How to Kill Bad Cold Email Domains (and Why You Should)
We killed 500 cold email inboxes last week. On purpose. All at once.
Most people would call that insane. We call it necessary infrastructure management.
Here's the thing most agencies get wrong: they think more domains equals more replies. It doesn't. It equals more reputation damage spread across a bigger surface area. When you're running cold email at scale, bad domains don't just underperform, they actively drag down your good ones.
Let me show you the framework we built to identify which domains to kill, and why this approach improved results across every client we manage.
Why Most Cold Email Infrastructure Strategies Fail
The default playbook looks like this: buy 50 domains, warm them up for two weeks, start sending, add more domains when volume needs increase. Repeat forever.
The problem? Nobody's cutting the dead weight.
You end up with a portfolio where 30% of your domains are actively hurting deliverability, 40% are mediocre, and only 30% are actually performing. But you're paying for all of them. Worse, the bad domains create noise in your data. You can't tell if a campaign is underperforming because of copy, targeting, or deliverability issues.
We needed a system that could isolate domain health from campaign performance, track trends over time, and make kill decisions based on data instead of gut feeling.
How We Score Every Domain Across Our Client Base
We built a scoring system that ranks every inbox against its peers across multiple dimensions:
Core metrics we track:
- Reply rate (total replies divided by sends)
- Bounce rate (hard bounces only, soft bounces are noise)
- Positive reply percentage (replies that show interest)
- Out of office rate (signal of inbox health)
- Warmup health score (pulled from our warmup provider)
Every inbox gets a tier ranking: top 20%, middle 60%, bottom 20%. But here's where most people stop, and where we go deeper.
Why Campaign-Adjusted Scoring Changes Everything
Raw metrics lie. If you put a domain on a test campaign that's intentionally aggressive or targeting a harder audience, the reply rate will naturally be lower. That doesn't mean the domain is bad. It means the campaign is harder.
So we built a versus campaign comparison system. We compare each domain's performance not just against the client average, but against other domains running the same campaign.
This isolates deliverability from campaign quality. If a domain is underperforming its peers on the same campaign, that's a deliverability signal. If all domains on that campaign are underperforming, that's a campaign signal.
This distinction is critical. It lets you kill bad domains without killing good tests.
Multi-Period Trend Analysis: The Kill Signal
A domain having a bad week doesn't mean it's dead. A domain having a bad month does.
We track performance across 3-4 week rolling windows. We're not looking for one-time dips. We're looking for sustained decline.
Our kill criteria:
- Bottom 20% performance tier for three consecutive weeks
- Sub 0.5% reply rate with 300+ sends
- Zero positive replies, zero out of office replies
- Declining trend (getting worse, not stable)
- Domain renewal coming up within 30 days
When all five signals align, the domain gets flagged for termination.
In this case, 500 domains hit all five criteria. They weren't just underperforming. They were dead weight actively hurting the domains around them.
Why We Killed 500 Domains in One Day
When we looked at the data, the pattern was clear. These 500 inboxes had collectively sent over 150,000 emails with reply rates below 0.5%. No positive replies. No out of office replies. Every metric screaming "this inbox is not reaching the primary inbox."
Keeping them active had two costs:
- Direct cost: We were paying for infrastructure that wasn't working
- Indirect cost: Bad domains can hurt the reputation of domains on the same sending infrastructure, especially if you're using shared IPs or the same SMTP provider
The second cost is the one most people miss. Email providers look at reputation signals across multiple levels: domain, IP, SMTP provider, even behavioral patterns. A cluster of bad domains can create guilt by association.
So we killed them. All at once. Automated through our integration with InboxKit directly from the dashboard.
How to Build a Rolling Domain Replacement System
Killing domains is only half the strategy. The other half is having replacements ready.
Here's our current infrastructure model:
- Active tier: 80% of domains, actively sending
- Ready tier: 10% of domains, warmed up and ready to deploy
- Warming tier: 10% of domains, currently in 2-3 week warmup
Every month, we cut the bottom 10-20% of the active tier and promote domains from the ready tier to replace them. The warming tier moves into ready. We buy new domains to refill the warming tier.
This creates a constantly evolving infrastructure. You're always rotating in fresh domains and rotating out underperformers before they become a problem.
The benefits:
- You never run out of capacity when you need to scale
- You're always working with your best-performing infrastructure
- You can isolate problems to copy and targeting instead of wondering if deliverability is the issue
- You avoid the renewal trap where you're paying for domains you should have killed months ago
What Changed After We Cut the Dead Weight
Results improved across the board. Not because we magically fixed 500 domains. Because we stopped letting them drag down the domains that were working.
Reply rates on remaining domains went up an average of 0.3-0.5%. That doesn't sound like much until you multiply it across hundreds of thousands of sends. Positive reply percentage improved because we weren't diluting the data with dead inboxes.
But the bigger win was clarity. When a campaign underperforms now, we know it's the campaign. We're not second-guessing deliverability because we've already cut the bad infrastructure.
The Takeaway: Infrastructure is Not Set and Forget
Most agencies treat cold email infrastructure like a one-time setup. Buy domains, warm them up, send forever. That worked in 2019. It doesn't work now.
Email providers are getting smarter. Reputation signals are more complex. The difference between a domain that reaches the inbox and one that hits spam is often a slow decline over weeks, not a sudden drop.
If you're not actively scoring, tracking, and rotating your infrastructure, you're flying blind. And if you're not willing to kill underperforming domains, you're paying for dead weight that's actively hurting your results.
We cut the bottom 10-20% every month. You should too.
