Your website is a starting hypothesis, not proof of who will buy. Turn its claims into testable account filters and a separate map of buying roles.
Start with the problem, not a list of job titles
A website can tell a research tool what you sell. It cannot tell you, by itself, which customers renew, which projects lose money or who finally approves the purchase. Those missing facts are why a polished AI-generated buyer profile can still produce a poor prospect list. Treat the first draft as a hypothesis that a salesperson must be able to challenge.
Start with one sentence: we help a particular kind of business solve a specific problem through a particular offer. For a fictional stock-planning consultancy, that might be: we help independent retailers reduce stockouts by rebuilding their replenishment process. This is narrower and more useful than targeting every company interested in growth. It also gives you a concrete reason to exclude companies.
Build a small evidence sheet
Read your homepage, service pages, pricing and strongest customer examples. For each claim, save the page address, a short description and whether it is evidence or an assumption. A named retail project is evidence that you have served a retailer. It is not evidence that all retailers can afford your offer. A price shown publicly is evidence of your offer, not of a prospect's budget.
Use three columns: what we know, what we think and how we will check. Put uncertain geography, size and technology requirements in the second column. This prevents the next person from mistaking a plausible model suggestion for established customer research. Remove unsupported customer logos or invented success metrics rather than using them to justify a target market.
Separate the account from the person
An ideal customer profile describes the business that fits. A buying-role map describes the people involved. HubSpot's Ideal Customer Profile Template distinguishes business fit using factors including industry, size, need, budget and geography. Use that distinction as a starting point, then replace broad labels with your own evidence. The worksheet below is our suggested method, not a universal scoring model.
For the fictional consultancy, start with retailers operating several locations in a region you can serve. Then map the operations lead who experiences the problem, the finance lead who challenges the return and the owner who may approve the work. In a larger account, those responsibilities may sit with different people. Searching for a CEO everywhere ignores that difference.
- Account rule: which businesses are eligible, and which are excluded?
- Problem owner: who deals with the operational consequence?
- Budget owner: who can approve the relevant spend?
- Evidence needed: what would confirm each assumption before outreach?
Translate your hypothesis into usable filters
A phrase like ambitious businesses is not a search filter. Replace it with observable information, such as business activity, supported locations or a defined employee range. Distinguish mandatory filters from preferences. If an integration is essential to delivery, make it mandatory. If a growing team is simply an encouraging sign, do not exclude every account where growth data is unavailable.
Keep a title dictionary rather than one exact title. Operations director, head of operations and retail operations manager may represent different levels of responsibility. Review actual responsibilities before treating them as equivalents. Also write down negative filters: competitors, existing customers, unsupported countries and business models your offer cannot serve. Exclusions are part of the profile, not an afterthought.
Audit a sample before buying contact details
Take a small sample of accounts and review them manually before scaling. Twenty accounts is a convenient working batch, not a statistically validated threshold. For each one, record fit, uncertainty and a reason. If many fail for the same reason, change the rule before enriching hundreds of contacts. If reviewers disagree, tighten the definition instead of averaging away the disagreement.
Track account fit separately from contact availability. A business can be a strong prospect even when no usable email is available. A verified email at an unsuitable business does not rescue a weak ICP. This separation helps you decide whether to improve the targeting, change the contact source or stop the campaign altogether.
Keep the profile editable
After real conversations, add the objections, buying triggers and exclusions you actually observed. Date each meaningful revision and keep the previous version with its campaign results. Otherwise a change in targeting can look like a change in message performance. Avoid rewriting the profile after every isolated rejection; look for a repeated, explainable pattern.
Sync Scala's website-to-profile workflow is designed to help organise this starting point. Review its output before approving downstream searches. A profile is useful when someone can explain why an account belongs on the list, not when the description merely sounds sophisticated.
Sources and further reading
Prepared with AI assistance and checked against the linked sources. Examples are illustrative, not customer results. See our editorial standards or request a correction.