Open rate is measured with a tracking pixel: a tiny image embedded in the email that requests a URL when the message is rendered. The request is logged as an open.
That mechanism has two long-standing weaknesses and one recent, decisive one.
Why the number is unreliable
Images can be blocked. A recipient who reads the whole email with images off never registers an open. This has always caused undercounting.
Pre-fetching inflates it. Privacy features that pre-load remote images on the recipient’s behalf fire the pixel whether or not a human ever looked at the message. Apple’s Mail Privacy Protection is the largest example, and because Apple Mail holds a substantial share of consumer email, the effect on a typical consumer list is not marginal — it can dominate the metric entirely.
Plain-text has no pixel at all. Text-only messages are structurally uncountable.
The result is a number that mixes real human opens, machine pre-fetches, and silent undercounts, in proportions that vary by audience and shift as vendors change defaults.
What to use instead
- Click rate — requires a deliberate action, and is far harder to trigger accidentally.
- Click-to-open rate (CTOR) — useful for comparing creative, with the caveat that it inherits the open-rate denominator’s problems.
- Conversion and revenue — the outcome you actually wanted. Attribution back to the campaign or automation that caused it is the metric worth optimising.
- Delivery, bounce, and complaint rates — the leading indicators of reputation trouble, and none of them depend on a pixel.
Where open rate is still worth keeping
Two places. As a trend against itself — a sharp drop for one brand, on one domain, week over week, still signals something worth investigating even if the absolute level is meaningless. And as an engagement input for sunsetting, where combining opens with clicks over a long window is more useful than clicks alone.
What it should not do is decide whether a campaign succeeded, or serve as the headline number in a report to anyone making budget decisions.