ICP Modeling

Layer three categories of data onto each account.

- Run all accounts through Clay to enrich:

- Firmographics: industry, headcount, funding, location, revenue

- Technographics: tech stack via BuiltWith

- Account-fit signals: hiring activity, web traffic, job postings, growth indicators

‍

Find patterns across wins and losses.

- Feed the dataset into Claude Code

- Use the Freckle Claude Code skill, which automates this step

- Identify:

- Shared traits across wins

- Shared traits across losses

- Negative indicators to exclude

‍

Set the baseline for accounts worth any outreach.

- Tier 3 acts as a broad filter rather than a target list

- Example output:

- Type: B2B SaaS or adjacent

- Funding: Any VC-backed round

- Location: US, UK, EU, Canada, Australia, Israel

- Headcount: 1–10 or 201–500

‍

Use your top closed won accounts to build Tier 1 criteria.

- Hand-pick the 150 best-fit closed won accounts

- Look at: highest ACV, fastest close, strongest product fit

- Build Tier 1 from what they share. Example output:

- Type: B2B SaaS

- Funding: Series A

- Location: US, UK, or EU

- Headcount: 20–150

- Industry: Marketing-tech / Sales-tech

- Tech: HubSpot

- Account fit: Sales team 5+

‍

Fill the gap between Tier 1 and Tier 3.

- Broader on location, funding, and headcount

- Example output:

- Type: B2B SaaS

- Funding: Seed – Series B

- Location: US, UK, EU, Canada, Australia, Israel

- Headcount: 11–200

- Account fit: Sales team 2+

- Tier 1 accounts go to manual outreach and Tier 2 accounts go into automated sequences

‍

Validate the scoring against historical data before rolling it out.

- Run the scoring across closed won and closed lost in Clay

- Check whether wins cluster in Tier 1 and losses cluster in Tier 3

- Pull sample lists to see how the scoring performs on fresh data using: Apollo, AI Ark, Ocean.io, Discolike and Clay

- Adjust criteria if the sample data doesn't match the historical patterns

‍