"Just find your ICP" is the most useless piece of advice in SaaS. Everyone says it. Nobody explains how to do it when you have zero customers, zero data, and no market research budget.
This is that explanation.
Why most ICP work fails before it starts
The classic mistake: founders build the ICP from their assumptions about who should want the product. They describe someone who sounds good on paper — "marketing managers at mid-market B2B companies" — but it's not rooted in any real signal.
Then they try to reach that persona and get no response, because they're not the right person, or they are the right person but the messaging doesn't connect, or the channel is wrong. The ICP feels like it's failing, but really the problem is that it was invented rather than discovered.
An ICP is discovered, not designed. Here's how to discover yours.
Step 1: Start with the problem, not the person
Before you can profile a customer, you need to deeply understand the pain you're solving. Not "they need better marketing" but: what does the day look like when this problem is at its worst? What does it cost them (time, money, stress)? What have they already tried and failed at?
Write this down as concretely as possible. Be specific enough that you'd be embarrassed to show it to someone who doesn't have this problem — that level of specificity is what makes ICP work actually useful.
Step 2: Find where people talk about this pain publicly
The internet is full of ICP research if you know where to look:
- Reddit threads: search for your problem in 3-4 candidate subreddits. Read the comments. Note the language people use.
- LinkedIn posts: search your problem keyword. Find the posts with real engagement (not just likes — actual comments). What job titles are engaging?
- Job listings: companies hiring for roles related to your problem are actively feeling the pain. If you see "hiring: Head of X" — that company probably has an X-shaped problem right now.
- Review sites (G2, Capterra): search your category or your competitors. The 3-star reviews are gold — they explain exactly what the product does wrong and what people wish existed.
You're not trying to confirm your hypothesis. You're trying to find patterns you didn't expect.
Step 3: Build a 5-field profile (not a persona)
Personas with stock photo headshots and fake names ("Marketing Mary") are useless. A functional ICP has five fields:
- Job title or role — specific, not broad. "Head of Growth" not "marketing professional."
- Company profile — industry, size range, business model (SaaS, agency, e-commerce).
- Pain trigger — what specific event or situation means they need this now? New funding round? Team size change? Competitive pressure?
- Status quo — what are they using today to solve this problem? Spreadsheets? A $300/month tool that only half-works?
- Unblocking event — what needs to be true for them to pay for your solution instead of keeping the status quo?
Six sentences. That's a functional ICP. It's more useful than a twelve-slide market research deck.
Step 4: Test before you commit
Once you have a candidate ICP, test it with ten outreach attempts before you treat it as validated. Personalised, specific outreach, targeting the exact person you described.
If you get a 0% response rate across ten attempts with genuinely good, personalised messages: the ICP or the pain framing is wrong. Adjust and try again.
If you get 2-3 conversations: you're pointing at something real. Keep going.
ICP validation is a number game at the start. The goal is to move fast through bad hypotheses until you find the one where the conversations feel effortless because the pain is obvious to both sides.
The compounding benefit
Once your ICP is real — meaning it's discovered from evidence rather than assumed — every other marketing decision gets easier. Your SEO targets the right queries. Your cold emails have hooks that land. Your Reddit presence resonates because you're speaking to the right people in the right way.
This is why ICP work isn't a branding exercise. It's the foundation that determines whether your distribution actually works.
Runlo's onboarding starts with a structured ICP discovery session — guided by Warden, tuned by your answers, upgraded as you get real signals. No guesswork.