Finding Leads in Your Connections' Engagement

Some links in this piece are affiliate links, each marked (affiliate). If you sign up through one, I earn a commission. I only ever link tools I actually run.

We naturally distrust strangers.

But if that stranger is a mutual friend, we build trust much faster.

Based on this, we made a theory: if you reference a post of a mutual connection, can you simulate this intro without facilitating tons of intros.

It worked. From an initial test to now, we're seeing

• 40-55% acceptance
• 15-22% real reply rate
• 10% reply immediately
• ~1 meeting booked per 20 connection requests.

And this little page shows how it works.

How to build it:

1. Download your connections. LinkedIn, Settings, Data privacy, Get a copy of your data. One of the files is Connections.csv. Google it if that menu has moved again, because it does.
2. Feed those profiles to Apify (affiliate). The actor is harvestapi/linkedin-profile-posts. Profiles in, their posts out, plus everybody who reacted or commented.
3. Run it weekly on last week's posts, no reposts. Claude Code will turn the CSV into the format Apify wants.
4. Add it up. Every 1,000 connections gives us about 1,100 posts and 2,000 reactions and comments. Posts cost around $2 per 1,000, and people who never post still cost a little, so set a cap. We run the example above at $150 a month.
5. Filter down to your buyers. Apify gives you each person's headline. Filter on that first, then look up real job titles for the ones you can't tell. About 1% make it through.
6. Send the ones who make it to HeyReach (affiliate) for LinkedIn, or Salesforge (affiliate) for LinkedIn and email.
7. Write the connection note. Something like: "Hey Tim, saw your reaction to Max's post, I help revenue teams go from exploring Clay to getting ROI from it. May we connect and I can share more?"

Steps 4 to 6 live in Clay. It matches each reaction back to the post it came from, so you know which of your connections led you there, and then it runs the filter.

The filter is the actual job

Apify gives you a headline, not a job title.

"CFO Advisor" is not a CFO, and "Helping CFOs scale" is usually a good clue that you are not one. Match on the headline alone and you will spend your whole week on people who sell to your buyers instead of your buyers.

So we filter cheapest first. A free formula on the headline catches the obvious ones. Anybody still unclear gets their real job title looked up. Whoever is unclear after that gets one pass from a small, cheap AI model, which might be wrong, but by then it is only handling the leftovers.

Match on the title or the AI check and you go to the outreach tool. Everybody else gets dropped before we pay for them twice.

Some things we learned

Who wrote the post matters more than how big the post was.

For one customer, a lead that came off a post written by one of their buyers cost a penny or two. The same lead off anybody else's post cost 17 to 20 cents. We were spending 74% of the budget to find 25% of the people, which is not our best trade.

Which makes sense once you say it out loud. People read the posts of the people they work with.

We started at 15 reactions per post, and 15 was just a number we picked.

When we finally plotted it, the counts fell off normally from 1 to 14, and then over a thousand posts landed on exactly 15. That was not a pattern we had found. That was our own setting looking back at us.

So we raised it to 40, and 98 people we had never seen before were sitting past position 15.

Then we tested the obvious version we should have tested first. Check who wrote the post, and only pay for reactions on the posts written by our buyers. It cost a third as much and found nearly the same people. Cheaper for the same result almost never happens. Fun.

The part I like

Your buyers tend to engage with each other.

So everybody who accepts is not just a lead, they are a new connection. Pull your connections again next month and their posts come along too, and the people reacting to those posts are more of the same kind of person.

The lead list grows, and your real network grows with it, in the direction you actually wanted. A few months in you are not hunting your buyers from the outside. You are in the room with them.

Bonus: find out when somebody started caring

The scraper above starts with a post. harvestapi/linkedin-profile-reactions starts with a person, and hands you six months or a year of everything they reacted to. There is a matching one for comments.

I ran it on four people I already knew were good buyers. Six months each, $0.41 total. One had 29 of her 66 reactions on my topic. One had 102 reactions and only 3 that were close. One had reacted three times in six months, so nothing I built was ever going to find her.

Then give the whole pile to Claude and ask what the good ones have in common. It is much better at seeing that than I am, and it finds things I would not have thought to look for.

Whatever pattern it comes back with, that is the thing you build into Clay.

Two small ones while you are in there. Somebody who never posts returns no posts, which is a correct answer and not a broken run, and you can still pull everything they reacted to. And none of these need a LinkedIn cookie, which is usually the part that breaks.