# Account Research for Outreach

Research before outreach has one job: find a true, specific reason this account should care right now, and the person who owns that problem. Dossiers on companies you compete with belong to the competitor-profiling skill.

## How much research each tier gets

| Tier | Time per account | Output |
|------|------------------|--------|
| Tier 1 | 10–20 minutes, or an agent run reviewed by a human | Full brief (below) |
| Tier 2 | 2–3 minutes per account, plus shared segment research | One sourced hook per contact + the segment brief |
| Tier 3 | None per account | The segment brief only |

Tiers are defined in [signal-plays.md](signal-plays.md).

## Where the useful facts are

In rough order of usefulness:
1. **Their job posts.** They describe the problem in their own words, list the tools they use, and reveal priorities ("own our experimentation roadmap").
2. **Their own pages.** Pricing, landing pages, changelog, careers, case studies. For a services pitch, the page you'd improve is the research.
3. **What leaders say publicly.** LinkedIn posts, podcast appearances, conference talks, interviews.
4. **Company news.** Funding, launches, acquisitions, leadership changes, new markets.
5. **Filings and reports.** For public companies, the 10-K risk factors and earnings calls name priorities explicitly.
6. **Reviews of their product.** What their customers complain about can show what they're under pressure to fix.

## The brief (Tier 1)

```markdown
## [Company] — account brief
**Fit:** [ICP criteria met, with any gaps]
**Signal:** [what happened, date, source URL]
**Likely priority:** [the problem the signal implies, in their words where possible]
**Evidence:** [2–4 facts, each with a source URL]
**Who owns it:** [name, title, why them; a second contact if multi-threading]
**Angle:** [one sentence connecting their situation to what you offer]
**Proof to use:** [the closest case study or result]
**Don't mention:** [anything sensitive, stale, or creepy, e.g. tracked website visits]
**Hook found:** yes / no
```

## Rules for agent research

AI makes research cheap, and it also makes invented hooks cheap. These rules keep the research honest:

1. **Facts first, angle second.** Run research as two steps. Step one collects facts, each with a source URL and date. Step two writes the angle using only those facts. Never let one step do both.
2. **"No hook found" is a valid result.** If the sources don't support a specific angle, say so and drop the account to the segment-level message. A generic hook dressed up as personal research is worse than an honest segment message.
3. **Bound the angles.** Give the agent your 3–5 value propositions and require the angle to map to one of them. Unbounded creativity produces promises you can't keep.
4. **Date everything.** A fact without a date can't be judged for staleness. Discard anything older than its freshness window.
5. **Quote, don't paraphrase, when it matters.** If the hook rests on something a leader said, keep the exact quote and link.
6. **Never use sensitive traits** (health, finances, politics, religion, family) even when public. See [compliance.md](compliance.md).

## Segment briefs (Tier 2 and 3)

For a segment, research the shared situation instead of each account:
- What the accounts have in common (the segment definition)
- The pain that commonality creates, with 2–3 public examples
- The proof that speaks to it
- The one-line opener each contact will get, plus the slot for a per-contact detail (Tier 2)

A tight segment does the personalization work. "Series A SaaS companies that just posted their first growth marketer role" needs little per-account research, because the segment already explains why you're writing.