Should you turn off open tracking in cold email? How to test it

My default cold email setup is plain text, open tracking off, and a custom tracking domain if I need link tracking at all. The reasons are real, but the proof is thinner than most vendor posts suggest, so here is what is established, what is not, and how to test it on your own list.

Why is open tracking unreliable?

Open tracking works by loading a tiny image from a server, and an open is recorded when that image is fetched. Apple's Mail Privacy Protection, launched in 2021, prefetches and caches messages including their images whether or not the recipient reads them, so the pixel can fire without a human opening anything. That is documented by email vendors such as SendGrid. I could not find Apple's own support page for it, so take it as well-documented vendor knowledge rather than a primary source.

The practical conclusion is that opens are a noisy number. I already made the point about judging subject lines by replies in cold email subject lines in 2026.

What does the evidence say about plain text and tracking?

Less than the claims suggest. Hunter's article "Is HTML harming your cold email deliverability?" analysed 2.2 million emails sent by Hunter users in 2025 and found HTML campaigns bounced more than plain text. Three things to know before quoting it:

• The article says it is not a controlled A/B experiment.
• It suggests plain text senders may be more experienced users who verify their lists, which would explain the gap on its own.
• Its own headline figure and its data section disagree with each other, and it does not separate tracking pixels from other HTML elements.

So the direction is plausible, the size is unknown, and the effect of tracking alone is not measured. That is the honest state of public evidence, which is why a test is better than a rule.

What is a custom tracking domain and when do I need one?

If you track clicks or opens, your tool rewrites links and inserts a pixel on a tracking domain. Instantly's help docs explain that without a custom one, that domain is shared among its customers. A custom tracking domain is a subdomain of yours (the docs use a CNAME pointing to the tool's tracking host) so the links belong to you.

Two clarifications. It does not affect SPF or DKIM alignment, which I cover in the DMARC alignment checklist. And it only matters if you track at all: with tracking off and plain text with no links, there is nothing to host.

How do I A/B test this on my own list?

Here is the test I would run. I am describing a method, not reporting a result.

1. Pick one variable. Test plain text against HTML, or tracking on against off, but not both at once.
2. Split one list randomly into two equal groups by sending domain and inbox, so each group gets the same mix of inboxes. Never send one version from your best domains and the other from your worst.
3. Hold everything else constant: the copy, the offer, the day and hour, the daily volume per inbox.
4. Pre-commit to metrics: bounce rate, spam-folder placement from seed accounts you control, positive replies per hundred sends. Not opens.
5. Set the sample size first. A few hundred sends per group is too few to read a small difference, as the maths in how many cold emails it takes to get one positive reply shows.
6. Run it long enough to cover the same weekdays on both arms.
7. Decide what a win is before you look. If the difference is smaller than the noise between two identical groups, call it a tie and pick the simpler setup.

How do I measure placement without opens?

Opens cannot tell you where a message landed. Use these instead:

• Seed accounts. Send to Gmail, Outlook and Yahoo accounts you control, and record inbox, promotions or spam by eye. Do this before and during the test.
• Bounce rate per arm, with the 2% line from why your cold email lands in spam as the alarm.
• Replies per hundred sends, split into positive and not.
• Google Postmaster Tools spam rate, if you send enough volume for it to populate. Google's sender guidelines say to keep it below 0.30%.

If the two arms are identical on all four, the setting does not matter for you, and you can choose on convenience.

What can go wrong with the test?

• Different inboxes per arm. The inbox is the biggest source of variation, so spread both arms across the same inboxes.
• Peeking early. Small samples swing wildly. Fix the sample size before you start.
• Changing two things. If plain text and tracking-off change together, you cannot say which did the work.
• Judging by opens. The exact thing you turned off will make one arm look worse for no real reason.

What is my default configuration?

How do I set up a custom tracking domain?

The steps follow Instantly's docs, and other tools use the same idea with different names:

1. Add a CNAME record on your sending domain. In Instantly's example the host is inst and the value is prox.itrackly.com.
2. Wait for DNS to propagate. The docs say this can take 24 to 72 hours.
3. Enter the subdomain (for example inst.yourdomain.com) in the inbox settings and run the status check.
4. Confirm with a CNAME lookup that it resolves to the value your tool gave you.

Check your own tool's documentation for the exact values, since they differ between tools. Set it up on each sending domain, not just one, or the unmatched domains will still use the shared one.

The short version

Plain text, no open tracking, and a custom tracking domain if you track clicks. Then test it once on your own list, because the public numbers are vendor-reported and uncontrolled.

Where this advice stops working

If you need click data for a specific offer, you will track links, and then the custom tracking domain matters. And if authentication is broken, a plain text email with no tracking will still land badly. Fix the basics first with why your cold email lands in spam and the deliverability audit. Provider differences are in Google Workspace vs Outlook for cold email.

For more on reading results without opens, see the outbound numbers nobody shows you.