Industry is a label. An ICP is a commercial decision rule.
Sales databases usually attach categories to companies: software, construction, healthcare, manufacturing, professional services and thousands of narrower variations. These labels are useful for narrowing a search, but they are not the same as knowing whether a company has a problem your product solves.
One organization may design both hardware and software. Another may distribute physical products but not manufacture them. Two businesses can have the same industry, funding stage and employee count while requiring completely different sales approaches. If the offer depends on an operating characteristic, that characteristic has to be researched directly.
An ideal customer profile (ICP) should be a set of observable, testable commercial criteria—not merely a saved search. The goal is to make “this company fits” understandable and defensible, including the cases where the answer is uncertain.
Translate the offer into facts you can verify
Start with the reason a customer would buy. Then work backwards to what would have to be true about a company for that reason to apply. Separate must-have conditions from supporting signals. A targeting team selling engineering simulation tools may care whether the prospect designs physical products and employs relevant engineering roles—not whether a database calls the company “hardware.”
| Weak targeting input | Stronger evidence-oriented question |
|---|---|
| “Manufacturing” industry | Does the company design or produce the relevant class of physical products? |
| “Healthcare” industry | Does it operate the specific facility or service model? |
| “Hospitality” industry | Does it run the locations or customer operations our offer addresses? |
| 50–500 employees | Is the relevant function staffed and is the account economically viable? |
| Raised funding recently | Does that change create a plausible need for our specific solution? |
Gather evidence from official sites, product descriptions, locations, teams, technical documentation or credible third parties. Note the source and date for decisive facts. AI classification can accelerate review, but uncertain classifications should retain an unknown state.
Define the correct entity before assigning fit
For some markets the unit of qualification is not the entire company. A healthcare group may operate different facility types. A real estate owner may have several properties with different tenants. A franchisor and its individual locations may be governed by different purchase decisions.
Model the parent-child relationship explicitly. A facility can be eligible even if the corporate parent is excluded, or the reverse, depending on the commercial and suppression rules. Without a clear record model, the system may incorrectly generalize a fact about one location to all the others.
Examples from TCA's market-research work
In TCA's published Dassault Systèmes case study, the target involved startups designing physical products across 13 European markets. The research used websites, team descriptions, funding information and accelerator portfolios to decide whether a company plausibly fit. Each account received a yes, no or unknown decision, rather than being included simply because of a firmographic label.
For Bounce Rental Solutions, the relevant segmentation depended on how party and event rental businesses accepted bookings. The published summary reports 4,800 businesses discovered across 42 states, with 3,620 classified by booking method. A conventional industry label could identify some rental companies, but not reliably answer the actual product-fit question.
These results illustrate an underlying design lesson: discover candidates broadly enough to cover the market, then apply more specific evidence to determine commercial fit. The discoverable universe and the qualified audience are not the same thing.
Build a usable evidence-based ICP
- Specify the purchasing problem: who suffers it and what must be true for the offer to help?
- Choose the unit: account, site, facility, product or buying event.
- Define must-haves and disqualifiers: use observable criteria where possible.
- Choose sources: select evidence sources for each criterion, not simply one master company database.
- Assign states: yes, no, unknown and excluded or held, with reasons.
- Test real accounts: compare known good and bad examples before scaling.
- Review drift: update the ICP as products, customers and market conditions change.
Common mistakes
A common failure is to confuse enrichment completeness with qualification accuracy. A row with 30 populated fields can still describe the wrong business. Another is treating one AI classification as authoritative without exposing its source. A third is forcing every ambiguous company into a binary decision just to achieve a convenient export.
Use industry filters to improve discovery efficiency, not to replace the commercial qualification process. For specialized markets, also see finding prospects that sales databases miss and the planning questions to settle before automation.
Bring the whole system together.
See how TCA builds and transfers the research, qualification and activation infrastructure—not just an individual workflow.
Schedule a conversation