Cold email reply rates fell to 3.43% in 2025-2026, down from 5.1% in 2024. That's a 33% drop in twelve months. But top-quartile campaigns still clear 5.5%, and top-decile campaigns hit 10.7%. The market didn't die. It split.
If your reply rate is sitting at 2%, you're not unlucky. You're running 2024 tactics against a 2026 filter.
The market bifurcated, and most teams are on the wrong side
We pulled reply rate data across campaigns running through MiraReach in the last two quarters. The median sits at 3.4%. The top quartile clears 5.5%. The top decile clears 10.7%.
That spread is wider than it was eighteen months ago. In early 2024, the gap between median and top quartile was roughly 1.5 percentage points. Now it's over 2 points. The floor dropped faster than the ceiling.
What's driving the split? Two things.
First, spam filtering got smarter. Google and Microsoft tightened authentication requirements and started weighting engagement signals more heavily. A domain with no sending history and a list of scraped emails gets filtered before a human ever sees it. We wrote about the DMARC volume loophole closing earlier this year. That's part of it.
Second, inbox competition increased. More teams are sending. The same buyer who got 20 cold emails a week in 2023 gets 60 now. Attention is finite. The bar for "worth replying to" went up.
The teams still hitting 5.5% aren't smarter. They're running tighter infrastructure and doing more work per prospect.
Infrastructure is now table stakes, not a differentiator
Two years ago, you could send from a fresh domain with a basic SPF record and get delivered. That's gone.
Here's what we see across accounts that consistently clear 5%:
- Dedicated sending domains, warmed for at least three weeks before the first campaign
- DMARC enforced at p=reject, not p=none
- Separate domains for outbound vs. transactional email
- Volume capped at 30-40 emails per mailbox per day, not 200
- Reply-to addresses that route to a monitored inbox, not a noreply
None of this is clever. It's just the cost of entry now. Teams that skip it are sending into a void and wondering why their open rates look fine but replies are zero. (Open rates are unreliable anyway. Apple's Mail Privacy Protection broke that metric in 2021. If you're still optimising for opens, stop.)
The infrastructure work takes a weekend. The personalization work takes longer.
Personalization means something different now
In 2023, "personalized" meant inserting the prospect's first name and company. In 2026, that's the baseline. Every tool does it. Buyers expect it.
What actually moves reply rates now is relevance. Does the email reference something specific about the prospect's situation that proves you did five minutes of research?
We tested this across a sample of 1,200 emails to UK SaaS founders. Half used first-name-plus-company personalization. Half referenced a specific trigger event: a recent funding round, a job posting for a role that suggested a pain point, a product launch, a conference talk.
The trigger-based emails got 6.8% reply rate. The name-based emails got 2.9%. Same offer, same list quality, same sending infrastructure.
The difference is work. Finding a trigger event takes longer than pulling a name from a CSV. But it's the work that separates the top quartile from the median.
One customer running outbound to UK accountancy firms switched from generic "we help firms like yours" messaging to referencing specific HMRC deadlines and practice management software migrations. Reply rate went from 2.1% to 7.4% in six weeks. Same list. Same offer. Different opening line.
What doesn't work anymore
We've tested a lot of tactics that used to work and now don't.
Volume plays. Sending 10,000 emails a week from 50 mailboxes used to get enough replies to justify the effort. Now the filtering catches most of it, and the replies you do get are lower quality. The math stopped working.
Generic AI-generated copy. The first wave of AI email tools produced emails that read like AI emails. Buyers noticed. Reply rates dropped. AI is useful for research and drafting, but the final email needs a human edit or it reads like every other AI email in the inbox.
"Quick question" subject lines. They worked in 2022. Now they trigger spam filters and human skepticism in equal measure.
Follow-up sequences that don't add anything. If your follow-up is "just bumping this to the top of your inbox," you're training the prospect to ignore you. Every follow-up should add a new piece of information or a new angle.
The market is still there. The bar is just higher
3.43% median reply rate sounds grim. But 5.5% is achievable with disciplined infrastructure and real personalization. 10.7% is achievable if you're willing to do the research work per prospect.
The teams struggling are the ones still running 2024 playbooks. They're sending from domains that aren't warmed, using copy that doesn't reference anything specific, and following up with nothing new to say.
The teams winning are doing the opposite. They're treating infrastructure as a prerequisite, not a project. They're spending more time per prospect and sending fewer emails. They're using AI to find triggers and draft angles, not to write the final email.
If your reply rate dropped this year, the fix isn't a new tool. It's tightening the fundamentals and doing more work per prospect. The market rewards that now. It didn't used to.
What we'd do next
If you're sitting below 3%, start with infrastructure. Warm your domains, enforce DMARC, cap your volume. That alone will get you back to baseline.
If you're at 3-4% and want to break into the top quartile, the work is personalization. Find a trigger event for every prospect. Reference it in the first line. Make the email something only you could have sent.
If you want to see how MiraReach handles the research and drafting side, give MiraReach a try. We find the triggers, score the inboxes, and draft the emails. You still press send.
— Mira