# Market Research Framework

Use this reference to select research depth, plan a study, or make a stage-gate decision.

At each level, package authorized durable output as one complete HTML report per
[html-report.md](html-report.md), unless the user requests another format. The
outputs below describe content and recommendations, not separate required files.
Response-only advice and reviews remain in the conversation.

## Research levels

### Level 1: Scan

Use for early ideas or comparisons across many directions. Answer whether deeper research is justified.

Minimum coverage:

- bounded market definition;
- demand and momentum signals;
- representative competitors, including weak outcomes;
- initial player and UA signals;
- rough opportunity, risk, and studio-fit assessment.

Use a small balanced seed cohort rather than an exhaustive universe. When comparing multiple directions, apply comparable sampling logic to each and expand only where uncertainty could change the gate decision. Public store and publisher signals may begin the Scan while proprietary data is pending.

Typical output: brief, compact market map, competitor shortlist, hypotheses, and Hold/Reject or Research further recommendation.

### Level 2: Exploration

Use when the receiving team may invest in a prototype. Answer whether an evidenced opportunity exists and is reachable.

Typical coverage:

- roughly 20-30 competitors across the five research groups for one direction when the market and decision justify that depth;
- 5-10 purposeful deep dives;
- performance snapshots and trajectory analysis;
- product, progression, economy, monetization, LiveOps, player, and UA patterns;
- opportunity map, comparative scorecard, assumptions, and prototype-oriented validation plan.

Typical output: Hold/Reject, Approve test, or Approve prototype.

### Level 3: Investment

Use only when considering meaningful production investment. Add:

- revenue and unit-economics scenarios;
- CPI, retention, payer conversion, ARPDAU, and LTV assumptions with ranges;
- production cost, schedule, content throughput, and technical risk;
- publishing, UA, platform, and regional constraints;
- soft-launch success and kill criteria.

Typical output: Hold/Reject (Kill for an existing project), a bounded iteration, or Greenlight production.

Do not present speculative financial outputs as forecasts. Use scenarios and sensitivity analysis.

## Seven phases

### 1. DEFINE

Start with a decision, not a game list. Lock the research question, entry route, market scope, inclusion and exclusion criteria, initial hypotheses, known risks, and gate criteria. For open discovery, use studio constraints and player needs to bound several opportunity spaces before narrowing to a direction. Keep comparable sampling and research effort across directions.

Define the investment being considered, business objective, hard constraints, timebox, and stopping rule. Stop collecting when further evidence is unlikely to change the next action; when the timebox ends with decisive gaps, recommend a targeted follow-up or Hold.

Define markets behaviorally where store taxonomy is too broad. Distinguish a defining mechanic from a superficial feature.

### 2. MAP

Describe demand, scale, growth, concentration, release activity, publishers, platforms, geography, business models, subgenres, and adjacency. Ask where the market is moving and how value is distributed, not only whether it is large.

### 3. DISCOVER

Build the competitor universe without relying only on top-grossing titles:

| Group | Purpose |
|---|---|
| Direct | Similar proposition and experience |
| Mechanic | Shares a defining mechanic |
| Audience | Competes for the same player need or attention |
| Business | Provides a relevant monetization or LiveOps model |
| Weak / failed | Tests the thesis but did not sustain or scale |

A game may serve different roles in different studies, but keep one canonical game record.

### 4. DECONSTRUCT

Analyze positioning, audience, fantasy, promise, core/session/meta loops, controls, mastery, strategy, RNG, build depth, progression, collection, economy, monetization, LiveOps, content, UX, production complexity, and scalability.

For important features, explain the player or business role. A feature inventory without causal reasoning is not a deconstruction.

### 5. UNDERSTAND

Use two lenses:

- **Player:** Why install, stay, pay, quit, love, and complain?
- **Marketing:** What fantasy, hook, action, transformation, and payoff are creatives selling?

Compare the advertised promise with the product. Repeated complaints across games may indicate a gap; one small sample does not.

### 6. SYNTHESIZE

Convert evidence through:

`Signal -> Pattern -> Insight -> Hypothesis -> Opportunity`

White space is not automatically opportunity. Check demand, marketability, retention, monetization, feasibility, and the reason the gap persists. Use the concept-development reference to create distinct candidate solutions, then compare their evidence and next-test value. Treat concepts as hypotheses until tested.

### 7. VALIDATE

State the product thesis, evidence, risks, falsifiable assumptions, and cheapest meaningful experiments. Possible sequence:

`Creative test -> Concept test -> Prototype -> Playtest -> Store test -> Soft launch`

Choose experiments by uncertainty reduction. Define success, failure, and decision thresholds before reading results when feasible.

## Stage gates

### Scan gate

Continue only with credible demand or momentum, a plausible gap, plausible execution fit, and a tractable next research question.

### Exploration gate

Prototype only when more than one evidence class supports the opportunity, important risks are explicit, differentiation is player-visible, and a bounded prototype can test the thesis.

### Investment gate

Greenlight production only when critical product, marketability, retention, monetization, production, and financial assumptions meet predeclared evidence requirements appropriate to the actual business model and investment size. Record residual risks, downside scenarios and stop criteria. Unknown life-or-death assumptions block production approval; a plan to test them justifies only a bounded test or prototype recommendation.

Apply feasibility constraints before any score. A high market score cannot offset an unaffordable production plan. Separate analyst recommendation, owner decision, authorized amount/scope, and actual test result.