ICP Development Methodology

What it is

A repeatable practice for building an Ideal Customer Profile (ICP): a scored, evidence-based definition of exactly which accounts are worth pursuing, built from firmographic data (industry, headcount, revenue), technographic data (what tools an account already runs), and behavioural or intent signals (what triggers a purchase, what an account is doing before it becomes a lead). Where the Segmentation–Targeting–Positioning (STP) Framework's Targeting step makes a strategic choice, which one or two broad segments to pursue, ICP development is the operational layer underneath that choice. It converts a targeting decision such as "mid-market fintech" into a scoring model that a sales development rep or a marketing automation platform can apply to a specific list of 10,000 accounts, ranking each one by fit. STP answers "which market are we going after?"; ICP development answers "given that market, which specific accounts should we call first, and how do we know?" The two are sequential, not competing: run STP first to choose the segment, then build the ICP to operationalise it.

When to use it

• Sales is prospecting broadly with no consistent qualification standard. Reps chase whatever account responds, regardless of fit, and pipeline quality is inconsistent from rep to rep.
• Marketing and sales disagree on what a "good" lead looks like. Marketing counts form fills; sales rejects most of them as poor fit. An ICP gives both functions one shared scoring standard.
• You are launching account-based marketing or outbound prospecting. ABM depends on a target account list; that list needs a defensible scoring model behind it, not a guess.
• Win rate or deal size varies wildly across won deals with no obvious pattern. A retrospective ICP build, scoring closed-won and closed-lost accounts against the same criteria, usually reveals which firmographic or behavioural traits actually predict a good fit.
• The customer base has drifted since the ICP was last defined. Product expansion, a new pricing tier, or a shift upmarket or downmarket can quietly invalidate an ICP that still sits unchanged in the CRM.

How to run it

1. Start from the target segment STP already chose. Do not run this as a second segmentation exercise; take the segment STP selected (for example, "mid-market fintech, 100 to 1,000 employees") as the starting scope for the ICP build, and narrow from there.
2. Pull the evidence base. Export closed-won and closed-lost deals from the CRM for the past 12 to 24 months, at least 50 to 100 accounts if the pipeline supports it. For each, gather firmographic data (industry, employee count, revenue band, geography), technographic data (what adjacent tools the account already runs, pulled from a data enrichment provider), and behavioural data (what content they engaged with, how they first made contact, how long the sales cycle ran).
3. Isolate what separates wins from losses. Compare the won accounts against the lost and never-progressed accounts on every attribute gathered in step 2. Look for attributes that appear disproportionately in wins: for example, if accounts already running a specific complementary tool close 3 times more often than accounts that do not, that tool becomes a strong-fit signal.
4. Build a weighted scoring model. Assign points to each signal in proportion to how strongly it predicted a win in step 3 (heavier weight to firmographic and technographic fit, lighter weight to softer behavioural signals unless the data clearly supports otherwise). Group the resulting score into three or four fit tiers, for example Tier A (strong fit, prioritise), Tier B (good fit, standard cadence), and Tier C (poor fit, deprioritise).
5. Validate against a holdout set. Score a set of accounts the model was not trained on, ideally the most recent quarter's closed deals, and check whether the tiers it assigns actually correlate with what happened: did Tier A accounts close faster and at a higher rate than Tier C? If not, revisit the weighting in step 4 before rolling the model out.
6. Operationalise the score. Push the model into the CRM and marketing automation platform as a scored field on every account and lead record, so sales development reps can prioritise outreach and marketing can route inbound leads by tier without a manual judgement call each time.
7. Brief sales and marketing on how to use it, not just what it says. A score with no explanation of what to do differently for Tier A versus Tier C sits unused. Give reps a concrete instruction: Tier A gets same-day outreach and an executive-level pitch; Tier C gets a slower, self-serve nurture track or is deprioritised entirely.

Cadence & ownership

PMM typically owns the ICP definition and its evidence base, in close partnership with revenue operations, who own the scoring implementation inside the CRM, and sales leadership, who validate that the tiers match what reps see in the field. Refresh the full model quarterly, re-pulling the latest closed-won and closed-lost data, and treat a product launch, a new pricing tier, or a deliberate move upmarket or downmarket as an automatic trigger for an off-cycle refresh, since any of these can shift which accounts are genuinely the best fit. Between full refreshes, spot-check the model monthly by comparing the current quarter's win rate by tier against the previous quarter; a Tier A win rate that quietly drops toward Tier B's is the first sign the model has drifted out of date.