Deliverability
Jun 8, 2025
Stop Tracking Open Rates in Cold Emails (They're Lying to You)
Bots and security filters trigger tracking pixels, so most reported opens never happened. Worse, the tracking itself can push your email into spam.
Open rates in cold email are unreliable by construction. Tracking pixels fire from spam scanners, security gateways, preview panes, and privacy proxies that pre-load images on the recipient's behalf, while genuine opens go uncounted whenever images are blocked. The pixel also costs you inbox placement. Remove open tracking from cold campaigns and score on replies and meetings instead.
How Open Tracking Works, and Why It Lies
Open tracking embeds a small invisible image in the email, hosted on a domain the sender controls. When a mail client requests that image, the sender's system logs an open.
Notice what is actually being measured. Not "a person read this". Only "something, somewhere, requested an image". Every inference beyond that is assumption, and in cold outreach the assumption is wrong a great deal of the time.
That was a tolerable approximation fifteen years ago, when the thing most likely to request an image was a human looking at a message. It is not a tolerable approximation now, because a large and growing share of image requests are made by machines acting on the recipient's behalf, before and regardless of whether the recipient ever looks.
Where Fake Opens Come From
Five sources, and between them they can account for a substantial share of reported opens on a cold campaign:
Privacy proxies that pre-load images. The largest single distortion. Apple Mail Privacy Protection, introduced with iOS 15, routes remote content through Apple's proxy and loads it in advance so the sender learns nothing about the recipient. The side effect is that the pixel fires for every message delivered to those users, whether or not it is ever opened. Given Apple Mail's share of business email on phones, this alone is enough to make the metric unusable.
Anti-spam and malware scanners. Security systems fetch remote content to inspect it. That is their job, and it logs as an open.
Email security gateways. Corporate filters commonly pre-render messages before delivery, firing every pixel in the process.
Preview panes. A message rendered in a reading pane while someone scrolls past counts identically to one that was read carefully.
Link and content prefetchers. Some clients and browser extensions fetch resources ahead of interaction.
The bias is not random, which is the part that matters. Random noise averages out across a large enough sample. This noise is systematic: it correlates with which mail client the recipient uses and what security their employer runs, which means it varies by segment. Enterprise recipients behind a security gateway will appear more engaged than a founder reading on a laptop with images off, purely as an artefact.
And the Real Opens You Never Count
The error runs in both directions. Opens go missing whenever the recipient's client does not load the image: images disabled by default, plain-text preference, privacy-focused clients, or a corporate policy that strips remote content.
So the number you see is inflated by machines and deflated by privacy settings, in proportions you cannot measure and that differ per segment. It is not a metric with a known margin of error. It is a metric whose error you have no way to characterise.
The Pixel Also Costs You Deliverability
Even if open rate were accurate, collecting it has a price.
Filters weigh signals that distinguish personal correspondence from bulk mail, and a tracking pixel is one of them. Specifically, what draws attention is an invisible image, loaded from a third-party domain that has nothing to do with the sending domain, in a message that is otherwise plain text. Genuine one-to-one email does not contain that.
The same applies to click tracking, which is worth stating plainly because the earlier version of this article recommended click-through rate as an alternative. Click tracking rewrites your links to route through a redirect domain. In cold outreach, a rewritten link pointing at an unfamiliar tracking domain is a stronger negative signal than the pixel, not a safer substitute. If you are removing open tracking for deliverability reasons, link wrapping has to go with it.
So the trade is: accept a measurable reduction in inbox placement in order to collect a number you cannot trust. Put that way, it is not a close call.
What to Measure Instead
Use signals that require a human to do something deliberate:
Reply rate. The primary metric. A reply is unambiguous: a person read enough to respond. It cannot be faked by a scanner.
Meetings booked. The outcome the campaign exists for, and the only one that connects to revenue.
Positive reply rate. Replies split by sentiment. More useful than raw replies, because a campaign can generate plenty of replies that are all variations of "not interested", and that is a targeting result rather than a copy result.
