Answers for research teams
evaluating Subtle Beacon.

Subtle Beacon is a game-focused platform for continuous Bayesian experimentation and preference research. It connects live tests, playtests, panels, surveys, telemetry, pricing research, and market simulation in one evidence system so studies inform one another instead of ending as isolated projects.

Subtle Beacon is designed for game research, product, analytics, monetization, marketing, live-ops, and production teams. Research leaders can control methods, design, quality gates, and interpretation, while less-specialized studio roles can use guided workflows for questions that would otherwise rely on gut feel or wait in a research backlog.

Game studios combine stated preference, playtest behavior, community signals, live telemetry, commercial outcomes, and rapidly changing product context. Generic survey or experimentation tools usually handle one channel well. Subtle Beacon is built to connect those channels around the same game, audience, and decision.

It reduces repeated study setup, fragmented fielding, hand-built analysis, and one-off reporting while making prior evidence reusable. Researchers can spend more time on framing, design choices, interpretation, and strategy instead of rebuilding the operating machinery for every question.

Yes. Subtle Beacon can run independently as a hosted experimentation and preference-research platform. Connections to Latent Spark, Patient Cartographer, and Anchored Horizon add game context, experiment triggers, and forecast handoff, but are not required.

Survey platforms are strong at collecting responses. Subtle Beacon adds game-specific study design, Bayesian fitting, adaptive task selection, live-product evidence, market simulation, reusable posteriors, decision rules, and follow-up workflows tied to the original product question.

Traditional conjoint software often treats each project as a separate design, fielding, fit, and export exercise. Subtle Beacon keeps conjoint inside a continuous evidence system that also supports live experiments, telemetry, pricing simulation, causal prior transfer, and portfolio monitoring.

Feature-flag platforms control exposure and may provide standard experiment analysis. Subtle Beacon adds Bayesian allocation and stopping, game-specific guardrails, stated-preference methods, pricing and bundle research, market simulation, cross-study learning, and proactive investigation when evidence channels disagree.

Analytics describes what happened in the product. Subtle Beacon helps determine what to test, what players value, whether an observed change is decision-relevant, how much uncertainty remains, and what evidence should be collected next.

Every study can become part of a living evidence state for the game. Historical research, live-market behavior, experiments, and new fieldwork can inform one another, reveal disagreement between channels, and reduce the cost and uncertainty of the next study.

Feature priorities, onboarding, progression, economies, live events, battle passes, cosmetics, DLC, bundles, editions, pricing, packaging, positioning, messages, store pages, audience segments, content roadmaps, and portfolio priorities.

A/B and multi-arm experiments; observational KPI studies; conjoint and adaptive conjoint; ACBC; MaxDiff; pairwise preference; menu-based choice; rating and constant-sum studies; concept and message testing; pricing, packaging, and bundle research; Gaussian-process preference learning; and behavioral preference modeling from telemetry.

Yes. Subtle Beacon supports HB-MNL for population- and individual-level part-worth estimation, attribute importance, willingness to pay, segment cuts, choice simulation, and downstream optimization.

Yes. A study can combine stated-preference collection with live exposure, telemetry, playtest behavior, or imported outcomes, allowing researchers to compare what players say with what they actually do.

Yes. Observational comparison and KPI-monitor studies are first-class study types. Their outputs are labeled according to evidence strength so descriptive or adjusted findings are not presented as randomized causal results.

Yes. It supports in-game assignment, exposure and outcome ingestion, Web SDK and Unity integration, analytics connectors, and remote-config or feature-flag integrations such as PlayFab, Statsig, Firebase, and LaunchDarkly.

Subtle Beacon can use Thompson Sampling to allocate more traffic to stronger-performing arms while preserving explicit exploration rails so uncertain alternatives are not abandoned prematurely.

Studies can stop on configured rules including probability of superiority, expected regret, uplift, credible-interval width, guardrail breach, non-inferiority, minimum exposure, stability windows, sample-ratio checks, or manual review.

