ICP quality directly determines AI SDR conversion rates. Tight ICPs produce 3-5x higher reply rates than broad ones. A well-defined ICP combines role titles, industries, company size, geography, tech stack signals, and growth triggers. Most AI SDR failures trace back to ICP that's too broad.
The 7 ICP components for AI SDR
| Component | Example |
|---|---|
| 1. Role titles | VP Marketing, Head of Demand Gen, Director of Demand |
| 2. Industries | B2B SaaS, fintech, healthtech (specific verticals) |
| 3. Company size | 50-500 employees, $5M-$50M ARR |
| 4. Geography | US + Canada + UK (or specific cities) |
| 5. Tech stack signals | Using HubSpot or Marketo (specific tools) |
| 6. Growth signals | Series A+ funding, hiring 3+ marketers this year |
| 7. Exclusions | No competitors, no current customers, no defunct companies |
Tight ICP example vs broad ICP
Broad (fails)
"B2B marketing leaders" — produces ~5M prospects, generic AI outreach, sub-1% reply rate.
Tight (works)
"VPs / Heads of Demand Gen at Series A-B B2B SaaS companies (50-500 employees) using HubSpot + Mutiny, hiring at least one marketer in the last 6 months" — produces ~3,000 prospects, AI can personalize per-prospect, 5-10% reply rate.
Common ICP traps
- Too broad: "All SaaS" or "All decision-makers"
- Too narrow: Under 500 prospects = exhaustion in 30 days
- Title-only: Missing company-level filters
- No signal triggers: Targeting cold companies vs ones with buying signals
- Geography too wide: "Global" produces compliance issues + dilution
- No exclusions: Wasting touches on competitors, current customers, defunct companies
ICP refinement process
- Define initial ICP (7 components).
- Run 200-500 cold emails over 2-4 weeks.
- Analyze who replied positively — what are common traits?
- Tighten ICP to overweight those traits.
- Identify who didn't reply — are there sub-segments to exclude?
- Repeat monthly.
Define ICP in Artra free — 10 minutes →
Frequently asked questions
What is ICP for AI SDR?
ICP (Ideal Customer Profile) for AI SDR is the precise definition of the buyer the AI agent should target — role titles, industry, company size, geography, technology stack, growth signals, and any other filters that distinguish good prospects from noise. ICP quality directly determines AI SDR conversion rates: tight ICPs produce 3-5x higher reply rates than broad ones.
How do I define an ICP for AI SDR?
Define ICP for AI SDR by combining: (1) role titles (3-5 specific titles, not 'all decision-makers'), (2) industries (2-3 specific verticals), (3) company size (revenue range or employee count), (4) geography (target regions), (5) technology stack signals (companies using specific tools), (6) growth signals (Series A+ funding, hiring patterns, executive transitions), and (7) exclusions (industries or companies to avoid). The combined filters should target 1,000-50,000 prospects total.
How tight should an ICP be?
ICP should be tight enough that the AI can generate genuinely personalized messages for each prospect. Specific rule: if your ICP filter produces 50,000+ prospects, it's too broad. If it produces under 500 prospects, you'll exhaust the list and need more. Sweet spot: 1,000-10,000 prospects, enabling 3-6 months of outbound at 50-100 emails/day before exhaustion.
What's the biggest ICP mistake reps make?
The biggest ICP mistake reps make with AI SDRs is being too broad. 'All B2B SaaS' or 'all marketing directors' produces generic AI spam with poor reply rates. Tight ICPs ('Series A SaaS marketing directors using Mutiny + HubSpot') produce dramatically better outcomes because the AI can write sentence-level personalization that resonates with that specific audience.
Should I have multiple ICPs?
Yes — most teams have 2-4 distinct ICPs and run separate AI SDR cadences for each. Example: ICP 1 (Mid-market SaaS marketing leaders), ICP 2 (Enterprise marketing leaders), ICP 3 (Indie SaaS founders). Each ICP gets its own sequence, voice, and signals. Artra supports running multiple ICPs in one account with separate cadence configurations.