How Cogny ICP Analysis Works

Your ideal customer profile (ICP) is one of the context layers every report and ticket reads. Cogny drafts it, you correct it, the reports check it against who actually converts, and the ad platforms get it as targeting. This page describes that cycle without the implementation detail.

Where the ICP lives

Settings → AI Knowledge → AI Context Generator has an ICP card next to the industry, company, competitor and strategy cards. It holds the profile as structured text: who the customer is, which segments matter and why, what they respond to, and who is explicitly not a fit. It also holds per-platform targeting notes, so the same profile can be expressed as Google Ads audiences and as Meta interests and lookalikes.

The card can generate the profile, you can edit every field, and Export targeting produces a checklist you can apply in the ad platforms or hand to an agent to apply through the integrations.

How the profile is generated

The first draft comes from what Cogny already knows: the company and industry context, the growth strategy, and, once data sources are connected, the observable buyers. Where the data allows, the draft is grounded in real conversion behaviour rather than assumption: which segments, sources, devices, regions and landing pages actually produce purchases, sign-ups or qualified leads, and how those segments differ in value and repeat behaviour.

Two things improve the draft more than anything else:

Named accounts, customer lists and lifetime value uploaded to the context tree, so the agent can reason about which segments to prioritise from your own history rather than from the web.
A sentence in the growth strategy about who you do not want. "B2B only", "no students", "enterprise over SMB" removes whole classes of off-target proposals.

Treat the generated profile the way you treat the other generated context: roughly right, review it, correct it before enabling reports.

How it is validated

The Audience Intelligence & ICP Validation report (Creative & Audiences, biweekly by default, needs GA4 or Segment) does the ongoing check. It maps the audience data available in your sources, builds the profile of who actually bought in the period, compares that with the stated ICP, scores each active audience on the ad platforms, looks for saturation, and proposes new segments with evidence. Its output is an audience scorecard, an actual-buyer profile, new audience recommendations and priority actions, which become tickets like any other recommendation.

Over time the stated ICP and the observed buyer profile converge, or the report tells you where they do not and why: a segment that converts but was never targeted, or a targeted segment that never buys.

How it is used

Reports receive the ICP with the rest of the context, so channel audits judge audiences against it and content reports write for it.
Tickets reference it when proposing campaigns, landing pages and content, and the ticket agent uses it to filter proposals that do not fit.
Targeting export turns it into concrete platform settings.
Your own agent can read it over MCP as part of the context tree.

What Cogny does not do

Cogny does not look your customers up in a demographic database or a data broker. It learns which segments and channels work from your own data and remembers the results. If you need a third-party enrichment source, connect it as an integration or load an export with Bring Your Own Data, and the ICP will use it.

Models

Cogny continuously tests and evaluates the best model for each purpose in the workflow, including profile generation and audience analysis, and routes each step accordingly. Workspaces that need data to stay in Europe can run on EU-hosted options.

Context and Growth Strategy: the other context layers and what to upload
Reports and Templates: enabling the Audience Intelligence report
How Growth Tickets Work: what happens to the recommendations