01 Govern & Research
Harvest by registry
Uses the 30-theme registry, bounded query batches, theme-specific source domains, provider limits, and a target of 3-5 active signals per theme.
The Scanner Foresight Methodology
Foresight creation is a verified signal system, not a future-story generator.
The Scanner creates foresight by separating four intellectual acts: governing the research territory through a Strategic Foresight Theme Registry, harvesting external signals into a company-neutral active Foresight Knowledge Base, synthesizing company-specific implications through a separate interpretation contract, and converting qualified forces into scenarios, diagnosis context, quantitative pressure views, Market Intelligence, and Observatory dossiers. The governing rule remains simple: trend evidence can describe external pressure, but it cannot prove current company behavior.
01Thesis
The Scanner treats foresight as external pressure mapped against company readiness.
Foresight is not a forecast that claims to know the future. It is an evidence discipline: the research space is governed through a curated theme registry; candidate signals are harvested, checked for source and semantic quality, classified into active or audit status, and stored under a company-neutral contract. Company relevance is created later through a separate interpretation layer.
Foresight is company-neutral at source.
Central signals use the central_foresight_signal/v1 contract:
claim_type=foresight_context, company_specific=False,
requires_company_specific_corroboration=True, and
not_evidence_of_deployment=True. Company-verdict metadata is stripped from the central KB.
Company relevance is synthesized later.
The company scan maps verified KB signals to domain themes, priority dimensions, current reality, archetype, financial capacity, and top findings.
Active-context eligibility precedes narrative.
Only signals marked foresight_context_eligible=True and
foresight_signal_status=active_context can enter normal company synthesis.
Rejected candidates are retained in a separate audit store, not mixed into the active KB.
Source meaning and company relevance stay separate.
The central signal describes what is changing in the world. Company-level
foresight_links preserve that neutral source_says content and place any
company-specific hypothesis in a separate possible_relevance field.
02End-To-End Pipeline
The foresight system has a registry-governed knowledge loop, a company interpretation loop, and a portfolio intelligence layer.
The registry and KB loop define where the system may search, which signals may become active context, and which candidates remain audit-only. The company loop normally reuses active KB signals, while a narrow, time-budgeted registry fallback can search only curated source domains when the KB is too thin. Portfolio intelligence then reuses the separated central and company-level records without collapsing their evidence roles.
02 Classify
Test active context
Checks source URL, liveness, registry fit, credibility, readability, semantic depth, freshness, theme match, and trend support.
03 Store
Split active and audit
Stores active context in the central KB, rejected candidates in a separate audit file, deduplicates, ranks, caps each theme at 15, and avoids wiping healthy data.
04 Select
Map to company domain
Domain mapping selects FP themes, core dimensions, priority dimensions, climate exposure, and near-fit related-theme signals when a requested theme is empty.
05 Interpret
Create company forces
KB-first synthesis creates dimension forces, cross-impact, scenarios, and climate exposure. If fewer than three live KB signals remain, a curated registry fallback may run within a protected time budget.
06 Apply
Gate diagnosis and market use
Only sufficiently supported forces feed AI Diagnosis. Separated foresight links then support FAVI, demand-creation metrics, Market Intelligence, cross-impact interpretation, and Observatory evidence dossiers.
Foresight creation path
The default company path is KB-first: central signals are harvested, classified, cleaned, and stored before a company scan uses them. When the active KB does not provide at least three live signals, the Scanner may use a narrow registry fallback constrained to curated domains, source-quality rules, URL verification, and a protected scan-time budget. If that bounded fallback cannot run safely, foresight is skipped rather than fabricated. Company interpretation is saved under a separate contract and remains explicitly non-factual unless corroborated by company-specific evidence.
03Foresight KB Harvest
The Strategic Foresight Theme Registry is the controlled research and source-permission layer.
