Test a campaign without risking a domain
Send the campaign from infrastructure that is not yours, at a sample large enough to mean something, and read what real recipients do with it. The complaints then land on the sender you borrowed rather than on the address your invoices and your customer mail travel on.
On this page
Testing a campaign is easy. Testing one without paying for the result twice is the part that stops people, because the usual way to find out whether a message works is to send it from the address the business depends on.
Why is testing from your own domain a trap?
Because the experiment and the asset are the same thing. A campaign that provokes complaints leaves a record on the domain your invoices, your password resets and your ordinary customer mail all travel through.
The damage is not immediate and it is not obvious, which is what makes it expensive. Weeks later ordinary messages start landing badly and nobody connects the decline to a campaign that finished long ago.
What is wrong with a small careful test?
The sample. Marketers who know the risk respond by testing to two or three thousand people, which is too small a number to separate a good offer from an unlucky day.
They then treat the result as evidence and scale it, which is the worst of both worlds: the caution bought nothing and the decision was made on noise anyway. A test that cannot be trusted is not caution, it is delay.
What does a real test look like?
Real recipients, a sample large enough to be read, and a sender that is not yours. Send the new campaign to a hundred thousand people and read how the market takes it, before your own domain has ever touched it.
Everything the send provokes lands somewhere designed for it. At PhantomFox the unsubscribes and the spam reports go to our sending infrastructure, so a campaign that goes badly costs you that send and never the address your business runs on. That is the reason to prove it here first rather than carefully at home.
| Way to test | What it can tell you | What it risks |
|---|---|---|
| Send to colleagues | the formatting works | nothing, and it proves nothing |
| Seeded inbox panel | a synthetic score | nothing, and almost nothing useful |
| Two thousand of your list | a number too small to trust | your own domain |
| A hundred thousand, off domain | the real response rate | that one send |
What should you measure?
Replies first, then complaints, then opens, in that order. A reply is a human deciding the message deserved an answer, which is the only signal that survives scrutiny.
Point the reply address at your own marketing inbox so the answers arrive where you already work and the results are yours to read directly. Complaint rate tells you whether the offer annoyed people, and opens tell you about the subject line and nothing else.
What happens after a good result?
You scale a message you have evidence for, rather than a message you hoped about. If the result was bad you have spent one send finding that out, which is the cheapest useful thing a marketer can buy.
The practical route is on how it works, the volume is on bulk sending, and the page for this exact situation is campaign proving at a hundred thousand. For background, read what a dedicated sender is and how to send bulk email without a domain, or start from the answers hub.