Building an ICP From Closed-Won Data
An ICP built in a workshop describes who you hope to sell to. An ICP built from closed-won and churn data describes who actually buys and stays, and those two lists overlap less often than anyone expects.
Key Facts
- Focus
- build ICP from closed-won data
- Category
- GTM Diagnostics
- Defined outputs
- 5 deliverables
- Regions served
- India · United States · United Kingdom · UAE · Singapore
- Last reviewed
- 2026-09-10
Winning Fast and Staying Are Predicted by Different Attributes.
Most ICP work looks only at closed-won and produces a profile of who converts. But the segment that converts fastest is sometimes the segment that churns hardest, and acquiring more of them actively destroys value. Without analysing churn alongside wins, the exercise can confidently point the whole go-to-market motion at your worst customers.
Closed-lost is ignored, so nobody learns what distinguishes the accounts you pursued and could not win.
Churn is treated as a customer success concern rather than as ICP evidence, which is where it is most valuable.
Segments are never sized, so beautifully specific profiles turn out to describe two hundred companies.
A Method You Can Run Yourself
Assemble Three Populations, Not One
Closed-won, closed-lost and churned, each with firmographic, technographic and behavioural attributes appended. The comparison across all three is what produces a usable profile. Closed-won alone tells you what you sold to, not what you should be selling to.
Look Past Firmographics
Industry and headcount are weak predictors and they are what everyone checks. Test technology in use, roles present on the org chart, recent trigger events, funding stage and operational characteristics. The strongest discriminator is usually something operational that nobody had thought to look at.
Weight for Retention, Not Only Conversion
Score attributes on both probability of closing and probability of retaining past payback. An attribute that predicts fast closes and high churn should reduce your ICP score, not increase it, and this inversion is the single most valuable output of doing the analysis properly.
Size It Before You Commit
Count the addressable accounts matching your derived profile using real source data. A segment that cannot support your revenue target needs broadening, and finding that out during the analysis is considerably cheaper than finding out after two quarters of a motion built around it.
Deliverables
- A three-population dataset covering won, lost and churned accounts with attributes appended
- Ranked discriminating attributes tested beyond firmographics
- A weighted scoring model balancing close probability against retention
- Defensible addressable account sizing for the derived profile
- The model deployed as a live score on your CRM records
Is This You?
Strong fit
- You have at least fifty closed-won accounts and some churn history.
- Your current ICP was written from intuition rather than derived from data.
- You suspect some of your customers are not worth acquiring.
Not a fit yet
- You have under twenty customers. Interview them all instead. It is better evidence at that scale.
- You cannot access churn data. The analysis is incomplete and potentially misleading without it.
Run the Comparison
Export won, lost and churned accounts. We will walk through what separates them on the call, and the attribute that discriminates most is almost never the one the team predicted.
Book a 30-Min Strategy CallSend a Request
We'll be in touch!
Expect a call within 1 business day.
Common Questions
How many customers do I need for this?
Fifty closed-won is a workable floor for directional patterns; a hundred and fifty or more supports proper weighting. Below fifty the statistics imply more certainty than the data supports, and structured customer interviews are genuinely better evidence.
What if my best segment is too small?
Then you broaden deliberately, accepting a lower average fit in exchange for reachable volume, and you monitor whether the broader segment retains. This is a legitimate trade and it should be made explicitly rather than discovered when the pipeline runs dry.
Should I include closed-lost in the analysis?
Yes. Closed-lost tells you which accounts you can attract but not convert, which is a different and useful signal from accounts you never engaged. If a segment appears frequently in lost and rarely in won, that is a positioning or product problem worth surfacing.
How often should I redo this?
Annually as a formal exercise, with continuous drift monitoring in between. Product changes, pricing changes and new competitors all shift the pattern, and the expensive failure is discovering the shift two quarters after your motion stopped working.
Related GTM Systems
ICP Definition & Segmentation
Ideal customer profiles derived from your closed-won and churn data, scored, sized and pushed into the systems your reps and campaigns actually use.
Lead Enrichment Automation
Multi-source enrichment waterfalls with coverage tracking, cost control and freshness SLAs, so every record your reps and agents touch is trustworthy.
GTM Metrics That Matter
Which go-to-market metrics predict outcomes, which ones only describe the past, and how to build a dashboard with fewer numbers and more signal.