account-researcher

Account Researcher

Overview

This skill produces a brief of one to two pages per account: a snapshot (size, revenue, HQ, keywords, team counts), dated trigger events, the technologies that decide which pitch to use, the people who own the decision with verified emails, a hypothesis with talking points that each cite a source, a contact plan, and the cost of the research.

The mistake it prevents is research that costs an hour and changes nothing: ten tabs of trivia, a "tech stack" pasted from a profile, and an email that opens with the company's founding year. Every item in the brief has to change who gets contacted, what gets said, or when. Paid lookups follow the same rule. Tech Stack Search costs $4.99 per technology, so it runs only for a technology whose presence or absence flips the pitch.

When to Use

• "Brief me on harborlinelogistics.com before Thursday's call."
• "Research our 10 tier 1 accounts before the AEs start."
• "What do we know about Acme, and who should I email there?"
• "Account plan for this enterprise target."
• A reply came in from a good account and the rep wants context fast.

Use a different skill when:

Prerequisites

Depth tiers

The user sets the time box for each tier.

Authoritative Foundations

Source details and the routing table: references/framework-notes.md.

Workflow

1. Scope the brief

Write one line per account: domain, what the user sells, purpose, depth tier, budget. List the 1-3 technologies whose presence would change the pitch (step 4). If the user names more than 10 accounts, switch to decision-maker-finder for the people and run this skill on the top 10 only.

2. Snapshot from the Lead Database

Call query_lead_database with companyDomain and limit: 1. The returned row carries company fields (industry, employee count, revenue, website, HQ city, state and country, keywords), and pagination.total is the number of people the database holds for the domain. Then size the target teams with more limit: 1 counts, one per department or title family (for example sales titles, engineering titles). These counts cost nothing and tell the rep how big the team they are selling into is.

3. Key people

Pick the roles from the product category (the role map in decision-maker-finder lists 22 categories). Query the Lead Database with companyDomain, the role's title list, seniority and the universal notTitle set, limit: 25. Rank by title match, seniority, email present. For an empty role, run scrape_website_finder on the domain with one jobTitle. For a name without an email, run find_emails. Validate every email that did not come from Email Finder with validate_emails. Keep the economic buyer, champion, evaluator and a backup.

For each person, add one person-level detail if it exists within the time box: start date in role, a post, a talk, a podcast. Cite it.

4. Tech stack: only what changes the pitch

Run the decision table in references/tech-stack-decision-guide.md:

Tech Stack Search (scrape_builtwith) takes one technology and returns every domain detected using it. Search the downloaded CSV for the account's domain. Keep the file: it answers the same question for every other account at no extra cost and seeds ecommerce-by-techstack or a TAM build.

5. Why now

Start with the seller's own history, which is free and the strongest opener: past meetings, demos, a closed-lost reason and date, and whether a champion from a current customer now works at the account (from the CRM, or a Sales Navigator lead list of customer champions filtered by Changed jobs). Then collect dated signals from primary sources: the company's newsroom and careers page, EDGAR for public companies, the executive's own posts or interviews, then press coverage. Sort each into Elias's three types (bad experience, transition, awareness). Signals older than the cutoff the user sets go to background. Job postings count as transitions when they describe a new team, a new tool or a rebuild.

6. Hypothesis and talking points

Write a two-sentence hypothesis: the likely pain and why this quarter. Then the talking points, each tied to a cited source. A talking point that cannot cite a source is cut. Map them to MEDDIC: which metric the buyer is judged on, who signs, who champions. Keep the phrases the buyer used in public (a post, a podcast, their own job posting) and write the talking points in those words.

7. Contact plan

Check for a warm path first: if the user's team has Sales Navigator, look up the economic buyer with TeamLink or the "Connections of" filter; a colleague who knows them makes Day 1 an introduction request instead of a cold email (LinkedIn's Sales Solutions blog says buyers are five times more likely to engage when outreach comes through a mutual connection). Then order the people, pick the channel and day for each, and the angle. Never email two people at the same account at once; Berman's guidance is to start at the top and move down a level after one to two weeks without a reply, since several people at once can get the sender blocked company wide. Add a mobile only for a planned call, after checking the do-not-contact list.

