How to Choose a Primary ICP Without Ignoring Secondary Opportunities

Who this worksheet is for

This worksheet is for teams with two or more plausible ICPs and limited research, service, inventory, localization, or sales capacity. It is suitable for early market entry, product-line expansion, a channel-versus-direct choice, or a reset after a broad account program produces unclear learning.

Contents

1 Define what primary means 2 Normalize the segment records 3 Apply hard gates first 4 Build a transparent priority score 5 Assign secondary opportunity states 6 Set scope and review triggers 7 Complete the primary-ICP worksheet 8 Illustrative example 9 Common mistakes 10 What this does not mean 11 Common questions 12 Sources and methodology

1. Define what primary means

A primary ICP is the account hypothesis that receives concentrated GTM work for a defined period. It controls research rules, data enrichment, messaging, qualification, learning reviews, and resource priorities. It does not mean “the only kind of customer allowed forever.” Write the decision boundary before scoring: For [product and geography], [ICP version] is primary for [period or milestone]. It receives [named resources and activities]. Other segments receive only the actions permitted by their state. The decision will be reviewed when [date or evidence trigger] occurs. Primary status should be specific to a product, region, and motion. The same company can fit a primary ICP for one offer and a secondary ICP for another. Avoid a company-wide label that outlives its assumptions. Public data can help describe candidate populations, but it cannot choose the primary segment. Statistics of U.S. Businesses publishes data by establishment industry and enterprise size, including firms, establishments, employment, and payroll within its stated coverage. U.S. Census Bureau - Statistics of U.S. Businesses Those counts do not resolve application, relationship, accessibility, or serviceability.

2. Normalize the segment records

Compare like with like. Every candidate segment needs the same fields:

• account unit and geography;

• product and application;

• commercial relationship;

• buying motion;

• hard requirements and exclusions;

• evidence state by criterion;

• estimated reachable account universe, with unit and method;

• service, delivery, certification, and channel dependencies;

• validation history and limitations;

• owner and version date. Do not compare an industry-wide establishment count with a hand-verified list of facilities and conclude that the larger number is more attractive. County Business Patterns provides establishment statistics by industry and geography. U.S. Census Bureau - County Business Patterns API Treat that as a population frame whose coverage and unit are retained, not as qualified demand. Normalize the evidence date as well. A current facility page and a five-year-old partner directory should not receive equal confidence merely because both are URLs.

3. Apply hard gates first

SellersGPT recommends five segment gates: Gate Pass question Examples of unresolved dependency Application Does the exact offer fit a repeatable use under required conditions? Installed process or alternative unknown Qualification Can mandatory technical, regulatory, or vendor requirements be met? Certification scope not confirmed Serviceability Can delivery, support, warranty, and lifecycle duties be fulfilled? Local response time not costed Commercial path Is there a viable OEM, distributor, end-user, or project buying path? Buyer and specification control unclear Conflict and permission Is the path permitted under channel, territory, suppression, and policy rules? Existing exclusivity under review Outcomes are pass, fail, or unknown. A failed critical gate removes the segment from scoring until the condition changes. An unknown critical gate creates a research dependency; it is not scored as zero or two. The gate is applied to the segment definition, then checked on a bounded account sample. One compliant account does not prove the whole segment passes. Conversely, one atypical failure should not automatically invalidate a properly bounded segment. Retain the sample and denominator.

4. Build a transparent priority score

Score only gate-passed segments. Use 0-4 anchors:

• 0: evidence contradicts the criterion.

• 1: weak, exceptional, or poorly supported.

• 2: mixed evidence with material limitations.

• 3: strong evidence with a manageable dependency.

• 4: direct, repeatable, and scope-matched evidence. Recommended criteria and default weights are an editorial starting point, not a universal standard: Criterion Weight What it measures Application strength 30% Specific, repeatable product-use fit Problem evidence 25% Credible operational consequence and alternative Relationship fit 20% Viable buying role and commercial model Accessibility 15% Ability to identify and appropriately reach accounts Criterion Weight What it measures Serviceability 10% Ability to deliver and support within boundaries Weighted priority = Σ (criterion rating / 4 × criterion weight) The result ranges from 0 to 100 only if every criterion has a supported rating. If a scored field is unknown, show an incomplete score and resolve it; do not substitute the mean. Keep a comment and evidence link beside each rating so the number remains auditable.

5. Assign secondary opportunity states

Every non-primary segment receives one state: State Meaning Permitted work Bounded test Viable and strategically relevant, but a named uncertainty needs evidence Fixed sample, question, owner, time, and stop rule Watch Action depends on an external or internal trigger Monitor only the named trigger at a defined cadence Park Not justified now, but retained for future review No active prospecting; preserve rationale and re-entry condition Retire Definition is invalid, duplicative, or persistently contradicted Archive version; do not silently reuse This is the mechanism that preserves opportunity without weakening focus. “Secondary” by itself is not a state because it does not say what work is allowed. A bounded test might verify 15 facilities for a specific application. A watch might monitor completion of a certification. A parked distributor segment might reopen when conflict rules or local inventory capacity change. Each state has a next review condition.

