How to Expand into Adjacent Markets Without Diluting the ICP

Who this framework is for

This framework is for manufacturers, exporters, industrial suppliers, growth leaders, and B2B product or sales teams that have learned something from a core market and are choosing where to expand next. It is most useful when many industries can technically use the product but only some share the same reason and path to buy.

Contents

1 Freeze the causal core of the ICP 2 Define five types of adjacency 3 Discover candidates with official data 4 Apply the SellersGPT Adjacency Transfer Test 5 Run a separate adjacency experiment 6 Illustrative example 7 Common mistakes 8 What this does not mean 9 Common questions 10 Sources and methodology

1. Freeze the causal core of the ICP

Before expanding, write the core ICP as evidence, not labels: [Account and facility type] experiences [material job] in [application] when [trigger]. [Buying group] purchases through [path] because [value and proof]. We can deliver within [product, commercial, service, and geographic boundary] at [viable economic conditions]. Separate invariants from observed attributes. ICP element Invariant candidate Attribute that may vary Operational job Outcome that creates value Industry terminology Trigger Event that moves the buyer to action Timing or source of event Application Process conditions required for fit End product made by the plant Buying group Functions that control risk and budget Job titles and reporting lines Proof Evidence required to approve change Format, standard, or reference Route Who specifies, buys, installs, and supports Named distributor or integrator Economics Order and lifecycle value sufficient to serve Price level and payment structure Delivery Required certification, logistics, installation, and service Exact region or partner An invariant survived in the core market but is not universal. Preserve it so the test can show transfer. Do not confuse NAICS with ICP The U.S. Census Bureau states that NAICS classifies establishments for statistical purposes according to economic activity. Census - NAICS An ICP can cut across multiple NAICS industries when the same application and buying logic recur. Conversely, one NAICS industry can contain facilities with different processes, relationships, and buying conditions. Use classification to organize data. Use causal buyer evidence to define the ICP.

2. Define five types of adjacency

Application adjacency The product performs a similar job in another process. Operating conditions, standards, proof, and users can change. Industry adjacency The application remains similar in another industry. Process, maintenance, risk, and buyer functions determine transfer. Geographic adjacency The account and application move to a new region, where tariffs, certification, language, logistics, channel, service, or procurement may change. Relationship adjacency The product moves among end users, OEMs, distributors, contractors, or integrators, changing specification control, margin, data, support, or cycle. Offer adjacency The buyer and job remain, but configuration, service, retrofit, consumable, rental, or delivery model changes. Do not test an industry, geography, and relationship change in one launch. If the result fails, the team will not know which assumption broke.

3. Discover candidates with official data

Official data generates adjacency hypotheses. It does not approve them. Product-by-industry evidence The 2022 Economic Census released products-by-industry and industries-by-product tables using NAPCS product codes and NAICS industry codes. The tables show which products are associated with which industries and related establishment and revenue measures. Census - New Product Statistics from the 2022 Economic Census This can reveal industries where a product or service appears outside its primary category. It does not show whether the same application, buying group, or supplier opportunity exists. Industry-input evidence BEA's input-output accounts show how industries produce and use commodities and how supply chains interact. BEA - Input-Output Accounts Use tables can suggest industries consuming related inputs; requirements tables can reveal direct and indirect relationships. Economic flows do not identify facilities, vendors, or triggers. Use them to choose where to investigate. Occupational evidence BLS Occupational Employment and Wage Statistics publishes occupation estimates across industries and geographies. BLS - OEWS Staffing patterns can suggest relevant functions, but not prove the buying committee. Account-density evidence Census County Business Patterns provides U.S. establishment counts by industry, geography, and employment size. Statistics Canada's employer-location table provides comparable research inputs for Canadian NAICS, metro, subdivision, and size bands. Statistics Canada - Table 33-10-1097-01 Use counts to find a serviceable cluster, then resolve parents and verify facilities. One enterprise may control several sites, and not every establishment uses the application. Trade-flow evidence USITC DataWeb provides official U.S. merchandise-trade access. Relevant product flows can suggest supply presence, import dependence, or geographic partners. HS trade data classifies merchandise, not buying accounts or end-use industries. Document the HS rationale and validate the application separately. Company evidence SEC EDGAR can reveal a public company's industries, end markets, acquisitions, customers, channels, capital priorities, and risks. Use company filings to refine an adjacency hypothesis and identify differences, not to infer an entire segment from one firm.

