496 emails, 7 interested, and a fight over infra cost
The best reply rate published this week came off 496 emails to local shops for a mural artist, with no signals and no variables in the copy. Meanwhile Mars and Taylor Haren spent two days arguing whether sending 1M emails really costs 10 cents each.
Landing in the inbox
Illingworth says parallel warmup took inboxes from 3-4 months to 8-12
Richard Illingworth says his inboxes used to be replaced every 3 to 4 months and now run 8 to 12, and credits never letting providers see a jump in behaviour. In his build each inbox sends 10 cold emails and 20 warmup emails a day, with warmup starting the same day the campaign does and running alongside it.
the warm-up starts the same day the campaigns do and runs next to them the whole time.
Campaigns and numbers
The week's best reply rate came off 496 emails to local shops
Of the campaign breakdowns published this week, the strongest rate came from the smallest list: Nick Abraham says a mural artist got 18 replies and 7 interested from 496 emails. No LinkedIn keyword signals, no clever variables, a raw list of local businesses and proof in the PS. Christian's landscaping build, by contrast, ran 1,000 emails a day at one interested lead per 576 emails.
A campaign like this will often beat a typical sales tech campaign, because nobody else is emailing these owners about this.
Copy that ran
Berman and Lieben both say the promise does the work, not the copy
Two operators landed in the same place this week: the line that carries a campaign is the offer, and most senders tune everything around it instead. Alex Berman's test is whether a 12 year old could repeat your promise after one conversation, with "We book qualified sales calls for B2B software companies in 30 days" as the version that beats five bullet points.
The highest ROI line in your campaign is the offer, and it's the one line nobody tests.
Workflows and tooling
Christian's Hermes agent gets written up in 14 parts
The Slack agent Christian has been posting about all month is now documented end to end by his Head of GTM, including the architecture and the briefs the team sends it. The writeup covers a $9 a month server that stays on 24/7, a 5 row sample that catches a bad brief before it hits 5,000, 69,204 businesses mapped for one client on 19 data points, and 5 reply agents.
how raw exports go from 50-60% ICP fit to 90-95%
Where they disagree
Mars says Haren's infra sheet prices backwards from what he sells
Mars accused Taylor Haren of building a 10 cent per email infrastructure number by taking a setup his own sheet prices under 3 cents and multiplying it by four. Mars says that implies 20,000 inboxes and 10,000 domains for 1M sends, plus 5 cents of data on every email including follow-ups, when a 3 step sequence to 1M emails is about 330k people.
Keep every one of your assumptions, just drop the 4x and stop buying a new lead for every follow up and you're at ~54k.
Frequently asked questions
How long do inboxes actually last before they need replacing?
According to Richard Illingworth, inboxes can last 8 to 12 months when you run warmup emails alongside cold emails without letting email providers see sudden behavior changes, compared to the typical 3 to 4 month replacement cycle.
What gets better reply rates in cold email, targeting or the offer?
Two operators this week found that the offer itself carries the campaign more than copy refinements do. Alex Berman tests whether a 12 year old could repeat your offer in one conversation, suggesting a clear promise works better than detailed bullet points.
Do I need advanced personalization to get replies from local lists?
The best performing campaign this week used a raw list of local businesses with no LinkedIn keyword signals or personalization variables, suggesting offer novelty matters more than targeting sophistication for local outreach.
What's the real cost difference between running email infrastructure yourself versus using an agency?
The gap between doing it yourself and paying an agency comes down to redundancy assumptions (extra inboxes and domains you might not actually need) rather than the core hardware cost itself.