Tips
How to measure an email marketing campaign
Mail privacy features and security scans limit what opens and clicks can tell you. Compare delivery, replies and later activity, and state what your filtering cannot establish.
Correction, 23 September 2026: click timing and session differences are diagnostic clues. They do not identify every bot or establish a complete human count.
Apple loads remote mail content in the background when Mail Privacy Protection is active. An image request therefore does not prove that the recipient read the email. Security tools can also inspect links, so a click count needs checking against later activity.
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. Read the provider’s failure reason. Remove invalid addresses and investigate other failures before retrying. Do not treat every rejection as proof that the address is dead.
Unsubscribes and spam complaints. Honour opt-outs and track complaints against the receiving provider’s published limits. Keep the mail platform’s filtering rules with the report.
Replies. Review them for useful responses and exclude automated messages such as out-of-office replies.
Revenue or pipeline attributed to the send. The number that decides whether the programme continues.
Investigate clicks that may be automated
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.
Fast clicks across every link are clues, not proof of automation. Compare click timing with later activity and your mail platform’s filtering method.
Different click and session totals need investigation. Consent, blockers and session definitions can explain part of the gap. Neither count automatically becomes the truth.
Tag the links so the site can count them
Put consistent campaign parameters on links to your site. Compare destination activity with the email report. Consent choices, blockers and measurement setup can leave gaps in either source.
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
Opens can help diagnose changes within a consistently measured audience. Privacy features and client differences can still affect that comparison. Do not present a subject-line winner from opens alone as proof of more readers or revenue.
Keep the metric’s definition and filtering method with any historical comparison.
The numbers to report
Delivered, complaint rate, unsubscribe rate, filtered clicks, useful 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
Watch complaint rates alongside authentication, delivery errors and the audience being contacted. A complaint increase needs investigation. It does not establish a fixed recovery period or a single cause.
Set thresholds from the current provider requirements. Review each campaign and segment, and pause sending where the evidence warrants it. Record what changed before testing a restart.
Gmail publishes spam-rate requirements. Use the current provider rules when choosing complaint thresholds.