How to measure an email marketing campaign

Open rate has been unreliable since Apple began preloading images in Mail Privacy Protection. Click rate in B2B is inflated by security appliances that visit every link before the recipient sees it.

So start from what is still true, and build the measurement on that.

What is still reliable

Delivered. Accepted by the receiving server. Not the same as inbox placement, and the closest thing you have to a baseline.

Bounces, split by type. A hard bounce is a dead address and should be suppressed immediately. A soft bounce is temporary. A rising soft bounce rate is an early warning about reputation, and it moves before anything else does.

Unsubscribes and spam complaints. These are real human actions. Complaint rate is the number to watch most closely, because it is what mailbox providers use, and the tolerance is small.

Replies. Unambiguous, human, and under-counted because they arrive in an inbox nobody exports.

Revenue or pipeline attributed to the send. The number that decides whether the programme continues.

How to tell a machine click from a person

Corporate mail filters follow links to check them. They produce clicks with no subsequent page activity, often within seconds of delivery, frequently on every link in the message including the unsubscribe.

Three filters remove most of them. Ignore clicks that happen in the first few seconds. Ignore recipients who clicked every link. Compare clicks against sessions that actually landed on the page, which your site analytics counts independently.

If your platform reports forty clicks and your site saw eight sessions from that campaign, the eight number is the real one.

Tag the links so the site can count them

Put campaign parameters on every link. Then the destination can be counted by your own analytics, server side if you have it. That sees the visit regardless of what the email platform believes.

Use the same naming convention every time. A campaign report is only as good as the consistency of the tags, and this is where it usually breaks.

What open rate is still good for

Relative comparison within one campaign to one segment, over a short window. Subject line A against subject line B, sent to the same list on the same day, still tells you something.

It cannot tell you whether 34 percent is good, and it cannot be compared to last year.

The numbers to report

Delivered, complaint rate, unsubscribe rate, real clicks, replies, and pipeline. Six numbers, every send.

If a seventh is added because it makes the report look better, that is the moment the measurement stopped being measurement.

The one that predicts trouble

Complaint rate. It moves before deliverability collapses, and once deliverability collapses the recovery is weeks of reduced sending.

Set a threshold, watch it per campaign and per segment, and stop sending to a segment that crosses it. That single rule prevents the failure that takes longest to undo.