Measuring email now that opens are broken

Email reporting gives you a row of percentages, most of which are either unreliable or beside the point. Three numbers matter, one of them has quietly stopped working, and one that is never displayed prominently is the one that can end your sending.

Why open rate broke

An open is measured by a tiny image loading when the message is displayed. Mail privacy features now pre-load images on the recipient's behalf, which registers as an open regardless of whether anybody read anything, and the share of your audience affected is not something you can see.

So open rate is inflated by an unknown amount that differs by audience, which makes it useless as a target and dangerous as a comparison. It still has one use: a sudden drop across all sends usually means a deliverability problem rather than a content one, and that is worth noticing.

The numbers that still mean something

• Clicks, as a share of messages delivered rather than of opens. A click is a deliberate act and is not pre-fetched in the way an image is.
• Replies. Rarely in any dashboard, and the strongest signal there is that a real person read a real thing. Count them by hand if you have to.
• Unsubscribes, watched per send rather than in total. A spike on one message tells you exactly what was wrong with it.
• Complaint rate, from the mailbox providers' own postmaster tools rather than from your sending platform. This is the number with consequences.
• What the email actually caused: orders, enquiries, bookings. Harder to attribute and the only one that is the point.

Comparing honestly

Two cautions that prevent most of the self-deception available here.

1. 01Do not compare your figures with published industry averages. They are aggregated across list sizes, sectors and collection methods that have nothing in common with yours, and the open rate component is now measuring something different for everybody.
2. 02Compare each send with your own previous sends, to the same segment, over a period long enough that one unusual message does not set the trend.
3. 03Expect a decay. A list that receives nothing new gets less responsive over time, and that is normal rather than a sign that something broke.
4. 04Record the figures somewhere dated. The platform keeps them until you change platform, and then the history is gone.

The question worth asking instead

Not how did this email perform. Rather: did anybody do anything differently because of it. On a small list that question is often answerable by reading the replies and looking at what happened in the two days after a send, which is less rigorous than a dashboard and considerably more informative.

Connecting the email to what happened

The number that matters is whether the send caused anything, and attributing that properly is harder than it looks because the person who reads an email on a phone often buys later on a computer.

1. 01Tag the links so your own analytics can see that the visit came from a particular send. A campaign parameter on each link costs nothing and is the minimum.
2. 02Look at the two days after a send rather than at clicks. Orders, enquiries and direct visits in that window, compared with a normal two days, is a cruder measure than attribution and closer to the truth.
3. 03Expect the sending platform's figures and your own to disagree, and decide which one you treat as the record. The platform counts clicks; your own side counts outcomes, and the second is the one worth keeping.
4. 04Count replies by hand and keep the useful ones. On a small list this is genuinely the highest-value output of sending.

And accept a limit honestly: on a small list, most sends will not produce a visible spike in anything, and that is not evidence they did nothing. The case for email at this size rests on the relationship over months rather than on individual sends, which is uncomfortable and true.