How do I define my ideal customer when I have no customers yet?
Start from the problem, not the product: write down who feels it most, what they do today instead, and what event makes them act, then test that guess on 20 real companies. With no customers, look at people complaining about the problem and at companies already paying for alternatives. GrowFast's scan_domain proposes a profile from your site for you to correct.
Published by GrowFast. Last updated 2026-11-13. We make GrowFast, so weigh our side accordingly. Markdown version: /solutions/define-your-ideal-customer-without-customers.md.
Why it happens
Early founders describe the customer by the product's features ("anyone who sends invoices") or by market size ("every SMB"). Both are too broad to write a single specific email to. Without customers there is no data to narrow it, so the profile stays a guess, and guesses sent to the wrong companies produce silence.
What you do have are proxies: public evidence of who has the problem. A competitor's customer page, a Reddit thread, a job post that lists the pain as a duty. Treat the profile as a hypothesis with a test date, not a finished document.
The fix that works without GrowFast
- State the problem in one sentence and name who has it worst. For example: "agencies with five or more clients lose a day a month reconciling invoices". The person with the problem is your target, not the person who might sign.
- Collect proxies. Read the competitors' customer logos and case studies, search Reddit and forums for the problem in people's own words, and look at job posts that list the pain as a duty. Each gives you a company type and a role.
- Write three hypotheses. Each one is a role, a company type, a size and a trigger event. Three is enough to compare; one is a bet.
- List 20 companies per hypothesis and email five of each. Small enough to write by hand, large enough to see a pattern. Track replies and what they said, including the no's.
- Write an exclusion list. Say who you will not target and why. An agency may match every keyword and still never buy a product built for in-house teams.
- Date the profile and revisit after 20 sends. Keep it to one page. Replace guesses with what the replies told you.
| Field | Too vague | Specific enough to act on |
|---|---|---|
| Company | SMBs | B2B SaaS with 5 to 50 staff and a public pricing page |
| Role | Decision-makers | Founder, or head of growth where there is one |
| Trigger | Needs our product | Just raised, is hiring a first marketer, or just launched |
| Evidence | Our gut | Named on a competitor's customer page or in a thread asking for this |
| Exclusions | None | Agencies and other tools in our own category |
When to use GrowFast
scan_domain reads your product's website and returns an ideal customer profile (title, industry, company size, pain points), a first email, and best-fit companies scored 1 to 10, each with the page its contact address was read from. It takes about 30 seconds, so it is a fast way to get hypotheses to react to.
A scan reads a website, so it can aim a tool for outreach at other outreach agencies. In autopilot you can state who to aim at and who to skip in your own words (the targeting field); that text outranks what the scan inferred, and scans drop companies below a fit score of 6, and autopilot only writes to companies scored 8 or better.
When it is not the right tool. Once you have paying customers, their real roles, industries and sizes beat any inference from a website. Use them first and treat the scan as a cross-check.
Plans and limits are on the pricing page.
Ask your AI assistant
Paste this into an assistant that has the GrowFast MCP server connected. The tools it uses are on the ChatGPT and Claude connectors, which research and draft but cannot send.
Assign to GrowFast: scan mynewsletter.com and tell me which kinds of companies would pay for it, with the 5 best fits and why each scored what it did. Show me what you did so I can review and correct the profile.
Tools used: scan_domain
FAQ
How many ideal customer profiles should I have?
One per product to start, with up to three hypotheses inside it. Split into separate profiles only when replies show two groups that need different emails.
How do I know the profile is wrong?
Twenty carefully written emails with no replies and no useful objections suggests the profile or the offer is off. Change one of them, not both, and test again.
Can software pick the profile for me?
It can propose one from your site and your competitors, which is a good starting hypothesis. Whether it is right is decided by who replies.
Where do I find companies once I have it?
See the guide on finding B2B leads from a domain, linked below.