AI Prompts for Ideal Customer Profile (ICP) Definition

An ideal customer profile is a set of testable criteria, not a persona poster. It should say which companies will buy, succeed and renew — and, just as usefully, which ones to walk away from. The evidence for it is in your closed-won and closed-lost data, your churned accounts and your best customers' shared characteristics, not in who you would like to sell to.

These prompts derive the profile from that evidence, turn it into explicit fit criteria with disqualifiers, and convert it into a scoring model that can be applied to a list. They need real deal and customer data; a model can structure the analysis and challenge weak criteria, but it cannot know your market.

Before you use these

Have these ready to replace the highlighted [variables]:

• Closed-won and closed-lost deals for the last 12–24 months with firmographics (industry, size, geography, tech stack) and outcome
• Customer health or retention data, and which customers expanded
• Deal size, cycle length and win rate by segment if available
• Product constraints (what you cannot serve well)

The prompts

• 1. Derive the ICP from won and lost data
• 2. Write the fit criteria and disqualifiers
• 3. Build the ICP scoring model

1. Derive the ICP from won and lost data

Act as a revenue operations analyst deriving an ideal customer profile from deal data.

Accounts: [paste: company, industry, employee count, revenue band, geography, tech stack, segment, outcome (won/lost/churned), deal size, cycle days, expansion yes/no]
Product: [what it does, who uses it, known constraints]

1. For each attribute, compare the distribution across won, lost and churned. Identify attributes where wins concentrate and losses or churn concentrate. Show the comparison as a table with counts and win rates per value.
2. Rank the discriminating attributes by how strongly they separate good outcomes from bad, and note where sample size is too small to trust.
3. Identify combinations that matter (e.g. mid-market AND a specific tech stack) rather than single attributes.
4. Draft the ICP as testable statements: firmographic fit, technographic fit, situational fit (triggers, maturity), and explicit disqualifiers derived from where you lose or churn.
5. State what the data cannot tell you (e.g. deals never pursued) and the bias that introduces.
6. Confidence rating per criterion and the additional data that would firm it up.

Do not describe an aspirational customer. Every criterion must trace to a pattern in the data provided.

2. Write the fit criteria and disqualifiers

You are converting an ideal customer profile into operational fit criteria for [product / segment].

Draft ICP: [statements]
Product constraints: [what we cannot serve, integrations required, compliance limits]
Go-to-market: [sales-led / product-led, deal size range, capacity]

Produce:
1. Must-have criteria: without these, do not pursue. Each with the reason and how it is verified (public source, first call question, data field).
2. Strong-fit criteria: raise priority when present. Same format.
3. Nice-to-have: influence messaging, not prioritization.
4. Disqualifiers: explicit conditions to stop pursuing (e.g. size below threshold, incompatible stack, regulated segment we cannot serve, recent competitor contract). Each with the verification method.
5. Edge cases: situations where a criterion conflicts with another, and the rule for resolving it.
6. A one-page version for reps: the ten questions that determine fit in order of how early they can be answered.

Keep every criterion observable. Replace vague terms ('innovative companies') with a proxy that can actually be checked.

3. Build the ICP scoring model

Act as a sales operations analyst designing an ICP scoring model.

Criteria: [must-have, strong-fit, nice-to-have, disqualifiers]
Weight basis: [our weights, or the won/lost data to derive them]
Sample list: [paste 20–50 accounts with attribute values]

1. Assign points per criterion (additive model), with disqualifiers as hard stops. Justify each weight from the data or the stated priority.
2. Define tiers (A/B/C) with score thresholds and the intended treatment per tier (outbound priority, nurture, ignore).
3. Score the sample list and show the distribution. Flag accounts that score high but look wrong, and low but look right — these test the weights.
4. Validation: apply the model retroactively to the won/lost data; report the win rate by tier. If tiers do not separate outcomes, say which weights to change.
5. Data requirements: fields the CRM must hold to score automatically, and the ones that need enrichment.
6. Review cadence and the trigger for re-weighting.

Keep the model simple enough to explain in one slide. Do not add criteria that cannot be populated for most accounts.

What the won/lost comparison looks like

Illustrative output of the first prompt for a mid-market SaaS vendor. The discriminating attribute is not the obvious one.

Related prompts

• Account Prioritization
• Sales Territory Planning
• Account Research
• Prospecting Strategy
• Win/Loss Analysis

Logical next step

After this, most sales teams move on to Account Prioritization.

All Sales prompts · Search the full library