4. Apply the SellersGPT Adjacency Transfer Test

Score transfer by evidence state, not intuition. Transfer dimension Question Evidence to collect Job Is the operational outcome materially the same? Workflow, cost, risk, downtime, requirement Trigger Does the same event create urgency? Recent projects, failures, customer or policy change Product Does the offer meet process and technical conditions? Application engineering, standards, testing Buyer Are users, influencers, approvers, and procurement roles similar? Buying-committee interviews and records Proof Will existing references, data, and validation satisfy the new market? Approval criteria and objections Route Can the same direct, distributor, OEM, contractor, or integrator path work? Specification and transaction map Economics Do order value, frequency, cycle, margin, and service support the model? Scenario P&L and cash timing Delivery Can logistics, certification, installation, parts, and response be fulfilled? Operational plan and owners Classify each dimension:

• Transfers: repeated evidence supports reuse with no material change.

• Adapts: the logic remains, but a bounded change is required.

• Rebuilds: the element needs a new capability, proof set, or operating model.

• Contradicted: evidence shows the core logic does not hold.

• Unknown: evidence is missing. Calculate transfer leverage transparently An optional internal calculation is: Transfer leverage = transferred validated elements / total elements tested The result is a calculated project metric, not a probability of success. Do not count Unknown as Transfers. Do not let a high ratio hide one fatal contradiction in Product, Economics, or Delivery. Use non-compensating gates SellersGPT recommends four hard gates: 1 the job and trigger create a plausible purchase; 2 the product can perform safely and credibly; 3 the market can be served at viable economics; 4 required delivery and support can be fulfilled. If one is contradicted, the adjacency should be redesigned, deferred, or rejected regardless of similarities elsewhere.

5. Run a separate adjacency experiment

Create an adjacency ICP Do not edit the core ICP into a longer list. Create a versioned adjacency record containing:

• changed dimension and reason;

• preserved hypotheses;

• new account criteria and exclusions;

• named account universe;

• buyer and route hypotheses;

• required proof adaptations;

• product and delivery gaps;

• economics range;

• learning goals and stop conditions. Select a bounded account set Choose verified facilities with the target application and one visible trigger. Include sufficient variation to test the changed dimension, but do not mix several different adjacencies. Reuse selectively Reuse only assets tied to transferred elements. A core case study may support the operational job but fail the new industry's validation requirement. A distributor may cover the geography but not the buyer relationship. Mark reuse as confirmed, adapted, or unverified. Maintain a learning ledger Observation Core expectation Adjacency evidence Transfer state Required change Decision impact Buyer role Trigger Required proof Route Service Economics Log contrary evidence. If a pattern appears only in one account, do not promote it to the adjacency ICP. Decide whether to merge

• Merge with core: repeated evidence shows the same causal buying and delivery logic, and the extra label no longer improves decisions.

• Keep separate: the adjacency is viable but requires different roles, proof, route, terms, or service.

• Narrow: only one application, trigger, or account type works.

• Redesign: need exists, but the offer or operating model must change.

• Stop: a hard gate is contradicted. Separate ICPs are not administrative failure. They prevent an apparently broad market from erasing the conditions that made the original program work.

6. Illustrative example

Illustrative example. All companies, account counts, product details, and findings are fictional. They do not represent a SellersGPT customer or actual market results. A maker of corrosion-monitoring devices has a fictional proven ICP among chemical-processing facilities. It considers water-treatment plants because both manage corrosive environments. The product team initially labels the new industry "high fit." The Adjacency Transfer Test shows:

• the monitoring job transfers;

• the trigger adapts from unplanned process loss to scheduled asset-integrity programs;

• the product needs different enclosure and documentation proof;

• the buying group shifts toward engineering consultants and public procurement;

• the route rebuilds because consultants influence specifications;

• service geography remains feasible;

• the sales cycle and bid effort change the economics. The adjacency is not added to the chemical ICP. A separate water-treatment ICP is created for plants with a named integrity program, applicable installation conditions, and accessible consultant path. Existing technical proof is adapted; commercial proof must be rebuilt. The decision is Test separately, with stop conditions for specification access and economics. The technical similarity generated the hypothesis, but buying and route evidence shaped the actual market.

