The three terms "ICP," "persona" and "TAM" get used interchangeably. They are not the same thing. Each answers a different question, and outbound goes wrong when one stands in for another.
This post separates the three by the question each answers and by the data each one needs. For full definitions, see the glossary pages on the ideal customer profile (ICP) and the buyer persona. We do not give a TAM formula or a TAM figure here.
Three questions, three jobs
This post uses the three terms in the following way:
- The ICP asks which companies.
- The buyer persona asks which people inside those companies.
- The TAM asks how many exist.
The first two shape who you contact. The third is a count, and a count is not a target list.
What the data looks like
The split shows up in the data you hold. Company data and people data are different records.
An ICP is built from company records and deal outcomes. In HubSpot, Outlevo reads company name, domain, owner and lifecycle stage. It also reads deal pipeline, stage, amount and closed-lost reason [[1]](#sources). In Salesforce, it reads accounts and opportunities with their stages, for the ICP and win/loss analysis [[2]](#sources). A CSV import of up to 200,000 rows maps your columns to domain, name, country, size, industry and deal fields such as stage, amount, outcome and loss reason [[3]](#sources).
Persona work lives in people records. Contact fields such as email, name and job title are optional. They are imported only after an admin turns contact import on with a stated purpose [[1]](#sources). A job title tells you who a person is. It says nothing about whether their company fits.
Where teams mix them up
Using a persona as the ICP. A list of job titles is not a list of companies. Filtering by title alone can pull in people at companies that never bought from you. Decide the company profile first, then ask who inside those companies matters.
Using the TAM as the ICP. Counting every company that could buy is a different exercise from describing the ones that did. A large count can sit next to a very narrow set of real customers. Outlevo's win/loss analysis counts companies in your deals [[4]](#sources). A market-size number is a different thing.
Treating the ICP as an opinion. An ICP written from memory is a hypothesis. Outlevo marks guesses as hypotheses and never presents them as findings [[5]](#sources). Do the same in your own documents. Label what comes from won deals and what is a guess.
Treating fit as permission. A company can match your ICP and still be one you should not email. Outlevo policies set who can be contacted, on which channels, how often and with what budget [[6]](#sources). The ICP says who fits. The policy says who you may reach. On the legal side, the FTC's CAN-SPAM guide states: "The law makes no exception for business-to-business email." [[7]](#sources) We read that on October 22, 2026.
Expecting a persona to work everywhere. Some channels never see a person. Outlevo's website-visitor script shows which companies visit and whether they fit your ICP, and it never identifies a person [[8]](#sources). LinkedIn Ads audiences in Outlevo are built from company names and websites, not from people [[9]](#sources). In Outlevo, both work at the company level.
What changed
The ICP question, which companies, can start from your own deal history.
As of the October 19, 2026 release, you can import past deals from CSV or your CRM and get a win/loss analysis that proposes ICP hypotheses [[10]](#sources). In HubSpot, Outlevo learns your ideal customer from the deals you actually won and lost [[1]](#sources).
Three rules shape how the analysis reads those deals:
- It counts companies, not deals [[4]](#sources). One company with five closed deals is one company.
- It labels small samples. Under 10 won companies is very low confidence [[4]](#sources).
- AI-proposed ICP segments are labelled experimental until a person approves them [[4]](#sources).
The same release added a free ICP assessment on outlevo.com that reads your public site and asks which countries you sell into [[11]](#sources). Outlevo never proposes markets [[4]](#sources). The country list is your answer to give.
How each one feeds outbound targeting
The ICP decides which accounts qualify. Each qualified account gets a Readiness Card with fit, intent and data quality scores [[5]](#sources). A card needs at least three observations, each with a source and a date [[5]](#sources). Facts read from a website are kept only when their quote is found on the page [[5]](#sources). Qualification costs credits: light qualification costs 1 credit and full qualification costs 3 credits [[12]](#sources).
If you connect HubSpot, Outlevo writes the fit score, the ICP version and the next action onto each company record, in fields it owns [[1]](#sources). Your standard HubSpot fields are never edited [[1]](#sources).
The persona decides who you write to inside a qualified account. It shapes the message and the choice of contact. It comes after the company decision, not before.
The TAM stays a planning number. Keep it in your planning documents, next to the ICP and separate from it.
What to do
- Gather your past deals. Export won and lost deals from your CRM, or import a CSV with company domain, outcome and loss reason. The CSV and Excel import maps those columns for you.
- Check the sample before you trust it. If you have fewer than 10 won companies, treat the result as very low confidence [[4]](#sources). Add more history or rely on the hypotheses only as a starting point.
- Approve or reject each proposed segment. Until a person approves it, a proposed segment stays experimental [[4]](#sources).
- Write the persona last, and keep it separate. List the roles inside approved companies. Contact import is optional, so decide whether you need it and state the purpose [[1]](#sources).
- Keep the TAM out of the targeting step. Write the number down once, somewhere else.
- Set the contact rules. Decide who can be contacted, on which channels and how often, and keep fit and permission as separate checks.
If you have no deal history yet, the free ICP assessment is a place to start. It reads up to 5 public pages of your website [[13]](#sources). It returns facts labelled as verified observations or AI inferences, plus up to 5 experimental segment hypotheses [[13]](#sources). Treat them as hypotheses to test against real deals.
For a step-by-step method, read How to Define an Ideal Customer Profile.