Most ICPs are adjectives: 'ambitious mid-market companies that value quality'. A usable ICP is a query. If you cannot run it against a company database and get a list, it is a positioning statement, not an ICP.
Start from closed-won, not from opinion
Export your last 100 closed-won accounts, match them to register records and look at the distribution of sector code, employee band, incorporation year, legal form and region. The pattern is usually narrower and stranger than anyone expected.
Then do the same for closed-lost and churned. The difference between the two distributions is your real ICP, and it is often one or two variables.
Express it as filters
Write the ICP as a filter set: sector codes in this list, employee band 10–49, incorporated more than three years ago, active status, within these regions. Add the negative filters too — dissolved entities, holding companies with no trading activity, sole traders if you do not sell to them.
Now the ICP has a size. Knowing your addressable market is 18,400 companies rather than 'the mid-market' changes quota, budget and channel decisions immediately.
Tier it and keep it alive
Split the result into tiers by fit strength, and give each tier a different motion: tier one gets research and ABM, tier three gets scaled email. Do not treat a filter hit as a lead.
Re-run the query monthly. New incorporations, status changes and relocations continuously move companies in and out of your ICP; a static list guarantees you miss the newest, best-timed accounts.
Frequently asked questions
Which firmographic fields matter most?
Sector code, employee or turnover band, entity status, incorporation date and location cover the majority of segmentation value for most B2B businesses.
How big should an ICP be?
Large enough to sustain pipeline for two years at your win rate, small enough that a rep can name why every account is on the list. If it exceeds tens of thousands, tier it.
