# Impact Excavation

Use this when a bullet lacks ownership, consequence, scale, or a
target-recognizable result.

## Contents

1. Hypothesis pass
2. Guided recall
3. Industry-recognition filter
4. Evidence and causality
5. Prompt design

## 1. Hypothesis pass

Do not begin with `What was the impact?` First reason through:

- the source company's product, customer, revenue model, and stage
- the operating problem at that moment
- the artifact, analysis, or decision Darsh owned
- the downstream team or leader who acted
- the likely causal chain

Map:

```text
insight or artifact -> decision -> operating change -> leading indicator ->
business outcome
```

Mark each link as confirmed or hypothetical.

Generate 2-4 plausible impact directions. For each, determine:

- why it follows from this specific mechanism
- what Darsh may have observed or heard
- what metric, proxy, document, or decision might survive
- whether the target reader will recognize its significance

Rank hypotheses by contextual likelihood, target value, and evidence
recoverability—not by how impressive they sound.

## 2. Guided recall

Ask one manageable question at a time:

1. Reconstruct what happened immediately after the work.
2. Name the strongest context-specific hypotheses and briefly explain why they
   follow.
3. Ask which concrete change the user remembers seeing or hearing.
4. Follow that thread into baseline, new state, time period, and magnitude.
5. Test whether the work produced the result, informed the decision that
   produced it, or merely coincided with it.
6. If no metric exists, recover an adopted decision, risk avoided, resource
   reallocation, operating cadence, or counterfactual.

Offer a defensible range, directional change, or scale proxy when an exact
number is unavailable. Label estimates as user-provided.

## 3. Industry-recognition filter

Continuously check whether the proposed impact is:

- a recognized success measure in the source function
- valued in the target industry and role
- appropriate for the target company stage
- immediately legible to the recruiter, hiring manager, or executive

Use three levels:

- **Broad business signal:** revenue, margin, cost, risk, growth, adoption,
  capacity, or speed.
- **Recognized functional signal:** quota attainment, attainment dispersion,
  forecast error, renewal rate, pipeline conversion, productivity ramp, or
  sales capacity.
- **Internal-only signal:** a proprietary score, product label, taxonomy, or
  process count whose significance is unclear outside the company.

Translate internal-only measures into the accurate decision or consequence
they represent. Do not substitute a fashionable target-industry metric for the
result that actually occurred.

When norms are uncertain or market-dependent, inspect the current JD, company
materials, and credible current functional sources before coaching.

Tell the user why a direction matters. Example:

> For a Sales Finance reader, attainment dispersion is a recognized signal of
> quota calibration and coverage quality. If that spread changed after your
> work, it will travel better than an internal customer label.

## 4. Evidence and causality

Recover:

- actual deliverable
- user or decision maker
- scale
- baseline and new state
- time period
- first observable result
- attribution strength
- counterfactual

Never infer that a suggested result occurred. Never insert a prompted metric
until the user confirms its meaning. Clarify ambiguous percentages with the
underlying before-and-after figures.

Use the strongest accurate causal verb:

- `produced` only for direct causality
- `drove` when the work materially caused the change
- `informed` when a leader made the downstream decision
- `coincided with` only as context, not claimed impact

## 5. Prompt design

Use known details and the strongest business hypothesis:

> Because the analysis changed both regional coverage and quotas, I would first
> look for better capacity deployment or narrower attainment dispersion. In
> the next planning cycle, what did leaders observe becoming better?

Avoid generic KPI menus. If none of the hypotheses resonates, reconstruct the
sequence instead of cycling through more outcome labels.

Do not draft the final bullet from vague answers. Return to `bullet-coaching.md`
once ownership, mechanism, and consequence are sufficiently supported.