Profile-to-ICP Reverse Engineering
Build an initial ICP from evidence about real target buyers
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 6
- Confidence
- 98%
Define the buyer role, geography, industry, and preferred company type, then identify several real people who plausibly match that target. Gather lawful public information from professional profiles, biographies, interviews, company pages, and other credible sources. Normalize the evidence into fields such as company characteristics, responsibilities, pain points, goals, buying triggers, preferred tools, decision authority, and communication style. Compare the profiles to find recurring patterns and use those patterns to draft an initial ideal customer profile, messaging guidance, and positioning hypothesis. The output is deliberately a first version rather than a permanent truth: inferred pains and motivations should be labeled as hypotheses, checked through customer conversations or sales evidence, and revised as better data becomes available.
Origin
Extracted from Marketing Against The Grain, where public information about five US B2B SaaS CMOs was analyzed to construct an initial ICP.
Core principles
- 01Start with real people who resemble intended buyers
- 02Use multiple profiles to identify recurring patterns
- 03Distinguish observed facts from inferred motivations
- 04Translate patterns into messaging and positioning
- 05Treat the result as a first version to validate
How to run it
- 1
Define the target cohort
Specify the role, geography, industry, business model, and any preferred company characteristics. Keep the cohort narrow enough for meaningful patterns to emerge.
Pro tip Begin with the people most likely to buy the current product, not an aspirational future audience.
Watch out A broad label such as “marketers” will produce a vague ICP.
- 2
Select representative people
Identify at least five real professionals who fit the target cohort and appear relevant to the offering.
Pro tip Choose people from several companies to avoid copying one employer's culture.
Watch out Public visibility can bias the sample toward unusually active executives.
- 3
Gather public evidence
Collect role history, company information, stated priorities, interviews, biographies, and other lawful public material. Record sources and distinguish direct statements from inference.
Pro tip Use company pages and interviews when professional-network access is limited.
Watch out Do not scrape restricted data or infer sensitive personal attributes.
- 4
Normalize profile fields
Summarize each person using consistent fields: responsibilities, goals, pains, buying triggers, tools, decision power, and preferred message style.
Pro tip Leave unsupported fields blank rather than manufacturing completeness.
Watch out Job titles alone do not establish budget authority or purchasing behavior.
- 5
Find recurring patterns
Compare profiles to identify shared company traits, responsibilities, needs, language, and purchasing conditions. Note meaningful exceptions as well as commonalities.
Pro tip Separate patterns observed across profiles from assumptions supplied by the model.
Watch out Five profiles support hypotheses, not statistically representative conclusions.
- 6
Draft and validate the ICP
Create the initial ICP, then derive positioning and messaging hypotheses from its evidence. Validate those hypotheses with interviews, sales calls, and conversion data.
Pro tip Attach confidence levels to uncertain fields.
Watch out Do not let a fictional synthesis replace contact with actual customers.
In the wild
A small B2B SaaS company identifies five US chief marketing officers, gathers public information about their roles and companies, and compares their responsibilities, goals, pains, buying triggers, tools, and decision influence. The team uses recurring patterns to draft its first CMO profile and message hierarchy.
→ The company gains a focused starting point for campaigns and a list of assumptions to test in customer interviews.
Common mistakes
Using titles as the whole ICP
A shared title does not reveal company fit, priorities, purchasing triggers, or decision authority.
Presenting inference as fact
AI-generated pain points and motivations must remain hypotheses unless supported by direct evidence.
Never validating version one
The first synthesized ICP should guide learning, not become an unchangeable description of the market.
Is it for you?
Best for
It is best for early-stage B2B companies that can identify representative target buyers but have not yet formalized their market profile.
Not ideal for
It is not ideal when public profiles are inaccessible, unrepresentative, or too sparse to support reliable inferences.
From the transcript
“That is a pretty good way if you were a smaller company and you want to be really scrappy to create a first version of…”
“It's going to identify five chief marketing officer profiles working in the US, preferably B2B SaaS.”
“Talks about the industries, talks about key responsibilities, pain points, common goals, buy-in triggers, uh, which is pretty cool.”
From the episode
The New ChatGPT Agent Promised to Save Me Hours - Did It?