Writing Great Lead-Discovery Prompts

Writing Great Lead-Discovery Prompts

The system's quality is bounded by your prompt. A vague prompt yields a vague mix; a sharp prompt yields hundreds of qualified leads. This is the same principle as a Google search — but the system is doing more reasoning behind the scenes, so your prompt should be a sentence, not a keyword string.

The 3-part formula

Every great prompt has three parts:

1. Role — who you want to talk to (CTO, Founder, Owner, Director of …)
2. Industry / Category — what their business is about (SaaS, fintech, dental, marketing agency, law firm)
3. Geography — where they are (city, region, country)

Optional 4th part: intent signal (raising Series A, hiring, recently launched, exhibiting at X conference).

Examples by ICP shape

Tech founder / SaaS:

Indian SaaS founders raising Series A
Series B fintech CEOs in Germany
Healthtech founders in Tel Aviv hiring engineers
B2B SaaS CTOs in Bangalore who recently launched a product

SMB / local services:

Dental clinics in Berlin Germany
Coffee shops in Sydney Australia with espresso machines
Law firms in Toronto Canada
Yoga studios in Lisbon

B2B service providers:

Marketing agencies in London UK
Custom software consultancies in Singapore
Real estate agencies in Dubai serving expats
Architecture firms in Mumbai

Tech engineers / individual contributors:

Senior backend engineers in Berlin using Go
Mobile developers in Brazil with React Native experience
ML engineers in Bay Area working on LLMs

B2G / public sector:

Companies bidding on IT tenders in Karnataka
EdTech vendors winning state govt contracts in Indonesia
Renewable energy contractors in California with state ITAs

What works well

• Specific geography — "Berlin Germany" > "Germany" > "Europe"
• Industry + role together — "SaaS founders" > "founders" alone
• Intent signals when relevant — "raising Series A", "hiring", "recently launched"

What to avoid

• Boolean-style searches — founder AND saas AND series-a NOT investor — the system isn't a SQL engine; use natural language.
• Database operators — industry:fintech or location:bangalore — these don't apply.
• Over-broad prompts — "founders in India" will return investment firms, students, and stealth-mode hobbyists. Add an industry.
• Listing 10 things — "founders, CEOs, CTOs, and VPs in SaaS, fintech, healthtech in Bangalore, Mumbai, Delhi" — pick one role + one industry + one geography per query.

Iteration tips

If your first run yields fewer than 30 candidates, try:

1. Broaden the geography one level. "Vilnius" → "Lithuania". The system will still favor Vilnius but will widen if needed.
2. Drop the intent signal. "Indian SaaS founders raising Series A" → "Indian SaaS founders".
3. Try a different role. "VP Engineering in Bangalore" may be sparse; "CTO in Bangalore" often more populated.
4. Run the same prompt twice. Some sources have rate-limit windows; a second run 30 minutes later may pick up additional candidates.

Multi-language prompts

The system understands prompts in any language, but the queries it generates are mostly English (because GitHub bios, LinkedIn titles, and news outlets index in English even for non-English geographies). For best results in non-English markets, write your prompt in English with the local geography name:

• ✅ "Pizzerias in São Paulo Brazil" (English)
• ⚠️ "Pizzarias em São Paulo" (Portuguese — works but yields fewer results)