How We Found Our ICP

Mar 2, 2026

Before we could run a single ad or send a single outbound email, we needed to answer a deceptively simple question: who are we actually trying to reach?

Ascend had 650+ clients and $20m in ARR, almost entirely from word-of-mouth. We had a strong sense of who our customers were, but we'd never formally segmented them or built the kind of structured ICP work that you need to run targeted acquisition at scale.

This is how we did it.

Finding Your ICP.md

Finding Your ICP

finding-your-icp.md

Drop this file into your .claude/skills/ directory

Step 1: Analyse the data we already had

We started with our booking data — 4,582 bookings across Jan-Oct 2025, representing ~$20m gross revenue and ~$3m net revenue. We ran a full analysis to understand who our most valuable customers actually are, not just who we think they are.

Key findings:

  • Customer base is well diversified — top 10 clients control only ~10% of revenue, top 10 groups ~16%. No dangerous concentration.
  • The profitability sweet spot is ranks 11-50 — these clients deliver 16-27% above-average margins. The very top clients often have volume discounts that compress margins.
  • EA-booked travel dominates — roughly 75% of revenue comes through executive assistants booking on behalf of their principals
  • Primary verticals: private equity, venture capital, hedge funds, family offices, investment banking, boutique consulting
  • Primary routes: North America to Europe (where our margins are strongest)
  • Notable clients: Google Ventures, General Atlantic, Charlotte Tilbury, Bitkraft Ventures, Bessemer Venture Partners, Stella McCartney

Step 2: Enrich our top 500 customers

We wanted to go deeper than booking data alone. We ran an enrichment project on our top 500 customers by gross profit, using Firecrawl for web scraping (LinkedIn profiles, company sites, bios) and then Perplexity's AI search for additional research when we hit Firecrawl's free tier limits.

What the enrichment revealed:

  • 32.7% are crypto/Web3 connected (vs less than 1% of the general population) — and they deliver 27% higher average margin than non-crypto clients
  • 70.9% are founders
  • 56.4% are active investors
  • 40% are dual founder-investors (serial entrepreneurs)
  • Geographic concentration: 51% US (SF/NYC/LA), 27% UK/Europe, 22% international

This confirmed two things: our primary ICP is executive assistants at finance and investment firms, and our secondary ICP is high-net-worth executives who travel frequently — with a higher concentration in sectors like crypto, banking, and venture capital. Both represent specific, targetable audiences.

Step 3: Build the prospect lists

With the ICP defined and segments identified, we built prospect lists across six target groups:

  • Finance/Investment Executives (target: 800-1,200 contacts) — MD/Partner/GP at PE/VC/HF/family offices, 2-50 employees, NYC/Miami/SF/London
  • Tech Founders/CEOs (600-1,000) — Series A-D SaaS/fintech/AI companies, 20-500 employees
  • Crypto/Web3 Founders and Investors (500-800) — GPs at crypto funds, protocol founders, hubs in Miami/NYC/Singapore/Dubai
  • Luxury/Fashion/Media Executives (400-600) — validated by Charlotte Tilbury, Stella McCartney, The Guardian already being clients
  • Consulting/Law Partners (400-600) — boutique/mid-tier firms with international practices (excluding Big 3 and AmLaw 50 who are locked into corporate travel management companies)
  • HNW Solo Operators (300-500) — post-exit founders, angel investors, family office principals

Data sourced from Apollo (email/phone enrichment), enriched via Clay, and then uploaded as:

  • Meta Custom Audiences — SHA-256 hashed PII, 1% US lookalikes generated
  • LinkedIn Matched Audiences — account-based company list of 148 private equity, venture capital, and hedge fund firms (Blackstone, KKR, Sequoia, Bridgewater, Citadel, Iconiq, etc.)
  • Seed data for outbound — fed into HeyReach (LinkedIn) and Instantly (email)

What this unlocked

The ICP work wasn't just a research exercise. It directly shaped everything that came next:

← Back to the full growth engine overview