# Lifesight Forecast Scenario

> Use when the user wants to project outcomes or compare paths forward — "forecast my revenue", "what will next quarter look like", "what budget do I need to hit $4M", "compare current vs +15% spend", "what if I shift budget to TikTok", "model this scenario", "target KPI planning", "project the impact before I commit". Routed to from the `lifesight` router after the workspace is calibrated.

- Skill: `lifesight/lifesight-forecast-scenario` (Agent Skill)
- Install (CLI): `npx skillmds@latest add lifesight/lifesight-forecast-scenario`
- Raw SKILL.md: https://api.skillmd.com/api/skills/lifesight/lifesight-forecast-scenario/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: lifesight (https://skillmd.com/u/lifesight)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/lifesight/lifesight-forecast-scenario

---


# Lifesight Forecast & Scenario Planning

Help the user see forward — a revenue projection, the budget needed to hit a target,
or a head-to-head between options — before they commit real money. **A forecast is a
projection with a range, never a promise.** The discipline here is honesty about
uncertainty and comparing options on the right axis.

**Prerequisites (router handles):** workspace calibrated, profile loaded. Operate
under `lifesight-core`; present under `lifesight-rendering`. Load both.

## The three shapes this takes

1. **Forecast** — "what will revenue be?" Project forward from a stated allocation.
2. **Target-KPI planning** — "what budget hits $X revenue?" Solve backward from a goal.
3. **Scenario comparison** — "current vs +15% vs my custom plan." Put options side by side.

Identify which one the user is asking before pulling anything.

## Flow

1. **Confirm the basis.** A forecast is meaningless without its assumptions: which
   allocation (current, or an optimized plan?), what budget, what period, which model.
   Lock these first (from the profile + one clarifying question if needed).
2. **One heavy call** via `ask_mia` (forecast / scenario / target-KPI). Walk its gates
   (`lifesight-core` Rule 4). One analysis per turn.
3. **Frame the result as a range with assumptions**, not a point promise.

## Judgment checks (mandatory)

- **Always a range + assumptions.** State the confidence range and the assumptions it
  rests on (budget, period, model, which scenario). A single number with no range is
  false precision — and dangerous if someone treats it as a commitment.
- **Compare scenarios on the right axis.** The higher-revenue scenario is NOT
  automatically the better one. Show incremental revenue **and** efficiency (iROAS) and
  the spend each requires. A target that demands value-destroying spend (lots more
  budget at sub-1.0 marginal return) must be called out — same discipline as
  budget-optimization Step 4.
- **Target-KPI realism.** "What hits $4M?" may have an answer that's inefficient or
  infeasible within guardrails. Give the budget *and* the efficiency cost of getting
  there; flag if the target requires spending past saturation.
- **Platform vs causal, and recency.** Name whether projections are on causal
  (incremental) outcomes; note data recency — a forecast off stale data inherits the staleness.

## Output shape

1. **The projection** — the number(s) with a confidence range, lead with the decision.
2. **The assumptions** — budget, period, allocation, model, in one tight block.
3. **Scenario table (if comparing)** — each option's spend, projected incremental
   revenue, and efficiency, so the trade-off is visible — not just top-line.
4. **The read** — which path is actually best and why (efficiency, not just size), and
   what would de-risk it.

## Next steps to offer

"Optimize the allocation to hit this" (→ budget-optimization) · "Validate the key
assumption with a test" (→ experiment-design) · "Build the board/CFO version" (→
board-briefing / cfo-translation).

## Red flags — STOP

- Presenting a forecast as a single certain number → range + assumptions
- Picking the scenario with the biggest top-line without showing efficiency/spend
- Answering "what hits $X" without flagging if it needs spend past saturation
- Treating a projection as a commitment the user can be held to
- Forecasting off stale data without saying so