Lightng CRM

Guide

How accurate is email open tracking?

Not accurate enough to quote as a number, and wrong in both directions at once. Open tracking works by loading a 1×1 image. Apple Mail preloads that image whether or not anyone reads the message, security gateways fetch it while scanning, and clients that block images never fetch it at all. So an open rate is neither an upper nor a lower bound on how many people read your email. It is useful for comparing two similar sends. It is misleading as a headline figure, and it should not be compared to an industry benchmark.

How open tracking actually works

There is less machinery here than people assume, and understanding the mechanism is enough to predict every way it fails.

When a tool reports opens, it has embedded a 1×1 transparent image in the message (a tracking pixel) at a URL unique to that recipient and that send. When the recipient’s mail client displays the message, it requests images from the sender’s server. That request is the open. The server logs it, attributes it to the recipient encoded in the URL, and increments a counter.

That is the whole of it. Nothing is installed on the recipient’s device. Their mail client does not report anything back. There is no read receipt and no cooperation from the recipient’s software. An “open” is one event and one event only: something, somewhere, fetched an image.

Which means the measurement is only as good as the assumption that the thing fetching the image is a person reading the email. Four separate mechanisms break that assumption, and two of them break it in opposite directions.

The four things that break it

Two inflate the number, one suppresses it, and one quietly destroys the metadata that open tracking used to carry alongside it.

Read from each vendor’s own pricing page and documentation on 26 August 2026. Prices move; the date is here so you can tell how stale this is.

MechanismWhat it doesEffect on your numbers
Apple Mail Privacy ProtectionIntroduced in 2021. Routes mail through Apple's proxy and preloads remote content (including the pixel) whether or not the message is ever opened.Inflates opens. Every affected message counts as an open, and it also masks the recipient's IP, so location and device data derived from tracking became unreliable at the same time.
Gmail's image proxyGmail fetches and caches images through its own servers rather than letting the client request them from you directly.The IP you see is Google's, not the reader's. Caching also means a later re-read may produce no second fetch, so repeat opens undercount.
Mail security gatewaysCorporate filters fetch images, and follow links, to check them for malware before delivery.Inflates opens, and inflates clicks too. This is the reason click tracking cannot be called clean either.
Image blockingSome clients and configurations do not load remote images by default; some people read in plain text.Suppresses opens. These are real reads that are never recorded, in the same dataset that is simultaneously being inflated by the three above.

The errors do not cancel out into a usable estimate. Their relative size depends entirely on which mail clients your particular list uses, which is precisely the variable you do not know and cannot control.

Why benchmarks are worse than useless here

“What is a good open rate?” is the most commonly asked question about this metric and the one with the least defensible answer.

Most published benchmarks were measured before 2021, when an open still approximated a human being. Rates measured after are inflated by an amount that depends on the client mix of the list being measured. Comparing your figure to a benchmark therefore compares two differently broken measurements taken under different conditions, and the comparison will usually flatter you, because the newer number is the inflated one.

What survives is internal comparison. This month’s send against last month’s, to a similar audience, through similar clients, is a fair test: the distortions are roughly constant, so the change is real even though the level is not. That is what “directional” means, and it is a genuinely useful property. It is just not the same thing as a measurement.

What to measure instead

Replies, first. A reply is a person deciding to type something to you. No proxy, scanner or preloader generates one. It is the only signal in email marketing that cannot be manufactured by software acting on the recipient’s behalf, and it happens to be the outcome you actually wanted.

Clicks, second, with filtering. A click is usually a choice. The exception is the security gateway following links during scanning, which the better tools mitigate by discarding clicks arriving within seconds of delivery and by filtering known scanner user-agents. Neither is perfect. Clicks are strong evidence, not proof.

Opens, third, as a trend. Read them as movement between comparable sends, never as a headline and never against an external benchmark.

And split it by template. An aggregate open rate across everything you sent is a number you cannot act on. “Which of my six follow-ups earns replies” is a question whose answer changes what you write tomorrow. Measure the thing that changes a decision.

How to read a report honestly

The same three questions work on any tool’s reporting, including ours.

  • Is the open rate presented first and largest?If so, the report is designed to flatter rather than to inform. The least reliable number should not be the most prominent one.
  • Are replies counted at all?Plenty of tools track opens and clicks and stop there, because replies land in your mailbox rather than in their system. The most trustworthy metric is the one that is hardest for a vendor to collect.
  • Does the vendor tell you any of this?A tool that quotes open rates to one decimal place without a word about Apple Mail is either not paying attention or hoping you are not.

For what it is worth, this is the reasoning behind how Lightng reports email: clicks and replies where your eye lands first, opens labelled directional, and every figure broken down by template and sending address rather than pooled into one number.

FAQ

Open tracking: common questions

How accurate is email open tracking?
Not accurate enough to quote as a number, and it is not merely imprecise: it is wrong in both directions at once. Apple Mail Privacy Protection preloads tracking pixels whether or not anyone reads the message, security gateways fetch them during scanning, and both inflate the count. Meanwhile clients that block images produce opens that are never recorded at all. An open rate is therefore neither an upper nor a lower bound on how many people read your email. It is useful for comparing two sends to similar audiences, and misleading for almost anything else.
How do CRMs track email opens?
By embedding a 1×1 transparent image (a tracking pixel) with a URL unique to that recipient and that message. When the mail client loads images, it requests that URL from the sender's server, which logs the request as an open. Nothing is installed on the recipient's machine and nothing is reported by their mail client; the entire mechanism is an image being fetched, which is exactly why it breaks the moment something other than a human fetches it.
Does Apple Mail Privacy Protection break open tracking?
Yes, comprehensively, and it has since 2021. When enabled, Apple routes messages through its own proxy and preloads remote content (including the tracking pixel) regardless of whether the recipient ever opens the message. Every such message registers as an open. It also masks the recipient's IP address, which is why location and device data derived from open tracking became unreliable at the same time. Apple Mail is a large share of consumer email, so this is not an edge case.
Are click rates more reliable than open rates?
More reliable, yes; clean, no. A click generally requires a person to decide to go somewhere, which is a far stronger signal than an image loading. But corporate mail-security gateways follow links to check them for malware, and that registers as a click nobody made. The usual mitigations are to discard clicks that arrive within seconds of delivery and to filter known scanner user-agents, which helps but does not fully solve it. Treat clicks as good evidence, not as proof.
What is a good email open rate?
The honest answer is that the question no longer has a useful answer. Published benchmarks were mostly measured before Apple Mail Privacy Protection changed what an open means, and a rate inflated by automated fetches is not comparable to one that is not: the size of the distortion depends on which mail clients your particular list uses. Comparing your open rate to an industry benchmark compares two differently broken measurements. Comparing this month's send to last month's, to a similar audience, is defensible.
What should I measure instead?
Replies first: a reply is a person typing, which cannot be faked by an image loader. Then clicks, with scanner filtering applied. Then opens, as a directional trend across comparable sends rather than as a headline figure. And measure by template rather than in aggregate: 'which of my six follow-ups earns replies' is a question with an actionable answer, where 'what is my open rate' generally is not.