Worked example: one intent, one day, one post
Intent: "people struggling with manual LinkedIn prospecting", French and English.
Anchors: the short and long names of the main paid LinkedIn search product, two automation tools the buyer would name, "prospects froids", "liste de prospects", "prospect list", "cold outreach".
Buyer description: anyone doing their own B2B prospecting. Exclude vendors, agencies, consultants, recruiters, sales reps at prospecting tools.
Search
Twenty 2 to 3 word queries, both languages, sorted by date, last 24 hours: 638 posts. The sixteen pain-phrase queries written first had found 18 posts and missed the target.
Engagement floor
reactions + 2 x comments above 20: 45 posts. Eleven above 100 reactions.
The post that ranked first
"Help, I pay 100 a month for [the paid LinkedIn search product] and still build my lists by hand." French, written by a founder, 96 reactions, 80 comments. Found by the two-word short-form query for the product; the full product name did not find it.
Bucket: rant from someone who has the problem (first person, present tense, a number, a plea). Score against the intent: 92. Kept.
The people
150 unique engagers.
- Hard gates first: vendors, agencies, consultants and reps at prospecting tools removed. That is where most of the 150 went.
- Commenters scored against the intent. Five above 70. Examples of what scored:
- "Same, and I still export to a spreadsheet every Monday" (intent 88)
- "Great post" (intent 5, dropped)
- "We solved this with our tool, DM me" (vendor, gated out before scoring)
- Reactors kept on fit only: role and seniority matching the buyer description.
- Author: read the profile. A founder doing their own sales and venting about it. Kept, and the best lead on the thread.
Yield: 22 fitting people, 5 with explicit intent, plus the author. Against the strict ICP used for people search (Head of Sales, 11 to 200 employees, SaaS, one country), the same post yields 1.
The approach, for the five with intent
Day 0: visit the profile, react to the post, reply to their comment with one specific sentence about what they wrote ("exporting every Monday is the part that never gets automated; how big is the list by now?"). Day 2: invite, with a note that names the thread in under 200 characters. No pitch in either. The conversation opens itself; the product comes after they answer.
What went wrong the first time
The first pass used the strict people-search ICP on engagers and reported "1 lead from a post with 150 engagers", which reads as "this play does not work". The play was fine; the fit rule was wrong for the source. Engagers are found by intent, so they are read with a loose fit.