800+ cold emails later — here's what actually moves the needle (and what's a complete waste of time)

Co-signing the plain-text finding hard — no logos, no signatures with four social icons, no tracking pixels you don't need. Plain text isn't just "performs the same," it actively avoids Promotions-tab classification.

One thing I'd add to the needle-mover list: the follow-up distribution. In most sequences I've seen, the majority of replies come on touches 2–4, not the opener. People aren't uninterested; they're busy. A 4-touch cadence (D1 opener / D3 one-line bump / D7 value-add / D14 breakup) roughly doubles the yield of the same list vs. opener-only.

And agreeing with your targeting point from the other side: agencies reply better than random SMBs because outreach is their native language — they grade your email professionally, and a good one earns professional courtesy.

Co-signing the plain-text finding hard — no logos, no signatures with four social icons, no tracking pixels you don't need. Plain text isn't just "performs the same," it actively avoids Promotions-tab classification.

One thing I'd add to the needle-mover list: the follow-up distribution. In most sequences I've seen, the majority of replies come on touches 2–4, not the opener. People aren't uninterested; they're busy. A 4-touch cadence (D1 opener / D3 one-line bump / D7 value-add / D14 breakup) roughly doubles the yield of the same list vs. opener-only.

And agreeing with your targeting point from the other side: agencies reply better than random SMBs because outreach is their native language — they grade your email professionally, and a good one earns professional courtesy.

Your 0.37% vs 2.0% split matches what I've measured from the other side, and I'd add one number underneath it. Before writing a single email I hand-checked 250 local business websites in four UK towns. Only 8 were broken in a way the owner would care about (domain no longer resolving, page 404ing, cert failing). So the list of small businesses with a provably bad website is about 30x smaller than scraped lists suggest, and most of the 820 you emailed probably had nothing wrong you could prove in one line. Agencies replied because your data was about their prospects, not about them. Second thing I'd separate: list quality from copy. In scraped small-business lists a chunk of addresses are pattern-guessed (info@, hello@) and never existed, and a chunk belong to domains that don't resolve at all, so they can't reply however good the email is. I ended up building the checker as a product (leaddrive.uk, free tier of 150 credits, no card) because doing it by hand for 250 sites took a week. Curious: on the agency batch, what share of the 149 had a published email on their own site vs one found elsewhere?

The scanning the prospect's website before writing part is the real test. We did outreach to 400+ agencies when we started Pick an Agency, and the ones that responded fastest were always the ones where the founder had clearly checked our site first,they'd say something like "I noticed you're doing reviews differently than Clutch" instead of generic copy. Also killed our fancy footer signatures and just use plain text,even less friction. The audience fit insight at the end is key too. You're right that agencies understand outreach better than most, so they respond to it when it's legit.