lp-forecast — Startup 90-Day Forecaster

Build P10/P50/P90 scenario bands and first-14-day validation plan for pre-launch or newly launched businesses with zero historical performance data.

Invocation

/lp-forecast --business <BIZ> [--horizon-days 90] [--launch-surface pre-website|website-live]

• --business: Business name (BRIK, intshop, etc.)

Business resolution pre-flight: If --business is absent or the directory docs/business-os/strategy/<BIZ>/ does not exist, apply _shared/business-resolution.md before any other step.

• --horizon-days: Forecast horizon in days (default: 90)
• --launch-surface: Stage of launch — pre-website (MVP planning) or website-live (post-launch tracking)

Operating Mode

READ + RESEARCH + FORECAST

• Read existing business context, offer hypothesis (from lp-offer/MARKET-06), and available market data
• Research competitor benchmarks, channel performance ranges, and market size signals
• Forecast P10/P50/P90 scenario bands with unit economics assumptions
• Generate first-14-day validation plan (THE primary deliverable)

Differs from lp-do-idea-forecast

This skill is NOT a rename or copy of lp-do-idea-forecast. Key differences:

1. Works from zero operational data: lp-do-idea-forecast requires historical performance data or existing market intelligence packs. lp-forecast builds forecasts from competitor benchmarks and channel ranges only.

2. No market intelligence pack prerequisite: lp-do-idea-forecast blocks when latest.user.md market research is stale and requires Deep Research prompt. lp-forecast proceeds with available evidence (competitor pricing, channel benchmarks) and marks confidence accordingly.

3. Simpler output contract: lp-do-idea-forecast delivers a full 12-section Output Contract with detailed channel plans and financial models. lp-forecast delivers P10/P50/P90 bands, unit economics assumptions, and a first-14-day validation plan.

4. No Deep Research gate: lp-do-idea-forecast stops and requires manual Deep Research when data is insufficient. lp-forecast proceeds with available evidence and tags assumptions as low-confidence when data is sparse.

5. Validation-first mindset: Both include first-14-day validation plans, but lp-forecast makes this THE central output — the forecast is a hypothesis to validate, not a plan to execute.

Inputs

Reads from:

• Business context: docs/business-os/startup-baselines/<BIZ>/
• Offer hypothesis: Output from lp-offer (MARKET-06) — docs/business-os/startup-baselines/<BIZ>/offer.md
• Channel selection (optional if available): Output from lp-channels (SELL-01) — docs/business-os/startup-baselines/<BIZ>/channels.md
• Market data: Any available competitor pricing, channel benchmarks, market size signals (opportunistic)

Artifact registry: Canonical producer paths are defined in docs/business-os/startup-loop/artifact-registry.md.

No formal market intelligence pack required. If none exists, proceed with web research.

Workflow

Stage 1: Gather Available Evidence

• Read offer hypothesis (pricing, target customer, value prop)
• Read channel selection (which channels, why chosen)
• Web research: competitor pricing, channel benchmarks (e.g., Google Ads CPC for niche, email open rates for vertical)
• Market size signals: search volume, competitor traffic estimates, industry reports (opportunistic)
• Tag confidence: high (multiple sources), medium (single source), low (extrapolated/assumed)

Stage 2: Build Scenario Forecast

Create P10/P50/P90 scenario bands for:

• Revenue: 90-day cumulative revenue range
• Orders: 90-day order count range
• Traffic: 90-day unique visitor range (if website-live)
• Leads: 90-day lead count range (if pre-website)

Scenarios:

• P10 (pessimistic): Low conversion, high CAC, slow ramp
• P50 (base case): Channel benchmark conversion, median CAC, steady ramp
• P90 (optimistic): High conversion, low CAC, fast ramp

Stage 3: Define Unit Economics Assumptions

For each scenario (P10/P50/P90):

• CAC: Customer acquisition cost per channel
• AOV: Average order value
• Margin: Gross margin per order
• Conversion rate: Visitor-to-order or lead-to-order
• Channel mix: Traffic/spend distribution across selected channels

Stage 4: Create First-14-Day Validation Plan

THE key output. Define:

• Metrics to track: Which KPIs confirm or refute the forecast
• Measurement cadence: Daily/weekly/milestone-based
• Decision gates: When to pivot, pause, or accelerate
• Validation thresholds: What metrics must hit what values by day 7, day 14
• Data sources: Where to pull each metric (analytics, CRM, ads dashboard)

Example decision gate: "If Day 7 CAC exceeds P90 CAC by 50%, pause paid channels and pivot to organic."

