I was this close to refunding a client earlier this year. Results were awful, ideas running dry, and i was convinced cold email was cooked for good. A mate of mine looked at my setup and dropped the truth bomb: 'Your infra is the problem, not your copy.' He was spot on. i'd been wasting months tweaking subject lines on emails that never made it out of spam.

First thing i did was rip out the cheap shared inbox provider i was on - those IP pools were trashed from other senders. moved everything to puzzleinbox and mailforge, split across both, tested maildoso for one campaign. Never put all your inboxes on one provider - last year an entire network i used got rate-limited by Gmail for a week. split your risk across at least two.

just from the inbox change, reply rate shot from under 1% to over 1.2%. Didn't touch a single word of copy. Emails were finally landing in primary.

Next, lead data. apollo-only lists are basically poison in 2026. everyone uses them, you're hitting the same prospects with the same noise. started using Apollo as a base and enriching everything in Clay - real signals like recent hires, funding events, tech stack changes. cut list size by about 40% and replies went up, not down. From 1.2 to 1.5%.

open rates? they're a joke. apple privacy and Outlook prefetching killed them years ago. i wasted months optimising for opens. now i only look at positive replies, meetings booked, pipeline. If your dashboard leads with open rate, you're watching the wrong stuff.

sequences got shorter - was running six steps, cut to three. day 1 initial, day 3 bump, day 7 breakup. reply rate went up. long sequences from old playbooks are hurting your domain reputation way more than helping.

Openers were the last piece. 'Loved your post about X' is dead - every guru taught it, every prospect spots it. i pulled real signals from Clay and built openers around them: 'You just hired your second revops person in six months - usually means {specific pain} - we helped {similar company} with this.' Nothing clever, just specific. That got me from 1.5 to around 1.7%.

stack for anyone curious: inboxes - puzzleinbox and mailforge mainly, sending - smartlead, but plenty of alternatives, data - Apollo + Clay + Findymail core, verification - millionverifier, intent - common room, CRM - pipedrive.

Stuff I wasted months on: A/B testing open rates, 8-step sequences from a 2022 course, openers using surface-level LinkedIn data, buying scraped lists, chasing volume before fixing reply rate, panicking and rewriting copy when deliverability was the real issue.

If you're stuck under 1% reply right now, it's almost never copy. It's one of: emails not landing in primary, stale or wrong-fit contacts, following sequence advice from three years ago, or openers that look like everyone else's. took me way too long to figure out - hope this saves someone the months i lost