Methodology
A game can look promising and still fail because the control feel is wrong, the learning objective is vague, the cultural references are shallow, the target device struggles, or the core loop is not understood. Our method puts those questions into builds early.
Evidence-gated production
The method is deliberately strict: every phase must produce evidence that changes what happens next.
We define the player, platform, core verb, entertainment target, educative intent, cultural context, business model, constraints, and non-goals. The output is not a pitch deck; it is a decision brief.
Player promise, target device class, concept risks, success measure, exclusions, and first prototype question.Every expensive unknown is written as a testable assumption: fun, comprehension, performance, scope, cultural safety, rights, monetisation, and production capacity.
Ranked risks with owner, test method, threshold, cost of being wrong, and go/narrow/pivot/stop consequence.We build the smallest loop that can answer the riskiest question. Fidelity is added only where the hypothesis needs it, such as animation timing, input feel, or UI feedback.
Playable build, instrumentation notes, known limitations, device observations, and a decision memo.When the prototype earns it, we build a representative slice with real art direction, UI, audio, controls, pipeline, performance, QA, and release discipline.
Source-controlled build, performance budget, content pipeline proof, save/load check, accessibility pass, and production estimate.Systems are split into simulation, presentation, authoring data, persistence, debug views, and tests. This keeps content growth from turning into untraceable controller code.
System specs, schemas, stable IDs, rule tests, save migrations, integration contracts, and debug tooling.We test with intended players and target devices, separating player value, usability, delight, technical feasibility, cultural safety, and commercial evidence.
Observed behaviour, tester context, metric limits, contradictions, and a clear advance, narrow, repeat, pivot, or stop recommendation.Gate criteria
These checks prevent a team from confusing movement with progress. If a proof gate fails, the honest answer is to narrow, repeat, pivot, or stop.
