DOCS / SIMULATE

Simulate

Deep uncertainty analysis and the NS Loop.

Simulation helps users examine strategies when the environment can change and other actors can respond.

Deep uncertainty analysis

The deep-uncertainty workflow generates varied futures, simulates strategy performance, reduces the results, and groups similar outcomes. It is designed for conditions where assigning trustworthy probabilities to every future would be misleading.

Use it to ask:

• Which actions perform acceptably across many futures?
• Under what conditions does a preferred action fail?
• Which uncertainties change the decision?
• What early indicators would justify changing course?

Cluster labels are summaries of simulated outcomes. They are not discovered laws of the market.

NS Loop

The NS Loop is a turn-based competitive simulation seeded from an approved market model. The human participant selects an action while model-controlled competitors respond according to their configured context and behavioural profiles.

One turn represents a complete campaign cycle. Within-cycle effects are treated as realised at the turn boundary. The loop can record:

• the state entering each turn;
• the player's action and rationale;
• anticipated competitor moves;
• resulting score and share changes;
• predictions made before the outcome;
• later observations; and
• calibration adjustments.

Contestable territory

The simulation does not assume that all market share moves instantly. Organisations can have different contestable fractions based on maturity. This represents structural protection such as switching costs, distribution, organisational inertia, and network effects without claiming to measure each mechanism separately.

Important assumptions

• Domain scores persist, but the current model does not add a separate carry-over momentum process.
• Revenue impact is presented per turn unless explicitly stated otherwise.
• Share-to-revenue conversion is linear in the current implementation.
• Results depend on the seed data, actor profiles, action set, and market-physics assumptions.
• Model-generated competitor actions are plausible strategic responses, not predictions of a named firm's actual intent.

Calibration

After real observations become available, compare them with recorded predictions. Update competitor profiles only with a documented reason. Calibration improves the usefulness of future turns; it does not retroactively make earlier simulations factual.

Related product page: Simulate — product overview

Documentation draft · Source baseline 16 August 2026