Experiment 005, infrastructure
Compare your campaign to itself.
A tool called two of our sending domains burned on three replies. Do not kill a domain because a label says so. Here is what we look at instead.
Four domains, four labels, three replies
Some of our fresh inboxes had sent two emails each when our deliverability tool marked them underperforming. After two emails. Not long after, it marked two of our sending domains as burned.
| One campaign | The tool's label |
|---|---|
| Domain 1 | Healthy |
| Domain 2 | Underperforming |
| Domain 3 | Burned |
| Domain 4 | Burned |
All four labels rested on three replies across the four domains.
Why the label cannot tell
The tool has its own algorithm, but its support told us reply rate is the biggest part, and they expect above 1%. With three replies across four domains, who gets one and who gets zero is basically luck. To really tell two domains apart, you need thousands of emails on each.
Why not a fixed 1%
Under 1% after 200 emails sounds fair. But your reply rate depends on your targeting, your offer, your list, everything. We tested that rule on our own history: it flagged 97% of domain days, the healthy domains too.
What we do now: the campaign against its own start
Let a campaign run its first four weeks with the same list, the same inboxes and the same copy. That is its baseline. Count replies the same way every time (see 004). Then every week, compare the last two weeks to that baseline.
| Last two weeks vs the first four | Read |
|---|---|
| Above half | Fine |
| Below half | Red flag, look closer |
| Below half, again and again | Probably done |
It is the fairest comparison we know, because it is your own number, not somebody else's.
To be clear: you need some volume for this. In the first days, no number means anything.
Try this
Take your campaign's first four weeks and write down the reply rate. That is your line. Then every week, check whether the last two weeks fall below half of it.