The harvester scans the 30-theme taxonomy through explicit theme-domain queries and carefully bounded Tier 1 exceptions. It combines run-scoped provider health controls, theme-specific batching, source-memory deduplication, freshness windows, semantic snippet requirements, content support checks, and a strict active-versus-audit storage contract.
01
Load, revalidate, and protect existing active signals
Existing theme records are rechecked. Dead, unreadable, stale, off-theme, or otherwise ineligible records move to the audit path; a failed refresh never replaces a healthy active theme with an empty file.
existing active KBaudit splitno destructive wipe
02
Harvest through bounded theme profiles
Focused, hard, and fragile themes use different micro-batch sizes and time budgets. OpenAI search-first discovery has a grounded Gemini fallback, while Claude can repair malformed JSON. Provider concurrency and run-level quarantine prevent one failing provider from destabilizing the full harvest.
bounded web researchprovider health3-5 active target
03
Apply the active-context contract
A candidate must have a real, live URL; match an explicit registry domain or a strongly matched Tier 1 exception; meet credibility, readability, semantic-snippet, freshness, theme, and trend-support requirements.
active_contextregistry fitfreshness status
04
Verify content support and semantic usefulness
The source-claim validation chain checks whether readable content directionally supports the macro signal and extracts a source quote. Semantic snippets must be rich enough for later company-scan selection; weak or unsupported records are not promoted into active context.
GPT verifierClaude Sonnet reviewfrontier fallback
05
Deduplicate, rank, separate, and store
Signals are keyed by URL or normalized title. The stronger active record wins, ranking considers source tier, credibility, snippet usefulness, registry fit, freshness, theme/trend support, and scan enrichment value. Active theme memory is capped at 15; rejected candidates remain auditable but cannot feed synthesis.
URL/title keyactive scoreMAX 15 active
| Harvest Rule | Implementation Meaning | Reliability Effect |
|---|---|---|
| Company-neutral horizontal signal | The record must describe a macro or industry force, use the foresight_context claim type, and carry no company-verdict metadata. |
Prevents the central KB from becoming a mixed store of trends, account claims, and inherited company conclusions. |
| Registry and freshness fit | Explicit theme-domain sources are preferred. Tier 1 exceptions require strong theme fit. Normal active windows are fresh within 12 months and current within 24 months; older official frameworks can remain as evergreen context. | Constrains search space and reduces stale or loosely related macro evidence. |
| Semantic usability | Readable source text, a sufficiently rich semantic snippet, theme match, and trend support are required for active scan-time use. | Prevents a live but content-poor page from becoming synthesis fuel merely because its title looks relevant. |
| Active / audit separation | active_context records stay in the central KB. rejected_audit candidates are stored separately under an audit contract and are not used by Scanner synthesis. |
Makes rejection visible without allowing rejected material to leak back into normal generation. |
| Provider-aware execution | Theme workers share provider-specific concurrency limits and can quarantine a timed-out provider for the remainder of the run. | Improves reproducibility and prevents cascading timeouts from degrading unrelated themes. |
04Company-Level Synthesis
A company scan converts macro context into company-specific relevance through a separate interpretation contract.
The company domain selects relevant foresight themes and priority dimensions. The Scanner retrieves active-context KB signals, re-verifies their URLs, and normally requires at least three live signals for synthesis. Near-fit themes may fill an empty requested theme. If the total verified set is still too thin, a narrow registry-source fallback can run only when scan budget permits; otherwise the system returns no new foresight instead of inventing it.
Domain Resolver
From company domain to foresight territory
SYN-01
The company domain maps to FP themes, core dimensions, priority dimensions, targeted source domains, and optional climate exposure.