8. Write and check

Fill templates/output-template.md, list every source with a URL and the date checked, and run scripts/check-output.py.

ScraperCity Implementation

Calls

Snapshot and team count:

{"companyDomain": "harborlinelogistics.example", "limit": 1}
{"companyDomain": "harborlinelogistics.example", "title": "Account Executive, Sales, Carrier Sales, Business Development", "notTitle": "assistant, intern, former", "limit": 1}

People for one role:

{"companyDomain": "harborlinelogistics.example",
 "title": "Chief Revenue Officer, CRO, VP Sales, Director of Sales Operations, Revenue Operations, Salesforce Administrator",
 "notTitle": "assistant, intern, former, retired, advisor",
 "limit": 25}

Missing role:

{"domains": ["harborlinelogistics.example"], "jobTitle": "Salesforce Administrator", "autoValidateEmails": false}

Email for a known name: {"contacts": [{"first_name": "Priya", "last_name": "Natarajan", "domain": "harborlinelogistics.example"}]}

Tech Stack Search, MCP: {"technology": "HubSpot", "fileName": "techstack-hubspot-20261013"}. CLI: npx scrapercity builtwith "HubSpot". REST: POST /api/v1/scrape/techstack. Poll check_run_status, then download_results, then search the CSV for the account's registrable domain.

Worked cost estimate

One Standard brief, two technologies that flip the pitch (Salesforce and HubSpot), four people:

Ten tier 1 accounts checked against the same two technology lists cost $9.98 for Tech Stack Search in total, not $99.80, so the per-account cost falls to about $1.11. Call check_wallet, show this table and wait for a yes before any paid lookup.

Error handling

Output Format

1. account-brief-[domain]-[date].md following templates/output-template.md: Account Snapshot, Why Now, Tech Stack, Key People, Hypothesis and Talking Points, Contact Plan, Sources, Cost Estimate, Open Questions and Next Steps.
2. account-people-[domain]-[date].csv: the people table with the person columns from references/list-quality-standard.md section 2a plus buying_role and found_by.
3. Saved Tech Stack Search CSVs named techstack-[technology]-[date].csv for reuse.

What the reader receives is the deliverable only: no internal file paths, reference section numbers, tool ids or skill names (name the product, such as Lead Database, Email Validator or Website Finder, and describe the next action instead of naming a skill); no sales-framework acronyms (keep the content: who signs, who champions); each figure stated once; every outside number with its source and date on first mention; no value left as "you set" when a sourced default or a dated step can carry it; no table of checks that were not run (one "Not checked:" line instead); no repeated identical rows where one template plus a list will do. Next steps are a table with Step, Owner, Date and Done when.

Quality Check

• Every talking point and every Why Now signal cites a source ID that appears in Sources with a URL and a date checked.
• Every signal has a date; anything older than the user's cutoff is marked background.
• Every Tech Stack row says whether it changes the pitch; rows that do not were checked with free signals only.
• Tech Stack Search ran only for technologies marked as changing the pitch, and the cost table shows $4.99 per technology.
• Key People names the economic buyer and champion (plus evaluator and backup when found), each with a buying role, an email status and a found_by tool; only valid emails are marked ready to send.
• The contact plan never puts two people at the account on the same day.
• The brief fits on two pages.
• scripts/check-output.py brief.md prints PASS.

Common Pitfalls

Execution Artifacts

Related Skills

• decision-maker-finder for the people at many accounts at once.
• icp-list-builder to find more accounts like the one researched.
• ecommerce-by-techstack to turn a saved Tech Stack Search CSV into a list.
• signal-triggered-outbound to build lists from the trigger types in Why Now.
• cold-email-copywriting to write the first email from the talking points.
• multi-channel-outreach to sequence email, phone and LinkedIn across the people.
• crm-enricher to write the people and brief back to the CRM.