Today’s plays › All plays › Prospecting & research
Build your whole market list wide first, then let AI check every website
Picked and checked by Olga Pearson, SalesUpLevel
Get tomorrow’s play in your inbox. Free ↓
25,763 companies checked and 14,710 kept as the market. 198 of 200 in the final spot check were correct (99%). 1,591 shops showed a hiring signal and 743 made the first campaign list. Tool cost was $230.
Database filters miss real customers and let in look-alikes. Jan’s team searched wide (every label their best customers use, ZIP code by ZIP code on Google Maps, plus manufacturer partner pages), removed duplicates, then had AI read every company’s website and answer one yes/no question with a quoted line of proof. Only after that did they look for signals and contacts.
By Jan Berning, Head of Growth and GTM engineer, writing an X article · Jan Berning on X ↗ · Oct 22, 2026
If you sell: run it before your 10am
- Pick 10 to 20 of your best customers.
- Use prompt 1 to write a short note: what they do in their own words, one yes/no test question, and the false matches to exclude.
- Collect every company that could fit from every source, filtering only by location and size. Remove duplicate websites before you pay for anything.
- Run prompt 2 on each company’s website. Tune it until it matches your own judgment on two rounds of 10 in a row, then check a random 200.
- Inside the companies that pass, find a buying signal (they used hiring), then the people and verified emails.
- Read 10 rows by hand before anyone sends, and mark current customers and open deals “do not contact”.
Prompt 1:
Here are our best customers: [websites]
Here is what we know about them: [call transcripts, CRM notes, won deals]
Read all of it and give me:
1. What they do, in their own words, and which products or brands they name
2. Who their customers are
3. The Google Maps categories and search terms that would find them
4. Directories, associations or manufacturer lists they appear in
5. One yes/no test question that every one of them passes
Write "not found" wherever the material doesn’t say.
Prompt 2:
You are checking one business for a market list.
Question: [your yes/no test question]
Answer NO if: [your false matches]
Input: name: {name}; category: {category}; website_text: {homepage + services page, first 4,000 characters}
Return JSON: answer ("yes" | "no" | "unclear"), evidence (the exact phrase behind the answer), confidence (0 to 1), no_reason (the first "Answer NO if" item that applies, or "none").
Rules: 1. Use only the input. If the text doesn’t say it, answer "unclear". 2. Never decide from the name alone. 3. Quote evidence word for word.If you lead a team: roll it out
- Have one person build the market list once and keep it in one table, then pull each month’s campaign from it.
- Report all three numbers together (whole market, signal, campaign list) so the team sees why the first campaign is small.
Watch out for
- It’s a big build. A seller can start with step 1 and prompt 2 on a list they already have.
- The article ends with “comment TAM and I’ll send you the checklist,” so it’s also lead generation. The numbers are his own.
- Some sites block scripts and only look dead. Open them in a real browser before you drop them.
- Cold email rules differ by country. Check yours before you send.
Source: Jan Berning, Jan Berning on X, Oct 22, 2026. Results are as reported by the source; we did not measure them.
Get tomorrow’s play before your first call
Free. One sourced play a day.