Consult market

Consult Market -- Lookalike Target List

After a win, find the next 5 companies in the same industry and size who would buy the same use case. One job: a ranked, sourced target list — not outreach, not a pitch.

Inputs (manual — ask if missing)

Input	Required	Notes
Industry / vertical	yes	e.g. "regional telco", "DTC supplements"
Company size band	yes	revenue, headcount, or both — e.g. "200–2,000 staff"
Use case	yes	one line on what was sold/won — the buyer's reason to care
Region / geography	optional	default: same region as the win
Exclusions	optional	the won client, named competitors, do-not-contact list

No engagement-passport.yaml needed. If the user points at a won proposal pack, use it only to sharpen the use-case one-liner — do not require it.

Data sources (degrade gracefully)

1. timeln-find — analogous engagements / accounts already in memory. Run first; cite titles only.
2. Web + Exa — public firmographics, recent buying signals (funding, hires, product launches, regulatory pressure, RFPs). Every company → source URL.
3. Apollo / Clay — only if connected — firmographic filtering and verified contacts. If absent, skip silently; do not block.

If Timeln MCP is unavailable, note timeln skipped — no MCP and continue with web + Exa only. Never fabricate memory results.

Workflow

1. Confirm the three required inputs; restate the use case in one buyer-centric sentence.
2. timeln-find for accounts/engagements that match the vertical + use case.
3. Web/Exa (and Apollo/Clay if present) to build a candidate pool; filter to the size band and region.
4. Score each candidate 0–3 on four axes — industry match, size match, use-case fit, signal recency — and keep the top 5.
5. For each kept company write a one-line why-it-fits tied to the won use case (not generic firmographics).
6. Add the best-fit contact: name + role + LinkedIn/company URL. Email only if a verified source returns it. Otherwise mark not verified — enrich. Never guess or pattern-build emails.
7. Write the markdown table, then the CSV (always) and offer XLSX via the xlsx skill. Save to the path the user names (default ./lookalikes/{use-case-slug}-targets.md + .csv).

Output -- exactly this shape

## Lookalike targets -- <use case> -- <YYYY-MM-DD>

**Seed:** <industry> · <size band> · <region> · use case: <one line>
**Sources:** timeln-find | web/Exa | Apollo/Clay (if used)

| # | Company | Why it fits (vs use case) | Industry | Size | Buying signal (date + source) | Contact (name · role) | Contact detail | Score /12 |
|---|---------|---------------------------|----------|------|-------------------------------|-----------------------|----------------|-----------|
| 1 | | | | | | | verified / not verified — enrich | |

**Scoring:** industry + size + use-case fit + signal recency, each 0–3.

**Excluded / parked**
- <company> — <reason>

**Next step:** `/cos-pursue {company} for {use case}` to draft outreach (cold-start).

CSV columns mirror the table (one row per company). Confidence = the /12 score.

Rules

• Top 5 only — tight beats wide. Park extras under "Excluded / parked" with a reason.
• Every company carries a source URL or a timeln-find title; no unsourced names.
• Contact emails are verified-source-only. No domain-pattern guessing, ever.
• why-it-fits must reference the won use case, not just "same industry".
• Do not write outreach copy here — hand off to consult-pursue.
• If a signal can't be found for a candidate, keep it but score signal-recency 0 and say no recent signal.

Common failure modes

Mistake	Fix
Generic "same vertical" rationale	Tie why-it-fits to the buyer's reason for the won use case
Guessed emails (first.last@)	not verified — enrich; only verified-source emails go in
20-company dump	Score, keep top 5, park the rest
Listing the won client or its direct competitor	Honor exclusions; default-exclude the won client
Drifting into a pitch	Stop at the list; point to /cos-pursue