Key takeaways
- Starting from one real customer forces a specificity that a from-scratch ICP description usually lacks.
- Write down what would have made the account a bad fit, not just what made it a good one, so the profile has real edges.
- Split what makes an account a fit into company criteria and buyer criteria before you go looking for similar accounts.
- Check the profile against a few more known-good and known-bad accounts before treating the list as final.
The fastest way to describe your ICP accurately is often not to describe it in the abstract. It is to look hard at one customer who is exactly who you want more of, and work backward from what actually makes them a fit.
Pick the right anchor customer
Not every customer makes a good anchor. Look for one who represents the segment you actually want to build a repeatable pipeline around:
- They renew or expand, not a one-time deal that closed and went quiet.
- They would give you a reference, or clearly get real value from what you sell.
- They resemble the kind of company you have more of in your addressable market, not an unusual, one-off fit.
Your single biggest deal is not automatically the right anchor if it does not fit that description. A large but atypical account can be a great customer and a poor template for list-building at the same time.
Derive what makes them a fit
Write down, in concrete terms, what this account actually is and why they bought:
- What they do: products, services, technology, and how they operate.
- The specific problem or need that existed when they bought, not a generic "they needed help."
- What a good-fit version of a similar account would look like, and just as importantly, what a bad-fit version would look like: wrong stage, wrong size, wrong function, wrong buying motion.
The bad-fit half of this exercise is easy to skip and is where a lot of the real value is. A description of only positive traits tends to be broad enough to match accounts that are not actually a fit; naming what would disqualify a similar-looking account gives the profile real edges.
Turn the profile into criteria
Split what you wrote down into the same two layers used for any ICP:
- Company fit: therapeutic area or modality, development stage, size, and the operational characteristics that mattered for this account specifically.
- Buyer fit: the function and seniority of the person who actually championed or approved the purchase, not just whoever the initial contact was.
See building an ICP as a CRO or CDMO for the fuller reasoning behind splitting company and buyer fit, and how to score account fit for turning these criteria into something scoreable.
Find companies that look like the anchor
Search using the concrete criteria from the previous step, not a loose "same industry" or "similar name" filter. Screening on therapeutic area, modality, development stage, size, and relevant technology or capability produces a materially tighter list than searching by broad category alone.
If you have more than one strong anchor customer, run this step against each and look for the accounts that show up as a match across multiple anchors. Those overlaps are usually the strongest additions to the list, since they fit more than one real example rather than just one company's specific profile.
Validate and prune
Before treating the list as final, sanity check the derived profile against accounts you already know the answer for:
- Run two or three more known-good accounts (other real customers, or accounts you already know are strong fits) through the criteria. They should score well.
- Run a couple of known-bad accounts (real losses, or accounts you know are not a fit) through the same criteria. They should score poorly.
If a known-good account fails the criteria, or a known-bad account passes, the profile is probably too narrow or too broad on a specific attribute, and is worth adjusting before the list goes to outreach. This is the same validation logic used for any ICP: the definition only earns trust once it is checked against real outcomes, not just the anchor it came from.
Checklist
- Anchor customer chosen for representativeness, not just deal size or recency.
- What made them a fit written down concretely, including a bad-fit counter-description.
- Profile split into company fit and buyer fit criteria.
- Prospective accounts searched on the concrete criteria, not a loose industry filter.
- Derived profile validated against known-good and known-bad accounts before finalizing.