intelligence-collection-disciplines skill

Intelligence Collection Disciplines

Purpose

Stop doing competitive research like a term paper. The intelligence community solved this problem decades ago: they don't collect "data," they run collection disciplines — independent channels, each with its own sources, tradecraft, and blind spots — and then they fuse them. This skill gives product managers that playbook: eight disciplines, each telling you (1) what to collect, (2) where to collect it, (3) which signal → inference chains to run, and (4) which PM artifact it feeds (TAM/SAM/SOM, ICPs, personas, battle cards, roadmap bets, win/loss, positioning, pricing). One signal is an anecdote; three correlated signals from independent disciplines is intelligence.

Input

Works best with the six instantiation variables below — fill them once and every discipline becomes engagement-specific. Leave them blank and this stays a teaching artifact.

~~~ [TARGET] = The competitor, partner, or acquirer you're researching [MARKET] = The category and its NAICS/SIC/NACE codes (or nearest equivalent) [GEOGRAPHY] = Regions in scope, at COUNTRY level, not continent level [BUYER] = Who signs the check (drives review sites, job titles, conferences) [CAPABILITY] = The strategic move you suspect (platform play, market entry, pricing shift, compliance land-grab, etc.) [DECISION] = What this research will change (roadmap bet, positioning, pricing, market entry, deal defense, ICP refresh) ~~~

If [DECISION] is blank, stop. Research without a decision is a hobby.

Anything supplied with the invocation itself — text after the skill name, a pasted context dump, or an appended ARGUMENTS: line — counts as variables already filled. Use it; don't re-ask.

Arriving empty-handed? That works too. Read the disciplines as a course in tradecraft, or ask the agent to walk the instantiation block with you one variable at a time.

Example invocation: Use intelligence-collection-disciplines — [TARGET]: Meridian Freight Systems, [CAPABILITY]: suspected move into warehouse robotics, [DECISION]: whether we accelerate our own integration roadmap.

Key Concepts

The Eight Disciplines

release can lie, but a customs record, a patent filing, and thirty job posts are three separate bureaucracies that would all have to be lying in the same direction.

• Independence is the design. The disciplines matter because they fail differently: a press

interpretation, labeled per the autonomous-investigation protocol (Fact / Inference / Assumption).

• Signal → inference chains are the tradecraft: each observable signal maps to a bounded

observable. See Guardrails in Common Pitfalls.

• What this is NOT: espionage. Every source below is published, filed, posted, or publicly

Application

1. OSINT — The Journalist's Desk

What a good beat reporter knows before the press release drops.

Signal → inference chains:

• [TARGET] exec suddenly posting about a new problem space → positioning pivot incoming (execs test messaging on social 3-6 months before launch)
• Sponsor tier jump at a [MARKET] conference → market entry or doubling down
• Review complaints clustering on one feature → their roadmap pressure point = your battle card ammo
• Analyst briefing requests (visible via analyst posts) → category creation attempt
• Webinar topics shifting → what they're teaching the market is what they're about to sell
• Sudden silence on a product line → sunset in progress
• Prediction-market odds moving on a regulation or milestone that gates [MARKET] → crowd-priced expectations for scenario planning; treat as a leading indicator of consensus, not ground truth, and check liquidity before trusting the number

Feeds: battle cards (objection handling from review mining), positioning (their words vs. customer words gap), win/loss context.

2. FININT — The Forensic Accountant

Follow the money. Companies lie in press releases. They lie less in filings, because lying there is a felony.

Signal → inference chains:

• Risk Factors section changes year-over-year in an annual filing → what genuinely scares them (they must disclose it)
• Segment reporting restructure → strategic reprioritization; follow which segment got promoted
• Earnings call Q&A dodges (analysts ask, execs deflect) → soft spot; probe it in your positioning
• New entity registrations or branches in [GEOGRAPHY] → market entry before any announcement
• Deferred revenue trends → actual sales momentum vs. stated momentum
• Merger filings → market definitions and named competitors, straight from [TARGET]'s own lawyers
• Sovereign fund or state-aid money backing [TARGET] → their runway math just changed; discount-pressure plays won't work
• Prior Information Notices and expressions of interest → tenders telegraphed 3-24 months out

Feeds: battle cards (financial stress = "they'll be desperate at quarter-end" plays), SOM capture rates (their revenue ÷ claimed customer count = deal size reality check, feeding the GEOINT/DEMOINT sizing recipe), account targeting (procurement award patterns).

3. GEOINT/DEMOINT — The Cartographer

The terrain map, not troop movement. Government statistics are free intelligence most PMs never open — and the backbone of every ICP, persona, and TAM that survives scrutiny.

Signal → inference chains:

• Establishment counts by industry code and employee band → the denominator for bottom-up TAM
• Regional industry concentration → where your SOM actually lives, and where field sales should live
• Occupation growth curves for your [BUYER] roles → is the population you sell to growing or shrinking
• Wage trends in buyer roles → willingness-to-pay ceiling shifts; pricing corridor validation
• Firmographic distributions (size bands, legal forms, sectors) → ICP boundaries drawn from data, not vibes
• Buyer-title prevalence by [GEOGRAPHY] → persona localization; the "VP of Product" you message in Boston is a "Head of Digital" in Frankfurt and may not exist in Riyadh
• Trade-flow shifts in product-specific codes → market entry or supply relocation before any announcement

The TAM/SAM/SOM Recipe (this discipline's signature dish):