6. Set scope and review triggers

Primary selection becomes dangerous when it cannot change. Set at least one time trigger and one evidence trigger. Examples:

• review after 30 researched accounts, not after an open-ended campaign;

• review when application confirmation falls below a stated threshold in the bounded sample;

• review when a mandatory qualification requirement changes;

• review when service cost exceeds the approved boundary;

• review when a channel conflict is resolved;

• review when the reachable universe is materially smaller than assumed. First-party company filings can help test assumptions about operations and segments. EDGAR offers public access to company filings. U.S. Securities and Exchange Commission - Search Filings Procurement notices may reveal project requirements and dates; SAM.gov describes its opportunity records as notices from federal contracting offices. SAM.gov - Contract Opportunities Neither source proves the account will engage with a seller. Record what would reverse the decision. A primary ICP that has no disconfirming condition is a preference, not a governed hypothesis.

7. Complete the primary-ICP worksheet

Decision field Segment A Segment B Segment C Segment D Definition and account unit Application and relationship Gate result Critical unknown Application rating (30%) Problem rating (25%) Relationship rating (20%) Accessibility rating (15%) Serviceability rating (10%) Weighted priority Decision state Test or review trigger Add evidence references and a decision owner outside the compact display. The worksheet should not become a score-only artifact.

8. Illustrative example

Illustrative example - explicitly fictional. Harbor Motion, a fictional component exporter, has three gate-passed segments and one distributor segment with an unresolved conflict gate. It scores only the three eligible segments. Each rating uses the 0-4 anchors above. Segment Application 30% Problem 25% Relationship 20% Access 15% Service 10% Calculation Total A: packaging OEMs 4 3 4 3 4 4/4×30 + 3/4×25 + 4/4×20 + 3/4×15 + 4/4×10 90.0 B: food plants 3 4 3 2 4 3/4×30 + 4/4×25 + 3/4×20 + 2/4×15 + 4/4×10 80.0 C: retrofit integrators 4 2 2 4 2 4/4×30 + 2/4×25 + 2/4×20 + 4/4×15 + 2/4×10 72.5 Segment D, broadline distributors, is not assigned ratings or a total. Its territory and exclusivity rules are unresolved, so the conflict gate is Hold. It remains Parked with one re-entry condition: documented approval of the territory and channel-conflict rule. Segment A becomes primary for one 90-day learning cycle. Segment B receives a 12-site service-cost and application-verification test. Segment C is watched for named retrofit projects. Segment D remains unscored and Parked until its hard gate is resolved. The totals rank three fictional, gate-passed hypotheses. They do not represent probabilities, market shares, response rates, lead volumes, or revenue. Segment D demonstrates that a plausible segment receives no weighted comparison while a hard gate is unresolved. If Segment A had failed a mandatory product-qualification gate, its 90.0 score would not restore eligibility.

9. Common mistakes

• Selecting the segment with the largest raw company count.

• Scoring before checking hard gates.

• Giving unknown fields a neutral midpoint.

• Adjusting weights after seeing which segment wins.

• Treating a score as a probability of purchase.

• Calling several ICPs “primary” to avoid a decision.

• Leaving secondary opportunities active without scope or account coordination.

• Failing to define what evidence would change the selection.

10. What this does not mean

Choosing a primary ICP does not guarantee that the segment will respond, generate qualified leads, create opportunities, or produce revenue. It allocates learning and execution resources based on current evidence. The worksheet does not replace product compliance, legal, channel, finance, or service review. It does not require rejecting inbound demand from outside the primary ICP; such demand should be evaluated under explicit qualification rules. It also does not make employee or revenue size a universal measure of attractiveness.

11. Common questions

Can two ICPs tie? Yes. Use the decision boundary: which segment offers the more valuable bounded learning, lower irreversible commitment, or clearer disconfirming test? If the tie remains, run two small tests rather than declaring two permanent primaries. How often should the primary ICP be reviewed? Use the earliest of a time milestone, sample milestone, or material evidence change. Avoid changing it after every isolated reply or loss. What happens to inbound accounts outside the primary ICP? Evaluate them against a documented secondary or exception rule. Do not automatically reject them, and do not let exceptions rewrite the primary ICP invisibly. Should account counts be a scoring criterion? Only with careful definition. Use qualified, reachable, deduplicated units and retain uncertainty. A large unverified establishment count should not dominate stronger application evidence. Can leadership override the score? Yes, if the rationale, evidence, owner, and review condition are documented. The score supports judgment; it does not replace ownership.

12. Sources and methodology

The gates, weights, anchors, score formula, opportunity states, worksheet, and review rules are SellersGPT editorial tools. Official sources support only their described statistical, disclosure, or procurement functions. Calculated illustrative totals are not forecasts. 1 ICP04-CENSUS-SUSB - U.S. Census Bureau, “About Statistics of U.S. Businesses.” https://www.census.gov/programs-surveys/susb/about.html Accessed August 24, 2026. 2 ICP04-CENSUS-CBP - U.S. Census Bureau, “County Business Patterns APIs.” https://www.census.gov/data/developers/data-sets/cbp-zbp/cbp-api.html Accessed August 24, 2026. 3 ICP04-SEC-EDGAR - U.S. Securities and Exchange Commission, “Search Filings.” https://www.sec.gov/search-filings Accessed August 24, 2026. 4 ICP04-SAM-OPPORTUNITIES - SAM.gov, “Contract Opportunities.” https://sam.gov/opportunities Accessed August 24, 2026.