DOMAIN_TO_FORESIGHTfp_themespriority_dimensions
Signal Selector
Active context first, near-fit second
SYN-02
Only company-neutral records with active foresight status and readable, verified support enter the normal prompt. If one mapped theme is empty, related themes can supply tagged near-fit signals without opening broad web search.
active_contextnear-fit fallbackno inherited company verdict
Synthesis Engine
KB-first, registry-bounded fallback
SYN-03
The normal path reasons over up to 12 re-verified KB signals without live search. If fewer than three remain, the fallback can search only curated registry domains and must stop when the protected scan-time budget is exhausted.
dimension forcescross-impactscenariosregistry fallback
Interpretation Contract
World signal and account hypothesis stay distinct
SYN-04
The same external force can imply different pressure depending on current reality, financial capacity, top findings, and adaptability archetype. The resulting link is still marked as foresight context—not as evidence of deployment or present company action.
source_sayspossible_relevancenot company fact
Why the company path is evidence-bounded rather than strictly search-free
Moving normal research to the central KB remains the primary reliability choice. The current code adds a narrower recovery path for thin themes: search is allowed only inside the Strategic Foresight Theme Registry, under time, source-domain, modal-language, URL-verification, and source-quality constraints. This improves coverage without returning to uncontrolled account-time browsing. Central signals and company interpretations are saved under separate lifecycle contracts, and classification namespaces prevent company archetypes, FAVI segments, adaptation segments, and foresight force profiles from being reused as if they were the same kind of label.
05Key Force Anatomy
A key force is a structured argument with a source contract, an interpretation contract, and a use boundary.
Each force binds a company-neutral macro signal to a Scanner knowledge dimension, a separately marked company relevance hypothesis, a time horizon, a source set, lifecycle state, and action timing. Only forces that retain readable, medium/high-credibility and verified support can feed generated diagnosis context.
Conceptual Object
Dimension Force
A dimension force answers two questions without merging them: what is changing externally, and why might that pressure matter to this company's readiness in a specific KB dimension?
external source meaning
dimension mapped
company relevance hypothesis
time bounded
TrendConcrete name of the force, tied to a central signal rather than a vague theme.
TypePlausible/Maturing, Probable/Emerging, Weakening/Declining, or Preposterous/Black Swan.
Timeframe6-12, 12-24, or 24-36 months.
DimensionAI Readiness, Operating Model, Service Portfolio, Strategy, or Values.
Source SaysCompany-neutral description of the external trend, retained separately from account interpretation.
Possible RelevanceModal company-specific hypothesis based on current reality, financial capacity, top findings, and archetype.
EvidencePrimary URL, verified supporting sources, source type, credibility, confidence, registry fit, readability, theme match, and trend support.
LifecycleWeak signal through market saturation/fading signal, plus momentum and action timing.
Contract
company_foresight_interpretation/v1, claim_type=foresight_context, and not_evidence_of_deployment=True.
Signal Type
Classifies uncertainty and maturity: emerging, maturing, declining, or black-swan style pressure.
Lifecycle Stage
Places the force on an adoption curve: weak signal, strong signal, hype, valley, go-to-market, breakout, established, saturation, weakening, or fading.
Action Timing
Translates lifecycle and company context into monitor, prepare, act now, last chance, or optimize.
Confidence & Credibility
Confidence reflects corroboration; credibility reflects the source tier and domain quality. Neither is treated as company certainty.
Context Eligibility
The force must retain active/readable source support, theme and trend support, medium/high credibility, and at least one verified source.
Scope & Namespace
A force profile, adaptability archetype, FAVI segment, and adaptation-capacity segment remain separate classification families.
06Verification And Guardrails
The active Foresight KB is governed by a macro-context contract, not by normal company-evidence promotion fields.
