Illingworth says a reply rate drop rarely needs new domains

Richard Illingworth posted the sequence he runs when a reply rate slides, and the point of it is that buying domains comes last, if at all. His example is a rate going 4%, 3.9, 4.1, then 1.9, which he says "almost NEVER" needs fresh domains, even though buying them is the first thing most people do when they see the drop.

Read the drift at domain level

Step one is to stop looking at single inboxes. He runs three inboxes per domain, takes the median of those three and plots that instead, because one inbox swings too much week to week to tell you anything. A wobble inside roughly 1% he calls healthy.

Step two is comparison. Put the underperforming domain next to the other domains in the same campaign. If they are all down together, the campaign is the problem. If only one domain is down, the problem sits in its infrastructure.

The bland email test

Step three is a deliberately boring send. Log into one of the sending accounts and email a Google inbox and a Microsoft inbox you own, subject "weekend", body "how was your weekend?". If something that plain does not land, Illingworth says you are looking at an infrastructure issue rather than copy or targeting.

Rebuild before rebuy

Step four is the part that replaces the purchase. Illingworth says Google assigns the sending IP through the admin panel an account is built in, so he moves the slipping accounts onto a new Google admin panel and they pick up a different IP. There is no full warmup again on the rebuilt accounts, and he says 2 to 3 days is the most it needs.

Richard Illingworth

@RCIllingworth

X

Google assigns the sending IP through the admin panel an account is built in.

This is one operator's process, not a dataset, and the no-warmup claim on rebuilt accounts is his own read. It also depends on the first step working. The median trick only means anything if you run several inboxes per domain, which he does at three, otherwise a single inbox's weekly noise looks like a death spiral and you end up buying domains to fix a number that was never real.

It sits alongside the lifespan argument he made earlier, where the fix for aging infrastructure was also process rather than replacement.