The Diagnostic (reference)

The Diagnostic is a 45-question survey about how your organisation makes decisions. It is not a personality test or a culture survey. It measures what you actually do, not what you aspire to. The output is a persona: a markdown file that describes how your organisation thinks and decides in enough detail to instruct a language model to reason like it.

#Overview

The 45 questions cover 15 strategic Ingredients across three domains: what you are trying to do, how you decide, and what you value. Each Ingredient is scored 0-100 from the survey responses.

Those 15 scores drive two outputs. First, the closest of 16 archetype centroids becomes your Flavour (a named strategic archetype). Second, the scores are compiled into a persona: a markdown document used as a system prompt. Load the persona into any language model and it answers as your organisation would.

The whole survey takes about 15 minutes. Answer as the organisation actually behaves, not as it would like to behave. That is the most important rule.

#The three registers

The 45 questions use three formats, called registers. Each measures a different aspect of behaviour.

• Preference (sliders).“Where does your organisation sit on this spectrum?” These capture stated position. Fastest to answer; least resistant to social desirability bias.
• Observed behaviour (multi-select).“Which of these have you seen your organisation do in the last 12 months?” These capture what actually happened. More reliable than preference answers because they are grounded in specific events.
• Forced choice (scenario questions).“Your organisation is in this situation. It must choose one path. Which one?” These capture revealed preference under constraint. The most reliable register; the hardest to answer honestly.

The gap between your preference scores and your observed-behaviour scores is usually where the most interesting signal is. A large gap means the organisation says one thing and does another.

#How scoring works

Each response maps to one or more Ingredient dimension scores. The mapping is not disclosed question-by-question. Revealing it would encourage answers optimised for the score rather than the truth.

After all 45 questions are submitted:

1. Each Ingredient receives a score from 0-100, weighted across all questions that touch it.
2. The 15 scores are compared against the 16 archetype centroids using Euclidean distance in the 15-dimensional ingredient space.
3. The nearest archetype is your Flavour. Distances to all five clusters determine your cluster assignment (conservative, aggressive, culture, system, independent). The cluster proximity bar on the results page shows all five distances.

You can retake the diagnostic at any time. Each submission creates a new persona entry. Previous entries are preserved in your library.

#The persona

The persona is a markdown document, typically 800-1,200 words. It is the primary output of the diagnostic and the file you load into language models to make them reason like your organisation. Structure:

• Overview. A plain-language description of how this organisation decides.
• Strategic Ingredients. The 15 dimension scores and a brief note on what each means for this profile.
• Principles. The decision rules this organisation operates by, derived from the Ingredient scores.
• Win conditions. What success looks like for this organisation.
• Red lines. What this organisation will not do, even under pressure.
• Posture notes (optional). How this organisation tends to read and react to other Flavours. Populated automatically if the relevant Ingredients are strongly scored; can be edited in Studio.

The persona is used as a system prompt by the twin chat interface and by ClaudeAgent instances in the wargame. You can also copy it and load it manually into any language model that accepts a system prompt.

#Perspective

The diagnostic can be filled from any of five perspectives, set at the /diagnostic/start screen.

• Self.You are answering about your own organisation. The resulting persona is framed in first-person-plural (“we would...”).
• Competitor. You are answering as you believe a competitor would, based on observed behaviour. The persona carries a banner noting it represents your read, not a self-report.
• Partner. Same as competitor but for partners. Framing adjusts in the persona.
• Regulator. For modelling regulatory bodies or government actors.
• Hypothetical.For building a fictional challenger, target, or scenario player. No “the organisation I am describing” framing; treated as a standalone entity.

The perspective is set before the questionnaire starts. If you selected the wrong one, retake from the /diagnostic/start screen.

#Saving to the library

After submitting, you land on the results page. A Save button appears at the top. You can save as:

• Private. Only visible to you.
• Public. Visible to other users browsing the library.

Saving creates a snapshot of the persona at that moment. A later retake creates a new library entry; it does not overwrite the old one.

When you import a persona from the library into a scenario in Studio, a snapshot of the briefing is copied into the scenario draft. It is not a live link. If you edit the library persona later, the scenario copy is not updated.