Email Marketing Goals: KPIs That Actually Matter
Which numbers tell you something, which changed meaning after Mail Privacy Protection, and how to pick three to report on.
On this page
The short version
- Open rate stopped being a clean measurement when mailbox providers began pre-fetching images on behalf of users. It is now a mixture of human opens and machine opens, in a ratio you cannot see.
- Click-to-open rate inherits the same contamination, because its denominator is opens. Click rate against delivered is the cleaner sibling.
- Report three numbers, not fourteen. The rest belong in a diagnostic view you open when one of the three moves.
Most email dashboards were designed when an open meant a person loaded an image. That assumption stopped holding when Apple's Mail Privacy Protection began fetching images for its users regardless of whether the message was read, and other providers apply their own pre-fetching and proxying. The number still appears, still moves, and no longer means what the label says.
This does not make open rate useless. It makes it a compound metric — some real opens, some machine opens, in a proportion that varies with how much of your list uses Apple Mail. It is still usable for spotting sudden drops, which usually indicate delivery problems rather than subject line problems. It is not usable for deciding which subject line won a test.
What each metric is actually measuring now
| Metric | What the label suggests | What it now measures | Safe use |
|---|---|---|---|
| Open rate | How many people read it | Human opens plus automated image fetches | Spotting sudden drops |
| Click rate | How many people clicked | Clicks against delivered, less bot clicks from scanners | Comparing sends |
| Click-to-open | Interest among readers | Clicks over a contaminated denominator | Directional only |
| Delivery rate | How many arrived | How many were accepted by the receiving server | Bounce diagnosis |
| Inbox placement | Where mail landed | Placement across a seed panel, not your list | Trend, not absolute |
| Unsubscribe rate | How many left | Exactly that — one of the few undamaged numbers | Direct interpretation |
| Complaint rate | How many marked spam | Complaints reported back by providers who report | Direct, with a hard ceiling |
| Revenue per send | What it earned | Whatever your attribution model credits it | Only against itself |
Delivery rate and acceptance are not the same as reaching the inbox. A message can be accepted and filed in spam, and your provider will count that as delivered.
Pick three numbers to report
A report with fourteen metrics does not get read, and worse, it lets anyone find a number that supports whatever they already wanted to do. Three numbers force a position.
For most programmes the three are: click rate against delivered, complaint rate, and whatever your one commercial outcome is. Click rate tells you whether the content is working. Complaint rate is the constraint — cross it and nothing else matters, because delivery collapses. The commercial number is why the programme exists.
The constraint number
Complaint rate is the one metric with a hard operational ceiling rather than a target to beat.
0.1%
The rate major providers treat as the line
One complaint per thousand delivered
0.3%
Where filtering typically becomes visible
Placement degrades before you get a warning
Weeks
Typical time to recover reputation
Damage arrives faster than repair
The diagnostic set, kept separate
Everything else belongs in a second view you open when one of the three moves. Bounce breakdown, placement by provider, unsubscribe by segment, click map, device split, time-to-open distribution. These answer why, and they are worth having, but they are not a report.
Keeping them separate solves a real problem: a monthly report that includes diagnostics invites the reader to interpret them, and most people reading a marketing report have no basis for interpreting a soft bounce distribution.
What to do about historical comparisons
If your open rate history spans the point where pre-fetching became widespread, the series is not continuous and comparing across it produces confident nonsense. Mark the break in your reporting and treat the two periods as separate datasets.
The cleanest response is to rebuild your baselines on click rate, which has a much smaller discontinuity, and to keep open rate only as an anomaly detector. A twenty-point overnight drop in opens still means something is wrong; a three-point drift over a year probably means your Apple Mail share changed.
Frequently asked questions
Should we stop reporting open rate entirely?
Keep it in the diagnostic view as a drop detector, but take it out of the headline report. Leaving it in the headline invites decisions that the number cannot support.
What is a good click rate?
Your own last twelve sends are the only benchmark worth acting on. Published averages combine lists of wildly different sizes, industries and consent standards, and the spread inside any one industry is larger than the gap between industries.
How do bot clicks affect click rate?
Security scanners follow links in email, which inflates clicks — usually noticeably in B2B. Look for clicks registered within a second of delivery and on every link in the message, and exclude those user agents where your platform allows it.