~~~ TAM: Establishment counts for [MARKET] (Census/NAICS, Eurostat/NACE, or national equivalent for [GEOGRAPHY]) × employment/spend benchmarks (BLS, Eurostat, trade associations) (validate against two independent analyst reports; if they disagree by 3x, say so)

SAM: TAM filtered by your actual constraints: [GEOGRAPHY], segment, compliance requirements, tech prerequisites (see TECHINT technographics), local-content and vendor-registration eligibility where applicable

SOM: SAM × realistic capture rate derived from [TARGET] public filings via FININT (their revenue ÷ their claimed customer count = deal size reality check) ~~~

Feeds: TAM/SAM/SOM (the backbone — see tam-sam-som-calculator), ICP definition, personas, messaging localization, market entry prioritization, pricing corridor validation.

4. TECHINT — The Patent Examiner

R&D leaves fingerprints 12-18 months before products ship.

Signal → inference chains:

• Patent clusters (5+ filings in one classification in 12 months) → committed bet on [CAPABILITY], not exploration
• Inventor names repeating across filings → the actual product team behind the initiative; track their conference talks and LinkedIn
• Trademark filing for a product-sounding name → launch inside 6-12 months (trademarks are cheap; companies file close to launch)
• [TARGET] repeatedly appearing in funded consortia → their long-range bet, 12-48 months of lead time
• Research pilot sites → likely launch customers, named in public deliverables
• [TARGET] chairing a standards committee → they intend to shape the rules of [MARKET], not just play by them
• [TARGET]-affiliated authors publishing preprints → R&D direction 6-24 months before patents; a paper cluster plus a hiring surge in the same specialty is one of the strongest fusion pairs available
• Author affiliations shifting from university to [TARGET] on successive papers → they hired the lab, not just the idea
• API docs adding endpoints for an unreleased capability → beta program running now
• Your prospects' tech stacks (technographics) → your SAM refinement: who can actually buy you

Feeds: roadmap bets (where to accelerate vs. concede), SAM refinement, battle cards (feature-gap countdown clocks), build/buy/partner decisions.

5. HUMINT — The Sports Scout

Organizations announce strategy through job boards long before press releases. People are the tell.

Signal → inference chains:

• Hiring surge in one specialty (30+ postings in a quarter) → building [CAPABILITY], not a feature
• Regional specialist roles appearing for a [GEOGRAPHY] you haven't seen them in → expansion pre-announcement
• Job posts naming specific technologies → confirmed stack choices → integration roadmap intel
• Senior product/tech leader exits within 6 months of a strategy announcement → the strategy is in trouble
• Your own alumni landing at [TARGET] → assume they know your playbook
• Employee reviews mentioning "pivot," "reorg," "leadership churn" → 2 quarters of internal distraction = your window
• Win/Loss interviews: the only source that tells you why deals actually close (everything else is inference)

Feeds: roadmap bets (their build signals), battle cards (org instability plays), win/loss program (ground truth for everything).

6. SIGINT — The Wiretap You're Allowed to Have

Companies broadcast constantly through what they change on the public internet. Most competitors never listen.

Signal → inference chains:

• New subdomain SSL cert (e.g., [capability].[target].com) → product launch staging, often weeks ahead
• Pricing page removes a tier → packaging overhaul, usually toward enterprise
• Sudden SEM bidding on your brand terms → they consider you the threat now (congratulations)
• Case study page pattern shifts (new vertical or [GEOGRAPHY] appearing) → segment push
• Messaging A/B visible via Wayback diffs → they're unsure of positioning; hit the wound

Feeds: battle cards (the freshest layer — this is what keeps cards from going stale), pricing strategy, positioning counter-moves.

7. MASINT — The Satellite Photo

Measure the physical and operational exhaust. Abnormal resource allocation never lies.

Signal → inference chains:

• 20%+ volume change in critical inputs → pre-launch or demand collapse (check which via FININT)
• New supplier geographies or country-of-origin shifts → market entry, tariff hedging, or resilience play
• Compliance certification "in process" listings → 12-36 month runway into a regulated segment, visible to anyone who checks the registry
• Product recalls or repeated safety-alert patterns → quality strain; battle card ammunition with a public citation
• Land allocation, power/water capacity reservations, or engineering-design contracts preceding equipment procurement → facility buildout 6-36 months before any launch announcement
• Support response times stretching + hiring freeze in support roles → cash constraint or overwhelmed by growth (disambiguate via employee sentiment)
• Office consolidations → cost compression; expect pricing aggression to follow

Note: supply chain and facility signals are strongest for hardware and industrial players. The software equivalent is ops capacity plus infrastructure-scale language in job postings.

Feeds: threat assessment, launch prediction and capacity estimates, battle cards (capacity-stretch objections: "ask them about their support SLAs lately").

8. All-Source Fusion — The Situation Room