什么是打开率？定义与指南

打开率是打开已发送邮件的收件人所占的百分比。

Optimization

Open rate is the percentage of delivered emails that register an open, usually calculated as unique opens divided by delivered emails, times 100. Most platforms record an open when a hidden tracking image loads, so it counts image requests, not proof that someone read the message.

That gap matters more than it used to. A person can read a message and leave no trace, while a privacy feature or a security scanner can load the image when nobody has looked at all. The errors run in both directions. Treat open rate as a record of what a tracking system could detect, and it becomes useful again, just for narrower questions than most dashboards imply.

Open rate is still the first number many teams check in everyday email marketing reports, so it's worth knowing exactly what it measures.

How Open Rate Is Calculated

The Formula: Unique Opens Over Delivered Emails

The common definition appears in vendor documentation such as Mailjet's email metric definitions and Infobip's email metrics reference guide:

Open rate = (unique opens ÷ delivered emails) × 100

Take a hypothetical example with invented numbers. If 2,000 emails are delivered and 500 different recipients register an open, the open rate is 500 ÷ 2,000 × 100, or 25%. The figures are illustrative only. They aren't a benchmark or a target.

The denominator deserves attention. Mailjet defines a delivered email as one that was sent and accepted by the recipient's server. Accepted isn't the same as placed in the inbox. A delivered message can still end up in a spam folder.

Unique Opens vs. Total Opens

Unique opens count each recipient once per message. Mailjet counts the first time a unique recipient opens the email, while total opens add up every open, which Infobip describes as all open events, including re-opens by the same recipient. Open rate normally uses unique opens, because one engaged reader can generate many total open events.

Why Two Dashboards Can Show Different Numbers

Platforms don't all count the same events. Mailchimp's open tracking help article says that when click tracking is enabled, a click on a link counts as an open, even if the tracking image never loaded. So two tools can report different open rates for the same send without either one being broken. Compare definitions before you compare percentages.

How an Open Gets Recorded

The Tracking Pixel

An open is usually recorded when an email client downloads a tiny, invisible image hosted by the sending platform. Nothing in that process checks whether a person looked at the screen.

Even the IETF standard for read receipts, a separate protocol from tracking pixels, says a message reported as displayed carries no guarantee that its content was read or understood.

Reads That Leave No Trace

The first error runs toward undercounting. Mailchimp states that open tracking won't work when recipients or their email client choose not to display images. Those people may read every word. Your report still shows nothing. Apple Mail adds a smaller gap: Twilio notes that repeat opens within Apple Mail are served from Apple's cache and aren't returned to the sender.

Opens No Person Made

The second error runs the other way. Some opens happen with no human attention at all.

Apple Mail Privacy Protection

Apple's Mail Privacy Protection page, dated December 12, 2025, says that with Protect Mail Activity, Mail downloads remote content in the background by default, whether or not the recipient engages with the email. A tracking image is remote content, so it can load before anyone opens the message. Twilio's guide to Apple MPP open events explains that Apple pre-fetches and caches messages, including their images, so a sender can't tell whether a given open came from the reader or from the prefetch.

Security Scanners and Other Nonhuman Activity

Mail security tools add another layer. The M3AAWG Senders Committee, part of an industry anti-abuse group, says in its post on nonhuman email interactions that these interactions typically happen during a mailbox provider's antivirus or security scans, which can produce recorded clicks and opens. Mailchimp says its system tracks clicks as opens even when a click comes from a bot or spam filter. One automated event can inflate two metrics at once.

Why Filtering Won't Give You the True Human Rate

Filtering helps, but it doesn't hand you a clean count of readers. Mailchimp's guidance on bot activity and bot filtering, published on Intuit's help center, says open and click metrics will drop significantly when bot filtering is turned on. Reported performance can change because a setting changed, not because readers did.

Apple traffic is the harder case. Twilio's open events include a field called sg_machine_open, which is set to true when Twilio believes an open was triggered by Apple MPP activity. Twilio also says a conventional open may still be associated with sg_machine_open set to true. So the flag can't separate human and machine opens cleanly. Drop every flagged open and you'll lose some real readers, but keep them all and you'll count prefetches as people. The filtered number is a different observation, not the true readership.

What Open Rate Can and Can't Tell You

Compare Within One Setup, Not Across Platforms

Open rate works best as a comparison inside one stable setup: the same platform, tracking and filtering settings, audience type, and reporting window. Across platforms, or across a filter change, the same movement can be measurement noise. Before reporting a rise or drop, record how the number was produced. A measurement note can record these choices as an editorial framework, not an industry standard.

