How Accurate Is Email Open Tracking, Really?

Every tracker shows you a number. Almost nobody tells you how wrong that number can be, in which direction, and why. We build one, so we have measured it. Here is what our own data says, including the part that is uncomfortable for us.

Table of contents

The honest answer, before the detail

Open tracking reliably answers one question: did a mail client somewhere load an image that was in your message. Everything you conclude beyond that is inference.

It errs in both directions at once, which is what makes it tricky. It counts opens that were not human, and it misses opens that were. The two do not cancel out, because they hit different recipients: proxies inflate some senders, image blocking deflates others.

This is not a reason to ignore it. It is a reason to read the pattern rather than the total, and we will come back to that.

What inflates the count

Image proxies. Gmail loads remote images through Google's servers. Some clients fetch them when the message arrives rather than when a person opens it.

Security scanners. Corporate gateways open messages and follow links before delivery, which is exactly what they are paid to do.

Your own rereads. Open a thread where your sent message is quoted and you have loaded your own pixel. On a long thread this is the biggest single source of noise, and we wrote about it in counting your own opens.

Forwards. A forwarded message carries the pixel, so a second reader adds to the first one's count without you knowing there are two people.

A number from our own data

Here is the measurement that changed how we built ours. Looking at first opens on emails sent to Gmail addresses over thirty days, grouped by how long after sending they arrived:

• 10 to 15 seconds: 10
• 15 to 30 seconds: 26
• 30 seconds to 2 minutes: 11
• 2 to 60 minutes: 29
• More than an hour: 24

Thirty six of the first hundred first opens land within thirty seconds of sending. That is not a hundred people who happened to be looking at their inbox at that exact second. That is the receiving infrastructure fetching the image on delivery.

If your tracker reports those as opens, your open rate is inflated by roughly a third, and the inflation is concentrated on the recipients whose provider behaves that way.

What hides a real open

Blocked images. Many clients do not load remote content until the reader clicks "display images". A careful reader can read your entire email and never register.

Apple Mail Privacy Protection. It preloads everything for everyone, which does not hide the open so much as drown it. See our article on it.

Plain text clients and previews. A message read in a preview pane that does not render HTML leaves no trace.

Caching. This one is underrated. Once a proxy has fetched your pixel, it may serve its cached copy to the reader forever. The human open never reaches you, and the email can stay marked as never opened for good.

The mistake we made, and what it taught us

We filtered out opens coming from the same network the email was sent from, reasoning that it was the sender rereading their own message.

It was a bad rule. It silently hid a recipient sitting in the same office as the sender, a colleague behind the same VPN, a household on the same connection, and everyone behind a carrier grade NAT. A customer reported his tracking as broken, and he was right: we were discarding his real opens without leaving a trace.

What we should have done from the start, and now do, is record everything with a reason attached and decide what to count at display time. A filter that deletes data is a filter you cannot audit, and a tracker you cannot audit is a tracker you should not trust.

How to read a tracker without fooling yourself

Ignore the total, read the times. One open thirty seconds after sending is probably a machine. One open at 8:40 the next morning is probably a person.

Treat a lack of opens as unknown, not as a no. It is the least reliable signal the system produces.

Watch for shape. Several opens clustered in one morning after days of silence means something is happening. That inference survives all the noise above.

Never quote the number to the recipient. Beyond being poor manners, you would be quoting a number you cannot defend.

Conclusion

Open tracking is a rough instrument that is genuinely useful when you know its error bars. Used as a precise measure of attention, it will mislead you, and it will mislead you differently for each recipient.

The right question is never "how many opens". It is "does the pattern suggest this is alive". That question a tracker can answer well.

Frequently asked questions

Is a click more reliable than an open?

Generally yes. There is no equivalent of an image proxy loading every link on delivery, though corporate scanners do follow links. See click tracking.

Can I get a true open rate?

No, and any tool promising one is selling you certainty it does not have. You can get a consistent number, which is enough to compare two subject lines with each other.

Does it work better for one to one email than for campaigns?

Yes, because you read individual patterns rather than an average, and patterns survive noise far better than averages do.

Why does my email show many opens from one person?

Rereads, several devices, or a forward. The article on why an email gets opened many times goes through the cases.

Should I use open rate as a KPI?

For one to one sales, no. Replies are the number that matters, and they are not inflated by anyone's infrastructure.

All articles