Concept Parti

Concept Parti

The parti is the single idea that organises everything else — the diagram you could draw on a napkin that the whole building obeys. "A wall of rammed earth that runs from garden to interior, carrying light, structure, and the memory of the site's soil." Not a style, not a mood: a structuring principle. This skill helps an architect find two or three strong ones and understand what each commits them to.

When to use

• The blank-page moment: brief in hand, no organising idea yet.
• When a scheme feels like a pile of moves with no spine.
• To generate genuinely distinct directions before committing — not variations on one, but different bets.

Input

• The brief, or at least the programme, site, and intent.
• Any fixed constraints or non-negotiables.
• What the architect is drawn to (a material, a view, a feeling), if anything — a parti often crystallises around one thing they can't stop thinking about.

Process

1. Find the generative tension or the given. Most strong partis come from resolving a tension (public vs private, heavy vs light, old vs new) or from seizing the site's one defining condition (the view, the slope, the light, the wall you must keep). Name it.
2. Propose two or three distinct directions — each a different bet on what the project is fundamentally about, not three flavours of the same plan. If they could all coexist in one building, they're not distinct enough.
3. For each, state the idea in one sentence, then draw it in words (the diagram), then list the moves it forces and the moves it forbids.
4. Be honest about the trade. Every parti buys something and pays for something. The courtyard scheme buys privacy and calm and pays in area and a dead street face. Name the cost.
5. Recommend, don't just array. One strong direction beats three weak ones. Say which you'd back and why — while leaving the choice with the architect.

What makes a parti strong (test each against these)

• Legible — a stranger could read it from the plan. If you can't draw it simply, it isn't one yet.
• Generative — it decides things. It tells you where the entrance goes, where the section steps, where light enters. A parti that doesn't force decisions is a slogan.
• Specific to this project — it answers this brief on this site. If it would suit any project, it's a cliché, not a parti.
• Honest — it can carry structure, circulation, and light, not just a render. A parti that collapses the moment you draw a section is decoration.
• One idea — not three good ideas negotiating. The discipline is subtraction.

Common failure modes (avoid)

• The slogan — "indoor-outdoor living," "form follows function." True of everything; organises nothing.
• The mood in disguise — "warm and minimal" is an atmosphere (that's space-feel / material-palette), not an organising principle.
• The everything-parti — five ideas bundled so none can be read.
• Retrofitted rationale — inventing a story for moves already made. The parti should drive the moves.

Output format

## Concept Directions — [project]

**The generative question:** [the tension or given the parti resolves]

### Direction A — "[the idea in one sentence]"
- **The diagram:** [describe the organising move in words — what a section/plan would show]
- **What it drives:** [entrance, section, light, circulation — the decisions it forces]
- **What it trades away:** [the honest cost — area, cost, a compromised elevation, a lost view]
- **Strongest when:** [the condition under which this bet pays off]

### Direction B — "..."
### Direction C — "..."

**Recommendation:** [which you'd back, and the reason — one paragraph. The choice stays with the architect.]

What you don't do

• Don't offer a mood or a style as a parti. A parti organises; a mood decorates.
• Don't produce near-identical options. If they're not genuinely different bets, collapse them and offer fewer, stronger ones.
• Don't decorate a decision already made. If the architect has a parti, use design-critique to test it instead.
• Don't hide the cost. Every direction pays for something; name it.
• Don't hedge into five weak options. Two or three strong ones, and a recommendation.