How do I find my ICP if I'm pre-PMF?

Closed-won deconstruction still works with 3-5 design partners. The attribute extraction surfaces signal even at small N. Don't wait for 50 customers to start.

What's the difference between a prospect list and a prospect graph?

A list is flat — names and companies. A graph captures account-to-contact relationships, buying committee mapping, and signal history. Multi-thread outbound and ABM motions require the graph.

How much enrichment data should I buy?

Director+ titles in the US run 80%+ accurate at Apollo prices ($0.02/record). IC titles or outside US drops to 30-50%. Waterfall enrichment ($0.10-0.50/record) lifts accuracy but the math only pencils above $50K ACV.

How often should I refresh the ICP?

Quarterly. Markets shift, your product shifts, and what worked 6 months ago decays. Closed-won analysis run quarterly catches drift before it kills the pipeline.

ICP + List-Building

From guessing your ICP to a tested prospect list.

How to find who actually buys, build a prospect graph, and turn it into a list your team can sequence. Without the data ops team you don't have.

See the 7-step plan Skip the build — book a call →

The plan

The 7-step ICP and list-building plan.

Seven steps from your closed-won export to a maintained, sequenceable list. Each step links to the deep reference if you want to go further. Each step also has the obvious alternative: have us do it.

Step 1
1. Pull every closed-won record
Export every customer who has ever paid (CRM if you have one, manual list if not). Even 3-5 customers is enough.
Why: ICP without closed-won data is a guess. Even tiny datasets surface non-obvious patterns.
→ Closed-won deconstruction Or have us do this step →
Step 2
2. Extract common attributes
For each customer: industry, company size, role of buyer, tech stack, growth signal, geography. Look for unexpected overlap.
Why: The pattern that surfaces is your first ICP hypothesis. Always different from what you assumed.
→ Attribute extraction Or have us do this step →
Step 3
3. Hypothesize 2-3 ICP segments
Don't pick one. Write 2-3 hypothesis statements: 'X-sized companies in Y industry, with Z trigger event, buying from W role.'
Why: Single-hypothesis ICP fails because you have no comparison. Multi-hypothesis testing surfaces the winner empirically.
→ ICP hypothesis testing Or have us do this step →
Step 4
4. Test via outbound experiments
Send 100-200 emails per hypothesis. Measure reply rate, meeting-booked rate, qualified rate. Winner is whichever has the highest qualified rate, not highest reply rate.
Why: Reply rate measures copy quality. Qualified rate measures ICP quality. The distinction is everything.
→ Testing methodology Or have us do this step →
Step 5
5. Build the prospect graph
For the winning ICP, build a structured prospect database. Account-to-contact relationships, buying committee mapping, intent signals.
Why: A flat list is enough for 100 prospects. A prospect graph is required for 1,000+. Multi-thread outbound needs the graph.
→ Prospect graph construction Or have us do this step →
Step 6
6. Enrich the data
Apollo, Clay, or Ocean to fill missing fields. Waterfall enrichment if budget allows. First-party signals (GitHub, conference talks, podcast guests) where third-party data is thin.
Why: Bad data is the silent killer of cold outbound. Director+ titles run 80% accurate in commercial enrichment; IC titles run 30%.
→ Enrichment vendor tradeoffs Or have us do this step →
Step 7
7. Operationalize as list management
Suppression list discipline. Re-engagement cadences. Quarterly ICP review. Treat the list as a living asset, not a one-time pull.
Why: Lists decay 30-40% per year. A pristine list at launch is a stale list 6 months later without active management.
→ List management Or have us do this step →

What you're building

The four layers, in plain English.

ICP and list-building has four layers. Each layer answers one specific question your future sequences will ask before deciding which accounts to hit and how.

Find the pattern

Closed-won deconstruction and hypothesis formulation. Extract the operational ICP from customers who have already paid, then write the hypotheses tight enough to test.

Build the graph

Prospect graph construction and segmentation architecture. Move from flat lists to account-to-contact maps with buying-committee structure and the segments that drive sequencing.

Fill in the data

First-party signals, intent data, and enrichment vendors. What to mine for free, what to buy, and where the waterfall pays back.

Maintain the list

Operational list management — suppression discipline, re-engagement cadences, and the quarterly hygiene that keeps a list from decaying into noise.

Reference

Go deeper — the technical reference.

For the operator who actually wants to learn this. Eight chapters covering every step from closed-won deconstruction to operational list hygiene at production scale. Read by chapter to debug something specific, or read in order to build the list from scratch.

Find the pattern

Reverse engineering

Closed-won deconstruction

Testing methodology

ICP hypothesis testing

Build the graph

Account architecture

Prospect graph construction

Segments

Segmentation architecture

Channel selection

Segment-channel fit