4.8% replied without the pixel. 0.8% with it on. Same copy. Same day. Natural experiment.
We turned off open tracking on half a send and left it on for the other half. Everything else was identical: same list, same subject line, same body, same sending account, same time window. The only variable was the tracking pixel.
This is not a thought experiment. This is production data from a live campaign.
The Numbers
| Metric | Tracking Off | Tracking On |
|---|---|---|
| Reply rate | 4.8% | 0.8% |
| Positive reply rate | Higher (measure replies) | Lower |
| Bounce rate | Same | Same |
Source: Open tracking experiment post
The bounce rate did not change. Deliverability did not improve because we fixed DNS. The only thing that changed was whether the recipient's inbox loaded a 1x1 pixel when they opened the email.
Why the Pixel Hurts
Google and Microsoft both show a version of this message when a tracking pixel fires:
"Images in this message are hidden. Mark as spam?"
On our test, 100% of Google inboxes served through Zapmail showed this warning when tracking was on. The recipient sees a yellow banner asking if they want to mark your email as spam. Before they even read your copy.
That is not a deliverability problem you can fix with SPF and DKIM. That is a UX problem you created by asking their mail client to load an external image.
What We Measure Instead
We measure:
- Reply rate — did they respond?
- Positive reply rate — did they respond positively?
- Bounce rate — did the email reach a real inbox?
We do not measure opens. Opens are a vanity metric that actively hurts reply rates.
If you are optimizing for open rate, you are optimizing for a number that (a) is inflated by bot scans and (b) triggers spam warnings in Gmail and Outlook.
Subject Line Variation: May 25
We ran a subject line test on May 25 with the same framework: half the list got variant A, half got variant B. No open tracking on either side. We measured replies.
The winner was not the subject line with the higher open rate in previous campaigns. It was the subject line with the higher reply rate. We would not have known that if we were tracking opens, because the open-tracking cohort was contaminated by the pixel warning.
Spintax: 5-10x "I Was Wrong"
We used to spintax subject lines 5-10 ways and pick the winner by open rate. That was wrong.
"Spintax 5-10x — I was wrong."
Spintax adds variance that makes A/B testing meaningless. If you are spinning "quick question" into 10 variants, you are not testing a hypothesis. You are adding noise. We stopped spintaxing subjects and started testing one variable at a time, measured by replies.
Aged Domains: 3-4 Months
Another variable we tested: domain age. Fresh domains with open tracking performed terribly. Aged domains (3-4 months) with tracking off performed significantly better.
"Aged domains 3-4 months."
Domain age matters. But domain age + open tracking is worse than domain age + no tracking. The pixel is the multiplier that kills you.
We saw reply rates move from 0.41% to 3.34% on a separate campaign when we fixed list quality and turned off tracking: 0.41% to 3.34% post.
FAQ
Should I ever track opens? Not for cold email. If you need engagement data on a nurture sequence to existing customers who opted in, maybe. For cold outbound, no.
What about click tracking? Same problem. External links loaded through a redirect trigger similar warnings. Use naked URLs when possible.
How do I A/B test without open data? Split the list 50/50. Same everything except one variable. Measure replies after 500+ sends per variant. Replies are the only metric that matters.
Does this apply to Instantly / Smartlead / Lemlist? Yes. The pixel is the pixel. The inbox warning is the inbox warning. Turn it off in your sequencer settings.
What I Would Do This Week
- Turn off open tracking in your sequencer
- Turn off click tracking (or use naked URLs)
- Pick one campaign and re-run the same copy to a fresh 500-row segment
- Measure replies at day 7
For the copy framework we use on those campaigns, see One Cold Email Template, Annotated.
Apply for a free test campaign at coldoutbound.com.
Key Takeaways
- 4.8% reply without pixel vs 0.8% with it — same copy, same day
- Google shows "Images hidden. Mark as spam?" on 100% of tracked sends through Zapmail
- Measure replies, not opens — opens are a vanity metric that hurts deliverability
- Stop spintaxing subjects — test one variable, measure replies
- Aged domains (3-4 months) + no tracking beats fresh domains + tracking
- Reply rates can move from 0.41% to 3.34% when you fix list quality and kill the pixel