The current architecture deliberately removes company verdict vocabulary such as
generation_eligible, analysis_verified, entity-match status, and account names from
central signals. Eligibility is instead expressed through foresight-specific source, freshness, semantic, theme,
trend, and active/audit statuses.
| Gate | Question | Stored Fields | Failure Mode |
|---|---|---|---|
| Source URL and registry fit | Is there a real live URL from an explicit theme domain or a strongly matched Tier 1 institutional exception? | source_url, url_verified, source_registry_match, source_tier |
Missing/dead or non-curated sources move to audit unless the strict Tier 1 exception applies. |
| Credibility and readability | Is the source medium/high credibility and does it expose usable text, quote, or semantic content? | source_credibility, source_readable, content_verified, source_quote |
Low credibility or unreadable sources cannot become active context. |
| Semantic depth | Is the semantic snippet rich enough to support later company-scan selection and interpretation? | semantic_snippet, active score, minimum snippet thresholds |
Title-only or shallow records become snippet_too_short audit candidates. |
| Freshness | Is the signal fresh/current, or an explicitly acceptable evergreen official source? | published_at, source_age_months, freshness_status, freshness_reason |
Aging, stale, or unknown-date non-evergreen sources cannot remain active. |
| Theme and trend support | Does the content match the requested theme and directionally support the stored trend? | theme_match, trend_supported, trend_support_status |
Off-theme or unsupported signals move to rejected audit. |
| Active / audit contract | Can the record enter company synthesis? | foresight_context_eligible, foresight_signal_status, source_audit_status |
Only active_context records remain in the active KB; rejected_audit records are stored separately. |
| Company interpretation and language | Does account relevance remain a modal hypothesis, separate from what the source actually says? | source_says, possible_relevance, interpretation contract, sanitizer metadata |
Company names are removed from neutral source descriptions; deterministic obligations, deployment claims, and expired deadlines are rewritten or flagged. |
Allowed
"External financial-sector signals suggest growing AI-governance pressure. This may be relevant to the company if regulated AI use is in scope and current governance evidence remains limited."
Forbidden
"The company is deploying AI governance and must redesign its operating model." A macro source cannot prove current deployment, entity-specific non-compliance, or a direct obligation.
Post-synthesis source quality gate
Company-level forces are rechecked after synthesis. At least one force must retain a verified primary source, roughly one-third of forces must have verified primary support, and the payload needs at least two verified URLs across primary and supporting sources. AI Diagnosis applies an additional gate: only medium/high-confidence, medium/high-credibility forces with readable trend support and verified sources can enter the final foresight paragraph.
Strategic Foresight Themes
The Scanner builds and maintains its own curated Foresight Knowledge Base from 30 macro-level change themes. Harvested signals are validated, classified, deduplicated, and stored under these themes before company-level synthesis can use them.
Technology & DigitalArtificial Intelligence; Data & Digitalisation; Communication & Media
Economy & BusinessBusiness; Finance & Ownership; Global Economy; Digital & Sharing Economies; Services
Industry & InfrastructureIndustry & Manufacturing; Construction & Urbanisation; Energy; Freight & Logistics; Private & Public Transport
Environment & ResourcesClimate Change; Sustainability & Recycling; Nature; Food; Water & Oceans
Society & GovernancePublic Governance; Geopolitics; Security & Safety; Values & Ethics; Society
People & Well-beingHealth & Wellbeing; Education & Knowledge; Work & Income; Management & HR; Leisure & Lifestyles
Science & DemographicsSciences & Research; Population & Demographics
07Methodological Foundations
The system combines futures studies, OSINT validation, evidence promotion, company-readiness analysis, and a Popper-compatible methods mix.
The implementation is pragmatic rather than purely theoretical, but its logic is academically defensible: govern the research space, validate sources laterally, separate active context from rejected audit, map macro signals to organizational capacities, create company relevance under a separate contract, and distinguish plausible pressure from proven company facts. Rafael Popper's Foresight Diamond remains visible as a background logic: the Scanner combines evidence, expertise, creativity, and limited interaction rather than relying on a single foresight method.
Futures Studies
Horizon scanning and signal taxonomy
M-01
The 30 FP themes create a wide scan field so the system does not only chase technology headlines. Signal type and lifecycle stage distinguish weak signals, maturing trends, saturation, and decline.
