Estimate Layoff Risk

Estimate Layoff Risk

Treat job-loss risk as a causal resource-allocation problem, not sentiment analysis.

Use this chain:

business model -> profit pools -> cash/margin pressure -> allocation decision -> business unit -> critical process -> role family -> employment mechanism -> individual exposure

Non-negotiable boundaries

1. Define the outcome before estimating it. Separate mass layoff, unit closure, role redundancy, performance termination, site closure, contract non-renewal, internship ending, and redeployment.
2. Use only public, legally accessible external information plus facts the candidate voluntarily provides. Never request confidential documents, credentials, private coworker data, or access-controlled material.
3. Do not ask for or use protected or highly sensitive traits such as race, ethnicity, sex, pregnancy, disability, health, religion, political affiliation, union membership, family status, or age except a legally necessary age-of-majority check. Do not use proxies for them.
4. Do not infer a private person's performance, compensation, health, relationships, or legal status. Treat self-reported performance evidence as optional and uncertain.
5. Distinguish observed_fact, candidate_report, derived_metric, model_assumption, and inference in the evidence ledger.
6. Give probability intervals and confidence. If calibration evidence is insufficient, report an ordinal risk band with a scenario range, not a pseudo-statistical percentage.
7. Analyze leadership through observable allocation behavior and governance incentives, never personality diagnosis or cultural stereotype.
8. This is decision support, not a guarantee or legal opinion. Search current jurisdiction-specific law when legal consequences matter.

Default deliverables

Create a case directory containing:

• case.json - scope, entity, role, employment type, location, horizons;
• candidate_answers.json - voluntary answers with date, provenance, and confidence;
• public_evidence.jsonl - claim-level evidence;
• source_manifest.csv - every source searched or used;
• profit_pools.json - segment/product/channel economics and strategic-option assessment;
• dependency_graph.json - profit pool to process to team to role mapping;
• model.json - risk-layer assumptions and sensitivity;
• report.json - structured report source;
• report.pdf - default user-facing report.

Initialize with:

python scripts/init_case.py --company "Company" --role "Role" --country "Country" --employment-type employee --out runs/company-role-YYYY-MM-DD

Workflow

Phase 0 - Define scope and competing outcomes

Identify the legal employer, parent/subsidiary, business unit, product, geography, worksite, role family, employment type, reference date, and 3/6/12-month horizons.

Treat these as competing risks:

• enterprise-wide workforce action;
• BU/product/site closure or material contraction;
• process redesign, outsourcing, automation, or location transfer;
• role-family reduction;
• person selection within a reduced role family;
• performance process;
• contract, internship, probation, or vendor non-renewal;
• redeployment instead of separation.

Do not equate non-renewal with a statutory layoff. Report them separately and optionally combine them as any_involuntary_exit.

Phase 1 - Run the adaptive interview

Read references/interview_protocol.md and references/role_playbooks.md.

Ask the 12 core questions in two short rounds, then only the relevant employment-type and role-family branches. Ask 12-25 questions total in normal cases. Ask follow-ups only when they can materially change the risk interval.

Generate a question set when useful:

python scripts/questionnaire.py case.json -o questionnaire.json

For every answer record the answer, observation date, provenance, candidate confidence, and whether the candidate consents to include it in the report.

Prefer objective signals over anxiety labels. Ask what changed in budget, roadmap, workload, hiring, reporting lines, customer commitments, and future tasks.

Phase 2 - Route the company and economic model

Read references/profit_pool_model.md.

Classify the company before searching: public/private, startup, founder-controlled, diversified, PE-owned/highly leveraged, manufacturing/hardware, software/platform, project/services, regulated finance/utility, biotech/long-cycle R&D, or public/nonprofit/state-owned.

Select the correct economic denominator. Prefer contribution profit, gross profit, cash generation, backlog, utilization, plant load, funding runway, or regulatory mandate over raw revenue when appropriate.

Phase 3 - Search in eight evidence passes

Read references/research_and_evidence.md and references/jurisdiction_playbooks.md.

Generate the initial plan:

python scripts/search_plan.py case.json -o search_plan.json

Run distinct passes for:

1. entity, legal employer, parent, subsidiaries, BU, product, and worksite;
2. solvency, cash runway, debt, covenants, financing, and working capital;
3. company/segment revenue, gross profit, contribution, operating profit, and guidance;
4. margin targets, restructuring charges, synergy commitments, and capital allocation;
5. BU/product lifecycle, customers, orders, roadmap, launches, cancellations, and strategic priority;
6. workforce, payroll, job postings, hiring freezes, location moves, outsourcing, and automation;
7. governance and leadership allocation history;
8. contrary evidence and comparable-company base rates.

Prefer primary sources. Deduplicate syndicated stories by underlying claim and origin. Record unavailable but decision-relevant sources.

Phase 4 - Reconstruct profit pools and cost pressure

Build a profit-pool table by material product x geography x channel x customer type x lifecycle combinations.

For each pool estimate ranges for revenue, gross/contribution profit, cash conversion, growth, orders/retention, inventory, customer concentration, incremental investment, avoidable cost, shared-platform dependency, strategic option, regulatory value, and management commitment.

Classify the action regime as survival, margin_reset, capital_reallocation, integration, demand_capacity, performance_system, or mixed.

Do not assume a currently loss-making business is disposable. Estimate future contribution, option value, synergy, required follow-on investment, and probability management still funds the next milestone.