Every lead tool I've ever used treats a lead like a fact. Name, company, problem, score. It sits in a table until you get around to it.
Watching thousands of real outreach conversations changed my mind. A lead isn't a fact — it's a moment. Someone is mid-problem, annoyed enough to post about it in public. That window closes fast.
The pattern I kept seeing:
Reply within a few hours of their post — they're still in the tab, still annoyed, still looking. It feels like two people talking.
Reply to the same person a week later — you're a stranger digging up their old laundry. Same fit. Same message. Completely different reception.
So I changed how scoring works: recency now outranks keyword match. A decent-fit thread from this morning beats a perfect-fit thread from last Tuesday. That felt wrong to build — you're deliberately skipping "better" leads — but reply rates said otherwise.
One exception worth knowing: Reddit's tail is longer than you'd think. A day-2 comment can still convert if the thread ranks on Google, because new people with the same problem keep landing on it. LinkedIn is the opposite — the feed buries anything without instant traction, so the mid-problem window barely exists there.
If you do any kind of outreach, manual or automated, here's the free takeaway:
— Sort by newest, always. Not by best match.
— Work in small windows — last 24-48 hours. Treat anything older as dead inventory.
— Found a great thread from last week? Don't DM the author. Answer the thread publicly instead — you're now writing for the next person who Googles it.
The uncomfortable part: this means your pipeline can't be a to-do list you clear on Fridays. Timing isn't a detail of the product. Timing IS the product.
I built LeadSynth around this — it watches Reddit, X, LinkedIn and YouTube in real time and surfaces people who are mid-problem right now, not last month. First leads are free, no card: https://www.leadsynthai.app
Question for anyone doing outbound: how old was the oldest lead you ever actually closed? I want to know if the half-life idea holds outside my own data.