reachable
← Back to blog
Deliverability

Open rates are broken

01/08/2026 · 121 Group, Melbourne · 8 minute read

The open rate at the top of your campaign report is not a count of people who read your email. It has not been one for years. If you are still using it to decide who to suppress, when to increase sending volume, or which subject line won, you are making real decisions on a number that a machine partly generates.

What actually happens when a mailbox "opens" your email

An open is recorded when a tiny tracking image inside the email body gets requested from the sending platform's server. That was always a rough proxy: it misses anyone who reads with images off, and it double counts anyone who reopens a message. But it correlated well enough with human attention to be useful.

Apple's Mail Privacy Protection changed that. When it is enabled, Apple's infrastructure pre-fetches remote content in incoming mail, including the tracking pixel, routing it through a proxy. The open is recorded whether or not the recipient ever looks at the message. It happens on Apple's schedule, from Apple's network, not from the recipient's device.

Two things break at once:

Apple is not alone in pre-fetching or proxying remote content, and mailbox providers change these behaviours without announcement. The honest position is that you cannot tell, per contact, whether a recorded open was a person or a machine. Any decision that depends on telling them apart is unsafe.

The decisions that quietly break

Sunsetting and list pruning

The classic rule is "suppress anyone who has not opened in six months". With inflated opens, contacts who genuinely stopped caring keep looking active, so they stay in your sending pool and keep dragging your engagement signals down. The rule now does close to the opposite of its purpose.

Warmup and volume ramps

Ramping a new sending domain by opening the taps whenever "opens look healthy" is the riskiest version of this mistake. Opens can look excellent while your bounce rate climbs and your complaints tick up, which are the signals mailbox providers actually respond to. You increase volume on a false green light and land the whole programme in the spam folder.

Subject line tests

A subject-line A/B test measured on opens can now be decided by which variant happened to reach more Apple mailboxes. That is not a creative insight. Measure the test on clicks, or on orders, or do not run it.

Engagement-based segmentation

"Highly engaged" segments built on opens will happily include people who have not read an email in a year. Sending your most aggressive offers to that segment costs you both the offer and the reputation.

What to use instead

The replacement signals are less flattering and considerably more honest.

SignalWhat it tells you
Hard bounce rateYour list quality and hygiene. Rising hard bounces mean stale or poorly sourced addresses, and mailbox providers treat repeated sends to dead addresses as a bad sign.
Complaint rateThe clearest negative signal there is. Someone pressed "report spam". Thresholds here are tiny, and small movements matter enormously.
Unique clicksA deliberate action. Not perfect, because security scanners follow links, but far closer to a human than a pre-fetched pixel.
Unsubscribe rateAn honest read on whether your frequency and relevance are working, and much healthier for you than a complaint.
Orders and revenueReconciled to your store's own order data, this is the only metric that survives contact with the finance team.

Clicks have their own machine problem: corporate link scanners and security appliances fetch every URL in a message. Count unique clicks per recipient rather than total clicks, be suspicious of a contact who "clicked" every link within a second of delivery, and never treat a single click as proof of a human. The point is not that clicks are clean. It is that they are much harder for infrastructure to manufacture at scale than an image request.

How Reachable handles it

Reachable's automated warmup gates on the signals that mailbox providers actually act on. Before a sending domain is allowed to step up to its next daily volume, the platform checks hard bounce rate and complaint rate against fixed thresholds, and confirms that SPF, DKIM and DMARC are still aligned. If a threshold is breached, the ramp holds where it is rather than climbing, and the team is notified.

Open rate is not one of those gates and never will be. We still record opens, because they are occasionally useful as a directional trend across a whole campaign, and because removing a number people have looked at for fifteen years without explanation helps nobody. They are simply labelled for what they are, and they do not control anything.

What to do this month

  1. Find every automation, segment or suppression rule in your platform that references opens. There will be more than you expect.
  2. Rebuild sunsetting on clicks, orders and time since last purchase, with a bounce and complaint check underneath.
  3. Move subject-line testing onto clicks or revenue.
  4. Put hard bounce rate and complaint rate on the dashboard your team actually looks at, next to revenue.
  5. Keep reporting open rate if you like the trend line, but stop letting it decide anything.

None of this makes your reporting look better. It makes it true, which is the only version worth acting on.

Want a second opinion on your deliverability?

We will look at your bounce, complaint and click data with you, and tell you what we would change.

Book a walkthrough
← Back to blog