introduced Forma's machine-learning rapid wind analysis in June 2023. Its current learning material distinguishes directional rapid predictions from simulation-based detailed analysis and identifies the latter for documentation and final comparison. Both provide wind-speed and estimated pedestrian-comfort results. This distinction should remain visible throughout the review. Autodesk announcement · Official wind-analysis guide

First pass: define a fair experiment

The following procedure is GAD's proposed review method. Begin with a baseline model and a short question: does opening one corner improve the intended sitting area without creating a new problem at the entrance? Record the boundary, north orientation, surrounding buildings, terrain assumptions, model revision and the analysis settings used. Mark missing context explicitly. Give every option the same comparison conditions.

Identify a small set of places that matter before looking at the results. Include the entrance, a route across the site, a stationary-use area and an exposed corner. This prevents an attractive overall colour map from distracting the team from a poor result at the location that actually determines the design.

Change one massing variable at a time. Save the baseline, the variation and a short description of the change. Compare each selected location using the same result type and display scale. A rotated block, lowered wing and new opening are different experiments; combining all three makes it harder to explain which move helped.

Second pass: challenge the shortlist

Run detailed analysis for the baseline and the shortlisted alternative. Carry their exact model revisions into the review. Record where the detailed result supports the earlier ranking and where it changes the conclusion. If the alternatives are close, keep the decision provisional rather than presenting a weak distinction as a clear winner.

Ask the appropriate wind specialist to review the assumptions, relevant directions, intended outdoor uses and any project-specific assessment requirements. Agree the required study scope before treating a result as evidence for a planning submission. A detailed software result still depends on the model and inputs supplied.

Make the decision traceable

An illustrative review package contains paired views, a location-by-location comparison, the model revisions, the proposed design response and named unresolved questions. Include a changed-context scenario if a nearby future building could materially alter the decision. This scenario is a sensitivity check, not a prediction of what will be built.

The acceptance gate is practical: another team member can reproduce the comparison and explain why the preferred option advances. Record setup and review effort alongside iteration time. The useful outcome is a defensible next design step with clearly assigned follow-up work.