Stage 5: Persist Output

Write forecast to:

docs/business-os/startup-baselines/<BIZ>/S3-forecast/YYYY-MM-DD-lp-forecast.user.md

Output format:

• Scenario summary table (P10/P50/P90 bands)
• Unit economics assumptions per scenario
• Channel-specific ranges (per channel selected in lp-channels)
• First-14-day validation plan (metrics, thresholds, decision gates)
• Assumption register with confidence tags
• Source list with URLs

Output Contract

File: docs/business-os/startup-baselines/<BIZ>/S3-forecast/YYYY-MM-DD-lp-forecast.user.md

Artifact registry: Canonical path defined in docs/business-os/startup-loop/artifact-registry.md (artifact ID: forecast).

Required sections:

1. Scenario Summary: P10/P50/P90 bands for revenue, orders, traffic/leads
2. Unit Economics: CAC, AOV, margin, conversion rate per scenario
3. Channel Ranges: Traffic, spend, CAC per channel (from lp-channels)
4. First-14-Day Validation Plan: Metrics, thresholds, decision gates, data sources
5. Assumption Register: All assumptions with these mandatory fields:
   • assumption_id — stable ID for tracking across review cycles
   • assumption_statement — plain-language claim being made
   • prior_range — plausible range before evidence
   • sensitivity — impact on CAC/CVR/revenue if assumption is wrong (Low/Medium/High)
   • evidence_source — URL or artifact reference
   • confidence_level — low (<60) / medium (60–79) / high (≥80)
   • kill_trigger — observable state that invalidates this assumption (e.g., "Day 7 CAC > 2× P90 estimate")
   • owner — named person responsible for monitoring and re-checking
   • next_review_date — ISO date for next assumption review
6. Source List: URLs for competitor data, channel benchmarks, market signals

Quality Checks

Self-audit before delivery:

• All three scenarios (P10/P50/P90) present with numeric ranges
• Unit economics defined for each scenario
• Channel-specific ranges match channels selected in lp-channels
• First-14-day validation plan includes metrics, thresholds, and decision gates
• Assumption register includes all 9 mandatory fields for every assumption (assumption_id, assumption_statement, prior_range, sensitivity, evidence_source, confidence_level, kill_trigger, owner, next_review_date)
• Confidence tier declared (low/medium/high) with corresponding spend cap, time cap, cadence, and allowed decision class
• Source list includes URLs for all claims (no unsourced assertions)
• Output file persisted to docs/business-os/startup-baselines/<BIZ>/S3-forecast/

Red Flags

Invalid outputs that require rework:

• Single-point forecast (no P10/P50/P90 bands)
• Missing first-14-day validation plan
• Unsourced claims (no URLs in source list)
• Unit economics missing for any scenario
• Channel ranges don't match lp-channels output
• Assumption register missing kill_trigger, owner, or next_review_date for any assumption
• Confidence tier not declared or spend cap / time cap not stated for the tier
• Validation plan has no decision gates

Integration

• Consumes: lp-offer output (MARKET-06); may consume lp-channels output (SELL-01) when available
• Feeds into: lp-prioritize (S4 prioritization), startup-loop S4 baseline (first-14-day tracking)
• Trigger: Called automatically by startup-loop at S3 stage, or manually via /lp-forecast

Forecast Guardrails (Confidence Tier Policy)

Apply the correct tier based on the overall confidence level of the forecast. Confidence level = the lowest confidence across all P50 scenario assumptions.

Guardrail breach handling:

• If actual spend is at or approaching the cap before the re-check cadence is due, pause and run an ad-hoc review — do not extend the cap unilaterally.
• If confidence improves at re-check (new evidence), tier may be upgraded. Document the upgrade with evidence source.
• Unknown or missing confidence values default to low tier — fail closed.

Reference: docs/plans/startup-loop-marketing-sales-capability-gap-audit/fact-find.md (Sparse-evidence forecast control surface)

Launch Surface Modes

pre-website

• Focus on lead generation metrics (email signups, waitlist, pre-orders)
• Traffic forecast not required (no website yet)
• CAC based on pre-launch channels (social, partnerships, waitlist ads)

website-live

• Focus on traffic, conversion, orders
• CAC based on post-launch channels (SEO, paid ads, email, partnerships)
• Include visitor-to-order funnel metrics

Notes

• This forecast is a hypothesis to validate, not a prediction to execute
• Low-confidence assumptions are acceptable — apply the low confidence tier guardrail (10% spend cap, ≤5 hrs/week/channel, Continue/Investigate only). See ## Forecast Guardrails above.
• First-14-day validation plan is THE critical output (the forecast exists to be tested)
• If evidence is sparse, proceed with low-confidence tags and wide P10-P90 bands
• No Deep Research gate — work with what's available and mark uncertainty