Unsubscribe and opt-out rate. Honest negative feedback about relevance, and free.
Bounce rate and complaint rate. Not engagement metrics, but the health signals that determine whether any of the above is even possible.
For deliverability specifically, stop trying to infer placement from opens and measure it directly. Seed-list placement testing and Google Postmaster Tools tell you where mail is landing, which is the question opens were being used as a poor proxy for.
The Incentive Problem
The strongest argument against open rate is not that it is inaccurate. It is that optimising for it actively degrades campaigns.
Open rate responds to curiosity. Vague, intriguing subject lines lift it reliably. Replies respond to relevance, which is a different and harder thing. So a team scoring itself on opens is being steered, week by week, toward subject lines that create an expectation the email cannot meet. The recipient opens, sees a generic pitch, and leaves. Opens up, replies down.
That is how teams end up with a metric improving and a pipeline shrinking, and why the dashboard offers no clue about what went wrong. The measurement is not merely noisy; it points the wrong way. Choosing what to measure is a decision about what your team will spend the next quarter getting good at.
The same logic applies to how you evaluate subject lines in the first place: subject line formulas are worth testing, but score them on replies.
What You Give Up
Being honest about the cost, because there is one.
Open rate is a fast signal. Replies to a cold campaign accumulate over days; opens appear within hours. Dropping tracking means losing an early indicator, and you will notice a badly broken campaign slightly later than you would have.
You also lose resolution in testing. Because reply rates in cold outreach sit in low single digits, distinguishing two variants on replies requires materially more volume than distinguishing them on opens. Some tests that felt conclusive before will now, correctly, be inconclusive.
Both are real, and neither changes the conclusion. A fast, high-resolution measurement of the wrong thing is not worth more than a slower measurement of the right thing. The practical mitigation is to watch bounce rate and complaint rate for the fast signal, since those move within hours and actually mean something.
Making the Switch
Record your current baseline first for replies, positive replies, and meetings per thousand sent. Without it you cannot tell afterwards whether anything changed.
Disable open tracking and link wrapping together in your sending platform. Half the change gets you half the deliverability benefit and all of the metric loss.
Redefine the campaign scorecard before the next launch, so nobody is still reporting the number you removed.
Set up direct placement measurement via seed testing and Postmaster Tools, to replace what opens were being used to guess at.
Compare replies per thousand sent across the change, not raw counts, and give it enough volume to mean something.
FAQ
Are open rates completely useless? For cold outreach, close to it. For opt-in email to an audience that expects you, open rate carries more signal because the machine-generated share is more stable and you are usually tracking relative change within one audience. The case against it is specific to cold email.
Does Apple Mail Privacy Protection affect all my opens? Only messages delivered to users who have it enabled, but because those opens register on delivery rather than on reading, they inflate your total and you cannot separate them out. A single unattributable bias of that size is enough to make the aggregate meaningless.
Can I track opens without hurting deliverability? Not meaningfully. The mechanism is a remote image load, and that is the part filters weigh. Serving the pixel from your own domain slightly reduces the signal and slightly increases the risk to that domain's reputation instead.
If I stop tracking opens, how do I know my emails are arriving? Measure placement directly instead of inferring it. Seed-list testing shows which folder mail lands in per provider, and Google Postmaster Tools reports Gmail's own view of your domain reputation. Both answer the question better than opens ever did.
Should I remove click tracking too? In cold outreach, yes. Rewritten links routing through an unfamiliar redirect domain are a stronger negative signal than the tracking pixel. If you need to know whether a specific link is working, use a distinct landing page rather than a wrapped link.
My agency reports open rates. Should I ask them to stop? Ask what they optimise against. If open rate drives their subject line decisions, they are being steered toward curiosity over relevance, and that shows up in your reply rate a month later. Reply rate, positive reply rate, and meetings booked are the numbers to hold them to.
Want campaigns measured on replies and meetings rather than phantom opens? Lidgen runs human-reviewed outreach on monitored infrastructure and reports on the outcomes that connect to pipeline. Book a demo.