Email outreach
Why cold email open rates are unreliable, and what to track instead
By Asen Levov, co-founder of Sellinger. Published . Last updated . 7 minute read.
An open is recorded when a tiny image in the email is loaded. Apple's Mail Privacy Protection loads it for the reader, security scanners load it before the reader sees the message, and many mail clients block it altogether. So the open rate can show opens that never happened and miss opens that did. For cold email, track what a person has to do: replies, positive replies, bounces, unsubscribes and meetings held.
This post explains how an open is counted, why the number stopped meaning what it used to, and what to measure instead. It does not give an industry average open rate, because an average of a number that is not reliable would not help you.
How is an email open measured?
The sender puts a one pixel image in the email. The image sits on a server. When the email is displayed, the mail program asks the server for the image, and the server records "opened". Nothing else is observed: not whether a human looked, not for how long, not whether the message was read.
That design had one assumption: the request comes from the person, at the time they read. Three things now break it.
What breaks the number?
1. Privacy features that load images for the reader. Apple describes Mail Privacy Protection as helping users "prevent senders from knowing when they open an email" and as masking their IP address (Apple Newsroom, June 2021, read 25 October 2026). In practice the mail app fetches remote images through Apple's servers whether or not the person looks at the message, so an opened status can appear for a message nobody read. How much of your list uses Apple Mail depends on your audience, and you can usually see it in your own data.
2. Security scanners. Many company mail systems open links and load images when a message arrives, to check for malware. The scanner is not a reader, but the pixel does not know that.
3. Image blocking. Some clients do not load remote images unless the reader allows it. A person can read the whole message and never be counted.
The result is an error in both directions. The number is high for some audiences, low for others, and you cannot tell which from the number itself.
Why does it matter more for cold email?
Because cold email is judged on action. A newsletter can live with a fuzzy open number because it has links and clicks to fall back on, and a long history to compare with. A cold email is a short note to one person, and the thing you want is a reply. A high open rate with no replies tells you only that your subject line worked or that a scanner ran.
There is also a cost. Tracking pixels and tracked links add elements that spam filters look at, and the sender rules from Google and Yahoo put weight on authentication, one click unsubscribe and a low spam rate rather than on opens. See Google and Yahoo bulk sender rules. Many cold senders therefore turn open tracking off and keep the email plain.
What should you track instead?
Five things, each of which a person did on purpose.
| Measure | What it tells you | Where it comes from |
|---|---|---|
| Reply rate | Whether the message and audience fit | Replies divided by people contacted |
| Positive reply rate | Whether the offer fits | Replies that show interest, divided by people contacted |
| Bounce rate | Whether your list and setup are clean | Messages that were not delivered |
| Unsubscribes and complaints | Whether you are being unwelcome | Opt outs and spam reports |
| Meetings held | Whether it matters | Meetings that took place, not only booked |
Write the denominator next to every number. "40 replies" means little; "40 replies from 800 people contacted" is a rate. The definitions are in define reply rate and the reply rate calculator does the arithmetic.
For an example of reporting with the base, see the Sellinger case study: it shows people messaged, replies and meetings, each with the count it was measured on, and says where the data comes from.
Can you still use opens for anything?
One careful use: comparing two versions of a subject line.
- Send both versions to similar people, on the same day, at the same hour.
- Compare the open rates, and also the reply rates.
- Treat a difference in opens as a hint only, and trust a difference in replies.
- Repeat before you change your default.
The reason to look at replies as well is that a subject line which gets opened by curiosity but does not lead to replies has not worked. This is a method, not a result: no number from it is a benchmark.
What would a worked example look like?
This is a made up example to show the arithmetic, not a customer.
You send a sequence to 500 people. The tool reports a 61% open rate. You also see:
- 25 replies (5.0% of 500)
- 9 positive replies (1.8%)
- 12 bounces (2.4%)
- 3 unsubscribes (0.6%)
- 4 meetings held (0.8%)
The 61% looks good, but it does not change what you do. The 0.8% meetings held does. If next week's version produces 7 meetings held from 500, that is a real improvement. If it produces a 70% open rate and 4 meetings, nothing has improved.
What should the agent do about opens?
An AI SDR should not decide anything from an open. Sellinger decides what to do next from the reply, from the state of the lead and from the follow up settings you chose: the default for email is up to five follow ups with a gap of 1 to 3 days, and every value can be changed per campaign. An out of office message is not counted as a reply, and neither is helpdesk mail; those are treated separately so that they do not inflate your reply rate. A reply, an opt out or a booked meeting changes the lead. An open does not.
A short checklist
- Decide that replies and meetings held are your headline numbers.
- Write the denominator beside each.
- If you track opens, use them only to compare versions sent at the same time.
- Check bounces and unsubscribes weekly; they tell you about your list and your tone.
- Do not quote an open rate as a benchmark, yours or anyone else's.
Opens were always an indirect measure. Now that they are also unreliable, the better habit is to stop at the person: did they answer, and did they come to the meeting?
What should you tell clients or a manager who asks for open rates?
Be direct and offer something better. A short note works: "Open rates are not reliable any more, because mail apps can load the tracking image without the person reading. We report replies, positive replies and meetings held instead, each with the number of people contacted." Then show the table above with their own numbers.
If someone insists, give the open rate with a caveat and put the reply rate beside it. Over a few weeks the reply rate will be the number they ask about, because it is the one that connects to meetings.
How do you set up tracking on a new campaign?
- Decide the three numbers you will report before you send: replies, positive replies and meetings held.
- Write the definition of a positive reply, for example any reply that asks a question about the offer or agrees to talk.
- Keep a simple weekly sheet with the base next to each number.
- Review after four weeks, not four days; replies keep arriving for a week or more after a send.
- Change one thing at a time, so you know what moved the numbers.
None of this needs open tracking. It needs honest denominators and the patience to wait for replies.
Common questions
Is the open rate of a cold email accurate?
Should I turn off open tracking?
What should I track instead of opens?
Can I still use opens to test a subject line?
What does Sellinger count?
How Sellinger helps
Read how it works on Read how analytics are organised, or see a real campaign in Read the Sellinger case study.