OSINT
SIFT and lateral reading
M-02
The prompts instruct Stop, Investigate source, Find better coverage, and Trace claims. The code reinforces that with explicit theme-source registries, URL checks, source tiers, freshness windows, semantic snippets, theme/trend support, and provider-aware execution.
Evidence Theory
Active context rather than generic promotion
M-03
The important methodological move is not gathering many links. It is deciding which company-neutral signals belong in the active macro-context store and which candidates must remain in a separate audit path.
Organizational Diagnosis
External pressure meets internal readiness
M-04
A force becomes strategically meaningful only when mapped against reality score, archetype, financial capacity, and Scanner knowledge dimensions.
Three evidence scopes, three contracts
The central Foresight KB stores company-neutral macro context. Company scans create separate
company_foresight_interpretation/v1 records that preserve neutral source meaning and isolate
account relevance as a hypothesis. The Observatory then builds locked evidence dossiers from the eligible
central KB and stored company portfolio, records input manifests and exclusion reasons, requires evidence IDs
for material claims, and independently verifies its strategic synthesis. These layers are related, but none is
allowed to inherit the evidentiary status of another.
Popper-compatible reading: the Foresight Diamond
The Scanner should not be described as a full participatory foresight programme in Popper's sense. It is not running citizen panels, Delphi rounds, or workshop-based stakeholder deliberation. But the methodological dynamics are clearly aligned with Popper's Foresight Diamond: evidence grounds the signal, expertise frames relevance, creativity opens alternative futures, and interaction appears through review, maintenance, and renewal loops.
Relationship to common strategy frameworks
The Scanner does not reject familiar methods such as PESTEL, SWOT, Porter's Five Forces, scenario planning, or horizon scanning. It treats them as partial lenses. Its own method combines horizon scanning, evidence qualification, entity-aware verification, knowledge-base classification, cross-impact reasoning, and company-specific readiness analysis.
| Method | Useful For | Limitation In This Context | How The Scanner Extends It |
|---|---|---|---|
| PESTEL | Structuring macro-environmental forces across political, economic, social, technological, environmental, and legal categories. | It names categories of change, but does not verify source quality, entity relevance, signal maturity, or whether a pressure connects to a specific company's operating reality. | Uses the broader 30-theme foresight taxonomy, then gates each signal through credibility, content support, classification, and company-readiness mapping. |
| SWOT | Communicating strengths, weaknesses, opportunities, and threats in a simple executive form. | It can become subjective and static when assumptions, opinions, and verified facts are mixed without source governance. | Separates verified company reality from external opportunity and threat signals, then keeps weak, unsupported, or wrong-entity evidence out of generated conclusions. |
| Porter's Five Forces | Assessing industry structure, bargaining power, substitution, rivalry, and entry barriers. | It is less sensitive to cross-domain forces such as AI governance, data infrastructure, climate regulation, autonomous operations, work transformation, and ecosystem shifts. | Combines industry/domain context with cross-impact chains and priority dimensions, so external pressure can be mapped to operating model, strategy, service portfolio, values, and AI readiness. |
| Scenario planning | Exploring plausible alternative futures and strategic uncertainty. | It can become speculative if scenarios are not tied back to evidence quality, timeframe, source credibility, and present company constraints. | Creates micro-scenarios only from qualified forces, then links them to evidence status, action timing, risk/reward framing, and readiness limits. |
| Horizon scanning | Detecting weak signals, emerging issues, and trend momentum before they become mainstream planning assumptions. | Raw scanning can over-collect links and promote loosely related material if there is no strict evidence-promotion boundary. | Preserves horizon scanning as the research layer, but adds a curated Theme Registry, freshness and semantic-depth checks, active/audit separation, URL/content verification, source quotes, and company-level interpretation contracts before signals influence synthesis. |
