We pay for every lead we pull. Then, before a single email goes out, we systematically delete somewhere between a quarter and 40% of them.
Clients occasionally push back on this. "We paid for 1,000 leads, why are we only emailing 650?" It feels wasteful. It's actually the single highest-leverage step in the entire campaign, and in this article I'll show you exactly what we cut, why, and what it does to the numbers.
A Bad Lead Costs More Than a Good Lead
The instinct to email everyone comes from thinking of a lead's cost as the $0.20-0.50 you paid for it. Sunk cost, might as well send.
But the real cost of a bad lead is incurred when you email it:
- A dead address bounces. Bounce rates above 2-3% tell every mailbox provider you're a spammer, and your deliverability drops for the whole campaign, including the good leads.
- An irrelevant prospect marks you as spam. Spam complaints are weighted far more heavily than bounces. A handful can poison a domain.
- An unqualifiable lead gets generic copy. If there's no data to personalize with, that lead drags your reply rate down and trains inbox algorithms that people don't engage with your mail.
- A wrong-fit meeting wastes the most expensive thing you have: your calendar. A booked meeting with someone who could never buy costs an hour and produces nothing.
Every lead you shouldn't email makes the leads you should email perform worse. Deliverability is shared across the list; one segment's bounces are everyone's spam folder.
The Pipeline: Enrich, Then Prune
Here's the exact sequence we run on every list. The order matters: enrichment comes before pruning, because the enrichment results are the evidence we prune on.
Stage 1: Verified Emails Only at the Source
This happens at pull time, not after. Every lead database offers "guessed" or "unverified" emails alongside verified ones. We never export them. A guessed email is a coin flip on a bounce, and per the math above, bounces are campaign-level damage.
If a prospect is genuinely valuable but their email is unverified, run it through a separate verification tool and keep only the ones that pass. Everything else stays out of the sheet.
Stage 2: Enrich Every Lead With LinkedIn and Website Data
Before judging any lead, we gather two pieces of evidence for every row:
- Their LinkedIn experience data: current role, tenure, and career history, scraped from the profile.
- Their website text: the actual copy from the company's site, scraped and stored next to the lead.
This isn't decoration. The LinkedIn data becomes the raw material for personalized first lines, and the website text becomes the input for qualification in the next stage. A lead without evidence can't be qualified or personalized, which brings us to the first cut.
Stage 3: The Missing-Data Cut
Any lead missing either piece of enrichment gets deleted. No LinkedIn profile found, or a website that's down, parked, or empty: gone.
This feels harsh until you think about what missing data actually signals. If the LinkedIn scrape can't find the person, the odds are good they've changed jobs and the database is stale, which means the email is heading to a dead inbox or the wrong company. If the company website doesn't resolve or has no real content, you're often looking at a defunct business or a shell.
The missing-data cut isn't discarding good leads; it's using enrichment failure as a cheap detector for stale ones. This stage typically removes 10-20% of a purchased list, and bounce rates drop accordingly.
Stage 4: Qualification From Website Text
Next, every surviving lead gets classified using its website text: does this company actually fit the ICP? We rate each one (fit, unclear, or not a fit, with a stated reason) and delete the clear misses.
Why bother, when the database already filtered by industry? Because database industry tags are wrong at a rate that would shock you. Companies self-select flattering categories, tags go stale after pivots, and "software" turns out to mean a two-person web design agency. On typical lists we find 10-20% of leads are simply not what the filter claimed once you read their actual website.
Reading a thousand websites by hand isn't realistic, which is why we automated this step. But even a manual spot-check of 50 random leads will tell you your list's true mislabel rate, and a VA with a clear rubric can work through a full list in a day or two. The rubric matters more than the tooling: define in one sentence what a "fit" company looks like (who they sell to, what they do, roughly how big), and classify against that sentence only.
Stage 5: Suppression and Dedupe
The last cut is against your own history. Before any send, remove:
- Existing customers and open opportunities (nothing torches trust like cold-pitching a current client)
- Anyone who has replied to a previous campaign, positively or negatively
- Anyone contacted in the last 60-90 days from any of your campaigns
- Unsubscribes and do-not-contact requests, permanently
- Competitors and partners
This requires actually maintaining a suppression list as a living asset: one master file of every address you've ever contacted, updated after every campaign. Most teams skip this, and it shows up as the embarrassing double-contact email six weeks later.
What the Math Looks Like
A representative run: 1,000 purchased leads. Verified-only export and the missing-data cut take it to about 820. Qualification removes another 130 mislabeled or wrong-fit companies: 690. Suppression and dedupe leave roughly 650 sendable leads.
Send generic-ish copy to the raw 1,000 and a realistic outcome is a 1-2% reply rate with a shaky bounce rate, call it 12-15 replies with maybe 5-6 positives. Send deeply personalized copy to the clean 650 (personalization you can only do because every lead has enrichment data) and 4-6% replies is normal: 26-39 replies, with a higher positive share because everyone on the list can actually buy.
Roughly double the meetings, from 35% fewer sends, plus healthier domains that keep performing next month. That's the trade. The list didn't shrink; the illusion did.
When Not to Prune: The Small-TAM Exception
Everything above assumes a big addressable market, where leads are cheap relative to deliverability and attention. There's one situation where the rules change: when your total addressable market is tiny.
If only 300 companies on earth can buy what you sell, deleting 100 of them because a scraper couldn't find their data is malpractice. In small-TAM campaigns, enrichment failure changes from a deletion trigger into a work order: leads missing LinkedIn or website data get routed to manual research instead of the trash. A person spends five minutes per lead finding the right contact, confirming the company is alive, and pulling personalization material by hand. At 100 leads, that's a couple of days of work protecting a third of your entire market.
The qualification stage still applies (a company that genuinely doesn't fit is still a delete, no matter how small the TAM), but the missing-data cut inverts. The general principle: the scarcer your leads, the more labor each one deserves before you give up on it. Prune aggressively when leads are abundant; repair aggressively when they're scarce.
This is also the test for whether your data vendor is the problem. If more than about a quarter of a fresh list fails enrichment, the issue isn't your pipeline; the list was stale at purchase. Track the failure rate per source and move your spend toward whichever database keeps its data freshest for your specific market. The differences between vendors, per niche, are much bigger than their marketing suggests.
Two Objections, Answered
"But I paid for those leads." You did, and that money is spent whether you email them or not. The only question that matters now is whether emailing a given lead increases or decreases expected meetings. For the categories above, it decreases them. That's sunk cost logic working exactly as it should.
"More volume means more meetings." Only when quality is held constant. Volume times reply rate equals results, and reply rate isn't independent of volume: junk volume degrades deliverability, which lowers the reply rate on everything else. Past the quality line, added volume is negative-sum.
The Checklist
If you take one thing from this article, take the sequence:
- Export verified emails only; verify or discard the rest.
- Enrich every lead with LinkedIn data and website text.
- Delete any lead missing either.
- Qualify from website text against a one-sentence ICP; delete clear misses.
- Suppress customers, past contacts, repliers, and unsubscribes.
- Send to what's left, with personalization built from the enrichment.
For the upstream half of this process (defining the ICP and pulling the list in the first place), see How To: Build a Lead List That Books Meetings.
Feel free to message me if you have questions!