7. Common mistakes

Expanding from "can use" instead of "will buy" Technical compatibility does not establish urgency, budget, proof, channel, or delivery. Start with the causal purchase logic. Adding every successful account attribute to the ICP Some attributes are incidental. Identify which conditions explain need, access, economics, and delivery before treating them as rules. Changing several dimensions at once A new industry in a new country through a new distributor tests multiple systems. Separate the changes or explicitly accept that the result will be hard to diagnose. Merging after one early win One account can be an exception driven by a champion, project, or unique condition. Require repeated patterns and document contrary cases. Reusing proof without checking approval criteria A reference that demonstrates performance may not satisfy a different standard, application risk, or buyer function. Verify required evidence. Ignoring delivery dilution New geographies and applications can create certification, inventory, installation, language, travel, and parts obligations. Model them before outreach. Protecting the adjacency by weakening exclusions If few accounts qualify, revisit the opportunity. Do not broaden the ICP until the account list looks large.

8. What this does not mean

• Similar NAICS, NAPCS, occupation, input-output, or trade codes do not prove a shared ICP.

• Product compatibility does not prove market accessibility or buyer willingness to change.

• Transfer leverage is not a forecast or probability of success.

• One adjacent customer does not justify merging segments.

• Separate ICPs do not require separate products, but they may require different proof, routes, or service.

• The framework does not predict leads, orders, revenue, or market share.

9. Common questions

Which adjacency should we test first? Prefer the candidate with a meaningful account universe, one major change, strong transfer across job and trigger, no known hard product or delivery blocker, and a viable route to evidence. Do not choose by market size alone. Can we test a new industry and geography together? Only when necessary and when the team explicitly maps both changes. The result will be less diagnostic. A same-geography industry test or same-industry geography test is usually cleaner. When should two ICPs be merged? When repeated evidence shows that the same job, trigger, buying group, proof, route, economics, and delivery logic drives both, and separate treatment no longer improves decisions. What if the product fits but the buyer is different? Keep a separate adjacency ICP. Rebuild the buying-committee map, message, proof, route, cycle, and economics while preserving valid product evidence.

10. Sources and methodology

Sources 1 U.S. Census Bureau - North American Industry Classification System. Used for the purpose and structure of NAICS as an establishment classification system. Accessed August 24, 2026. 2 U.S. Census Bureau - New Product Statistics from the 2022 Economic Census. Used for official products-by-industry and industries-by-product data scope using NAPCS and NAICS. Accessed August 24, 2026. 3 U.S. Bureau of Economic Analysis - Input-Output Accounts. Used for the scope of supply, use, and requirements tables linking industries and commodities. Accessed August 24, 2026. 4 U.S. Bureau of Labor Statistics - Occupational Employment and Wage Statistics. Used for occupation estimates across industries and geographies. Accessed August 24, 2026. 5 U.S. Census Bureau - County Business Patterns. Used for establishment density by industry, geography, and employment size. Accessed August 24, 2026. 6 Statistics Canada - Canadian Business Counts, Table 33-10-1097-01. Used for Canadian employer-location counts by NAICS, geography, and size range. Accessed August 24, 2026. 7 U.S. International Trade Commission - DataWeb. Used for official U.S. merchandise-trade access and product-flow hypothesis generation. Accessed August 24, 2026. 8 U.S. Securities and Exchange Commission - Search Filings. Used for first-party public-company evidence on markets, customers, channels, capital, acquisitions, and risk. Accessed August 24, 2026. SellersGPT analysis and framework The causal ICP statement, invariant and attribute distinction, five adjacency types, Adjacency Transfer Test, transfer states, transfer-leverage calculation, hard gates, learning ledger, and merge decisions are original SellersGPT editorial tools. Official data is used for hypothesis generation, not as proof of adjacency success. Update policy Review classification and official data as new versions and reference years appear. Keep each adjacency ICP versioned, date account evidence, and revisit transfer decisions when product, route, service, regulation, or economics changes.