| Foresight Diamond Capability | How It Appears In The Scanner | What The Wording Should Not Overclaim |
|---|---|---|
| Evidence | Strategic Theme Registry, active-context classification, URL/readability/freshness checks, semantic snippets, source quotes, theme/trend support, separate interpretation contracts, FAVI and evidence-ID-preserving Observatory dossiers. | Evidence of a trend is not evidence that a specific company has already acted. |
| Expertise | Curated 30-theme taxonomy, domain-to-foresight mapping, KB dimensions, source tiers, lifecycle classification, action timing. | The system uses codified expert frameworks; it does not replace domain expert judgment. |
| Creativity | Weak-signal scanning, black-swan category, micro-scenarios, cross-impact chains, and alternative future pressure narratives. | Creative foresight remains bounded by source support and modal language. |
| Interaction | Admin review, Verify KB, Verify Evidence, scan renewal, Market Intelligence feedback, and human interpretation of outputs. | This is the lightest Diamond capability in the current product; it is not a participatory panel or Delphi process. |
| Question | Scanner Answer | Methodological Role |
|---|---|---|
| What is changing? | FP theme signal harvested from authoritative external sources. | Horizon scanning. |
| How reliable is the signal? | Registry fit, source tier, credibility, URL/readability status, semantic depth, freshness, theme/trend support, corroboration, active/audit status, and later force-level source gate. | OSINT validation and evidence promotion. |
| Where does it hit the company? | KB dimension, domain theme, current findings, and archetype-specific impact. | Organizational readiness mapping. |
| How urgent is it? | Timeframe, lifecycle stage, trend momentum, and action timing. | Strategic timing analysis. |
| How does it cascade? | Cross-impact chain and micro-scenarios. | Systems thinking and scenario reasoning. |
| How are methods mixed? | Scanning, source review, signal classification, cross-impact, scenarios, scoring, and renewal functions are combined as one methodological system. | Popper-style methods mix. |
08FAVI And Market Use
Foresight becomes quantitative pressure only after external context and company readiness have been separated.
The quantitative layer uses validated company-level foresight_links, where each link retains its
central source meaning, possible company relevance, theme, dimension, timeframe, force type, credibility,
context status, and priority-dimension flag. The current implementation keeps legacy FAVI, but also separates
future exposure, present vulnerability, adaptation capacity, proactive demand creation, and urgent remediation.
Core foresight measures:
Foresight Pressure Score (FPS)
= average geometric mean(force weight, timeframe urgency, source credibility)
= priority-dimension links only
Intent Readiness Gap
= dimension pressure × (1 - current reality score / 100)
= may be conservatively bounded upward by validated current-vulnerability signals
Financial Adaptation Capacity (FAC)
= investment signal 40% + financial stability 30% + profitability 30%
Legacy FAVI
= geometric mean(FPS, 1 - FAC, readiness gap, megatrend exposure factor)
= sparse-data additive fallback when link or dimension coverage is insufficient
Separated decision views:
Current Vulnerability v2
= present weakness from reality, financial stress, intent-reality gap,
negative findings, evidence weakness, and quality-risk flags
Adaptation Capacity
= FAC 40% + current reality 35% + evidence confidence 15% + current stability 10%
Future Exposure Gap
= FPS 45% + intent readiness gap 40% + megatrend exposure 15%
Demand Creation Potential
= future exposure 30% + adaptation capacity 25% + latent need delta 20%
+ readiness gap 15% + score confidence 10%
Urgent Remediation Potential
= current vulnerability 35% + future exposure 25% + readiness gap 20%
+ capacity constraint 10% + score confidence 10%
Trapped
Low adaptation capacity, high future exposure
The company has limited ability to absorb a strong future-side pressure and may require careful prioritization before the need becomes explicit.
Capable but Exposed
High adaptation capacity, high future exposure
The company has a meaningful cushion and ability to act, but validated forces are building and may justify proactive demand-creation conversations.
Limited but Sheltered
Low adaptation capacity, lower future exposure
Capability is constrained, but the current foresight portfolio does not yet create the same external urgency.
Future-Ready
High adaptation capacity, lower future exposure
The company appears better positioned to absorb the validated external forces in the current evidence window.
