Build a target account list from scratch
Go from ICP criteria and no list at all to a complete, deduplicated target account list with firmographics, buyer contacts and fit scores.
Verified against the live platform
1 step rests on platform behaviour that has not been decisively probed. Everything else was checked live.
Where the list comes from
Database and web sourcing against your stated criteria, then hard filters on geography and size before anything is enriched.
The steps
- 1
Orient
Confirm the inputs, compute the expected cost, and tell the user what the run will do before anything spends a credit.
- 2
Source the accounts
Run two sources and merge. ICP Prompt and Companies Search return overlapping-but-different sets; the union beats either alone, and dedup in Step 3 is nearly free.
- 3
Hard filters: geography and size
Do this before enrichment, using whatever the builder already returned. Free filter groups on the imported location and size columns. Where the builder did not return them, defer to Step 4 and filter there.
- 4
Deduplicate and normalize
Clean Company Names (0.1c), Normalize Website URL (0.1c), Clean Location (0.1c).
- 5
Enrich firmographics
Industry Classifier (0.2c), Company Headcount (0.2c), Company Annual Revenue (1c), Company Address (0.5c) or Clean Location (0.1c), Find Target Audience (0.2c, B2B/B2C/Both).
- 6
Classify business model
Business Model Finder (0.2c), SaaS / PaaS / IaaS etc. Website Technology Detector (0.25c) when the ICP mentions a tech stack. Revenue Stream (0.1c) for how they actually monetise. Gate: classification present on rows that matter.
- 7
Remove bad-fit categories
WorkaroundFree filter groups against the exclusions gathered in preconditions. {{COMPETITOR_DOMAINS}} and {{CLIENT_DOMAIN}} per primer §9. Where "bad fit" is a judgement rather than a field: no negative/exclusion classifier.
- 8
Score accounts by ICP fit
Scoring Agent (FREE) with the full ICP from preconditions. Gate: scores spread; reasoning references real criteria.
- 9
Segment
Free filter groups by vertical, size band, or pain point. Segments should map to how the user will actually write copy. Segmenting by something nobody will write differently for is wasted work, ask.
- 10
Run the shared outbound tail
Personas: from the user's ICP. Prefer Find People Agent (3c) here, a from-scratch list spans company sizes, and the right title genuinely differs between a 40-person and a 4,000-person company. Signal column: none yet.
Where the run should stop
- Geography and size are hard filters, applied before enrichment, not scoring criteria.
- Business model classification runs before the fit score, so bad fit categories are removed rather than scored low.
- Segment at the end rather than deleting, so the middle band stays available to nurture.
What you should end up with
A deduplicated account list with firmographics, a business model classification, an ICP fit score and buyer contacts, segmented into working and nurture bands.