Sample-ratio mismatch detection, exposure and ingestion validation, drift detection, guardrail monitoring, holdout validation, respondent quality checks where relevant, and configurable statistical release gates.

Yes. Portfolio-level monitoring, scheduled health scans, drift alerts, anomaly detection, and investigation workflows can operate across concurrent studies and imported KPIs.

The system maintains probability distributions over outcomes rather than reducing evidence to a significant or not-significant label. Teams can see which option is most likely best, the plausible effect range, how much uncertainty remains, and the expected cost of choosing incorrectly.

Probability-of-best answers how likely each option is to win. Expected regret estimates the downside of acting on the wrong option. Together they connect uncertainty to the actual decision rather than only reporting whether an effect passed a threshold.

Randomized, observational, adjusted, transfer-supported, and descriptive evidence can be labeled separately. Prior transfer is discounted when mechanisms differ, and observational results are not silently presented as experimental causality.

Hosted surveys, tenant-branded domains, external survey vendors, gamer panels, Discord and community studies, CRM audiences, email lists, LLM interviewers, manual CSV, playtests, and live-product channels.

Yes. Subtle Beacon can orchestrate external providers rather than requiring all collection to move to a proprietary panel or hosted survey surface.

Yes. Manual CSV, outcome templates, analytics connectors, warehouses, event pipelines, stores, attribution sources, CRM systems, and external KPI feeds can supply evidence.

Yes. Subtle Beacon can use posterior evidence from previous studies as priors for a new study when the population, mechanism, context, and decision are sufficiently related.

Potentially. Informative prior evidence can reduce the amount of new data needed when transfer is justified. The system preserves uncertainty and validates the transfer rather than assuming every superficially similar study is reusable.

Yes. Signal-divergence rules can detect cases such as community sentiment conflicting with playtest behavior, preference research conflicting with live outcomes, or portfolio KPIs moving against experiment expectations.

Yes. Adaptive conjoint, MaxDiff, and related workflows can select tasks or design rows based on expected information gain, reducing respondent burden or concentrating evidence where uncertainty matters most.

Signal Intelligence is the proactive layer that monitors portfolios and evidence channels for drift, anomalies, and disagreement, then proposes the next research needed to understand what changed.

Depending on method: posterior estimates, credible intervals, probability-of-best, expected regret, segment differences, part-worths, attribute importances, willingness-to-pay ranges, choice shares, guardrail state, diagnostics, charts, decision briefs, and recommended follow-up work.

Yes. It supports price sensitivity, bundle optimization, segment scenarios, choice-share simulation, revenue and profit tradeoffs, source-of-volume, TURF-style reach, and sensitivity analysis.

For common questions, guided setup can move from a plain-language decision to a recommended method, validated configuration, sample and cost estimate, and fielding-ready preview in minutes. Collection time still depends on audience and channel.

Yes. Guided flows, templates, recommended methods, validated defaults, approval gates, and plain-language outputs can support designers, producers, marketers, and product managers while preserving research-defined controls.

Bayesian posterior updates, Thompson Sampling, conjoint and MaxDiff fitting, optimal design, market simulation, guardrail rules, SRM detection, drift detection, causal transfer assessment, ingestion, and governance are statistical or operational systems rather than language-model outputs.

Yes. Product analytics, data warehouses, event pipelines, stores, attribution, revenue systems, CRM, feature flags, remote config, survey vendors, and panels can remain in place.

No. Federation is based on controlled summary statistics and privacy-safe artifacts, not raw respondent rows or unrestricted cross-tenant access.

Subtle Beacon is offered as hosted multi-tenant SaaS operated by Iridae. Alternative deployment requirements can be discussed separately.

Tenant isolation, role-based access control, secure tenant-scoped data handling, idempotent operations, rate limits, health probes, observability, runbooks, entitlements, usage reporting, and governed human and agent access.