Current vulnerability and future exposure are also shown as separate axes
The Market Intelligence UI distinguishes Critical (high current and future pressure), Emerging Risk (lower current weakness but high future exposure), Current Recovery Need (high current weakness but lower future exposure), and Well Positioned. Portfolio-relative percentiles are attached to the derived scores, while a separate confidence band reflects link sufficiency, dimension coverage, evidence confidence, and data completeness. The score and confidence are therefore not the same thing.
Market Intelligence and Observatory reuse
At portfolio level, Market Intelligence counts eligible themes, computes theme-by-dimension density, ranks future exposure, adaptation capacity, demand-creation potential, and remediation potential, and segments companies using calibrated portfolio cutoffs. The Cross-Impact interpreter reads the matrix in three ways: vertical dominant forces, horizontal convergence points, and a focused structural pattern; deterministic checks protect numeric grounding, required structure, and public-output hygiene.
The Observatory uses the independent central Foresight KB plus stored company-portfolio data. It records an input manifest with eligible and excluded records, builds locked L1-L3 evidence dossiers, preserves evidence IDs for material claims, and runs independent verification. It does not directly read the Company Evidence DB, company source ledger, or stakeholder layer.
09Boundaries And Reliability
The most important reliability feature is what each foresight lifecycle is prevented from claiming.
The Scanner separates central macro signals, company relevance hypotheses, quantitative portfolio indicators, and Observatory synthesis. The layers can inform one another, but they do not inherit each other's evidence status. This prevents trend context, company diagnosis, and market-ranking outputs from becoming one undifferentiated truth.
What foresight can do
Identify active external pressures, map them to readiness dimensions, express timing and momentum, create bounded company relevance hypotheses, support scenarios and cross-impact reasoning, and feed future-exposure and demand-creation views.
external pressurescenario contextfuture exposure
What foresight cannot do
Prove that a company has deployed a capability, is non-compliant, must act by a date, owns a current problem, or has buying intent unless separate company-specific evidence supports that claim.
no deployment proofno hard obligationno invented intent
Active Context
The central record passed registry, source, semantic, freshness, theme, and trend checks and can enter normal KB-first synthesis.
Rejected Audit
The candidate is retained separately for traceability but is contractually excluded from Scanner synthesis.
KB-First Generated
At least three live active KB signals were available and synthesis returned structured forces, cross-impact, and scenarios.
Registry Fallback
The KB was thin, so a time-budgeted search ran only against curated registry domains and still had to pass URL and force-level source gates.
Skipped / Retained
The verified set remained too small or the protected fallback budget was unavailable. The system produces no new foresight rather than filling the gap with speculation.
Scoped Interpretation
Neutral source_says content and company-specific possible_relevance remain separate and modal.
Diagnosis Eligible
Only medium/high-confidence and medium/high-credibility forces with readable trend support and verified sources can enter AI Diagnosis.
Score Confidence
Quantitative score confidence is calculated separately from the score using signal sufficiency, dimension coverage, evidence confidence, and data completeness.
Auditable
The report retains lifecycle contracts, active/audit status, source meaning, company relevance, force metadata, exclusion reasons, and Observatory input manifests.
Bottom line
The current foresight method is best understood as a governed chain of distinct objects: a curated registry defines the research territory; a candidate signal either becomes active macro context or rejected audit; active signals are re-verified and interpreted against company reality; the interpretation remains a hypothesis under a separate contract; only sufficiently supported forces enter diagnosis and quantitative views; and the portfolio layer builds evidence-ID-preserving Market Intelligence and Observatory dossiers from those separated stores. In Popper's terms, this remains a methods mix with a strong evidence and expertise bias, bounded creativity in scenarios and weak-signal logic, and a lighter interaction layer through review and renewal. Its strength is not prediction. Its strength is preserving the distinction between evidence of change in the world, a hypothesis about relevance to a company, and evidence of change inside that company.