TL;DR: Tell Copilot who your best customer is, in one sentence, and it generates a complete ICP — segments, parameters, weights, and parameter modes. Refine from there in the Qualification page or by talking to Copilot.
What an ICP is
An ICP — Ideal Customer Profile — is your company-level targeting definition. It answers "what kind of company should we even bother chasing?". Every record in your workspace gets an ICP code (which ICP it matches) and an ICP score (how well). Audiences, signal qualification, enrichment prioritization, and pipeline reporting all key off these two fields.
A good ICP is sharper than a marketing persona deck — it's executable. Each parameter (industry, employee count, revenue band, tech stack, geography, growth signals) operates in one of three modes:
- Filter — hard gate. Records that fail the filter are not in this ICP, period.
- Score — weighted contribution. Records get points for matching; non-matchers are still in but score lower.
- Exclude — hard removal. Even matching records are tossed if they trip an exclude rule (e.g. competitors).
Copilot understands these modes natively. When it builds an ICP for you, it picks the right mode per parameter — not everything gets crammed into the same bucket.
The Copilot path
From any page, open Copilot. Type a single sentence:
> build me an ICP for mid-market SaaS companies (51–500 employees) with a
security-first culture, primarily North America and EMEA, growth-stage,
excluding competitor accounts
Copilot will:
- Propose a draft ICP — name, segments, the parameter list with modes (Filter / Score / Exclude) and weights
- Show you a preview of how many existing records would match (and which Personas inside the ICP would qualify)
- Wait for your confirmation before saving
Refine in conversation. > drop the geo filter and replace with a score. > add a "uses Snowflake" boost. > exclude companies with fewer than 20 engineering hires in the last 12 months. Copilot rewrites the ICP in place; you approve when it looks right.
Copilot can also build an ICP from your historical performance. Try: > look at my closed-won customers from the last 12 months and propose an ICP that overweights what they have in common. This pulls from the same lead and account tables that power the Insights dashboard — no setup needed.
The manual path
Open Qualification from the left sidebar. You'll see your existing ICPs grouped, with their Personas nested underneath. Click + Add → ICP in the toolbar. The drawer opens on the right and walks you through the parameter list, organized by category (Firmographic, Tech, Growth, Engagement, Custom). For each parameter:
- Pick the mode (Filter / Score / Exclude)
- Set the value (single value, list, range, or "any of")
- For Score mode, set the weight (relative contribution)
The drawer is non-modal — Copilot stays open on the right and can answer questions as you fill it in. > what's a reasonable weight for industry vs revenue band? works mid-edit.
When you save, the new ICP scores all existing records in the workspace within a few seconds. The ICP score column on the Records page updates live.
When to use which path
Use Copilot when... | Use the manual page when...
You're defining an ICP from scratch | You're tuning a single parameter on an existing ICP
You want a draft based on historical wins | You need to inspect or copy a specific parameter config
You're not sure which parameters to include | You're auditing why a specific account didn't qualify
Both paths write to the same underlying ICP definition. There's no "Copilot ICP" vs "manual ICP" — they're just two editing surfaces over one model.
Reading the result
After save, the ICP appears in the Qualification table with these columns: Status (Draft / Live / Paused), Type (ICP), Params count, Qualified count (records currently matching), L30D sparkline (last 30 days of qualification flow), Enriched %, Synced %, Last qualified, Last modified, Modified by. The Qualified count is the simplest correctness check — if it's 0 after publishing a Live ICP, your filters are too tight.