Open Rate Isn't an Inbox Placement Signal

Google's email sender guidelines page is direct about this. Its open rate section says Google doesn't track open rates, can't verify the accuracy of open rates reported by third parties, and that low open rates aren't necessarily an accurate indicator of deliverability or spam classification issues. Where a message lands is an email deliverability question. It needs its own evidence, not an inference from opens.

Open Rate vs. Click Rate, CTOR, and Conversion Rate

Each related metric changes the numerator, the denominator, or both. Click rate compares unique clickers with delivered messages. Click-to-open rate (CTOR) compares unique clickers with unique openers. Conversion rate measures a chosen outcome against the audience or visits defined in the report. For a wider view of click metrics, the click-through rate (CTR) entry is the closest reference point.

Notice what happens to click-to-open rate when opens are inflated. Prefetched opens grow the denominator, so CTOR can fall even when the same people clicked. Bot clicks push the number the other way. CTOR doesn't escape the open problem; it inherits it from both sides.

Decisions Open Rate Shouldn't Make Alone

Opens can be missing or invented, so decisions about one person's open status need a second signal.

Removing subscribers for "never opening." A recipient with images turned off can read everything and still look inactive. Check clicks or replies first, and remember that scanners can click too.
Automatic resends to non-openers. A missing open is an unknown, not a confirmed miss.
Scoring buying intent from opens. A prefetched open looks just like an eager reader in the data. Treat it as a weak signal at best.
Judging subject lines on open rate alone. When some opens come from machines, a gap between two versions is harder to read. Pair it with a later action you care about.

When email supports a broader content marketing program, the useful question is usually what readers did next. When you review published results, including case studies in an agency portfolio, ask how any open rate was measured and whether bot or Apple traffic was filtered.

Privacy Rules in France and Italy

Regulators in France and Italy have published guidance on email tracking pixels. The summaries below describe those two countries only, as of October 2026. They don't describe rules in the United States or the rest of the European Union, and they aren't legal advice.

France. CNIL's recommendation on email tracking pixels, Deliberation No. 2026-042, was adopted on March 12, 2026, and announced in final form on April 14, 2026. It lists analyzing email open rates to measure and optimize performance as a purpose that requires the recipient's consent. It also allows an exemption for individual open tracking when that tracking is strictly necessary to adjust sending frequency or stop sending to inactive recipients, but only for emails the recipient requested or that relate to a service they requested, such as transactional emails. For that use, CNIL recommends keeping only the day of the most recent known open, without the time, and overwriting it with each new open.

Italy. The Garante's guidelines on tracking pixels in email, Provision No. 284 of April 17, 2026, were published in the Gazzetta Ufficiale, Serie Generale No. 98, on April 29, 2026. The Garante's press release places tracking pixels under Article 122 of the Italian Privacy Code, requires prior consent in ordinary cases, and allowed six months from that publication to comply, a period that ended in late October 2026. Recipients must be told about pixel use whatever the purpose of the email. The guidelines allow an exemption from consent for a statistical count of the overall open percentage, provided the data is anonymized, and they suggest using the same pixel for every recipient of a campaign and anonymizing related data such as IP address and client.

Frequently Asked Questions

How do you calculate open rate?

Divide unique opens by delivered emails and multiply by 100. Check how your platform defines both, since some count tracked clicks as opens.

What is a good open rate?

There isn't a universal target that holds across platforms, audiences, and tracking setups. The most useful comparison is your own history under the same settings, read alongside clicks and outcomes.

What's the difference between unique opens and total opens?

Unique opens count each recipient once for a message. Total opens count every recorded open event, including repeats by the same person.

How does Apple Mail Privacy Protection affect open rate?

With Protect Mail Activity on, Apple Mail downloads remote content in the background, so the tracking image can load without the recipient engaging. According to Twilio, repeat opens in Apple Mail are served from cache, so some later opens never reach the sender.

Can someone read an email without it counting as an open?

Yes. If images aren't displayed, the tracking image doesn't load and the open goes unrecorded, unless the platform infers an open from a tracked click.

Does Gmail use open rate to judge senders?

Google says it doesn't track open rates and can't verify open rates reported by third parties. It also says low open rates aren't necessarily a sign of deliverability or spam classification problems.

Why did my open rate change suddenly?

Check the measurement before you blame the message. A changed bot filtering setting, a tracking change, or a different audience mix can move the number without any change in how readers responded.

Related Terms

Conversion rate for email campaigns - the share of a stated base that completes a defined action.
A/B testing email variables - comparing two message versions against a result chosen in advance.
Click-to-Open Rate (CTOR) - unique clicks divided by unique opens.
Email Bounce Rate - the share of sent emails that the receiving server rejects.
Pixel Tracking - embedding a remote image to record when an email loads.