Email Open Rate Accuracy: What You Can Still Measure

Most outbound dashboards still show an open rate at the top. For the majority of your recipients, that number is no longer a measurement of anything.

Your sequence tool says 48% of your last campaign was opened. Your reply rate was 1%. Somebody on your team concludes the subject lines are working and the copy is not, and rewrites the copy.

That conclusion rests on a number that Apple has publicly said it does not report. Email open rate accuracy broke somewhere between 2021 and now, and the awkward part is not that opens are inflated — it is that the replacement metric everybody moved to has no agreed definition either. Two of the largest published cold-email studies disagree about reply rates by roughly seven and a half times, and both are correct, because they are measuring different things under the same word.

Here is what is actually knowable, and what to steer on instead.

Apple's documentation is the clearest source on email open rate accuracy

Open tracking works by embedding a tiny image. When the recipient's mail client requests that image from your server, you record an open. Mail Privacy Protection breaks the link between that request and the human.

Apple's own support page for iPhone says it plainly: Mail Privacy Protection "hides your IP address so senders can't link it to your other online activity or determine your exact location. It also prevents senders from seeing if you've opened the email message they sent you."

That is not a vendor's interpretation. That is the company controlling the mail client telling you the signal is gone. The same page adds one detail most summaries skip: "Mail Privacy Protection does not extend to links or email attachments such as PDF or other files." Clicks still register. Opens do not.

The scale question is how many of your recipients this covers. Litmus measures email client share from over 1.5 billion opens in its own analytics; for September 2026 it puts Apple at 76.56% of opens, Gmail at 16.10% and Outlook at 3.48%. Litmus notes that MPP-affected opens "are not considered reliable opens," and is upfront that client detection itself depends on images loading, so the shares carry their own distortion.

Litmus's summary also states that MPP affects "roughly 55-60% of all email opens," attributed to unnamed studies. That figure circulates widely. It is worth treating as an estimate rather than a measurement, because the page does not say whose studies produced it.

The replacement metric has no fixed denominator

When opens died, everyone moved to reply rate. Reply rate sounds like a solid number. It is not, because nobody agrees on either half of the fraction.

Belkins published a 2026 study covering 7,530,489 cold emails sent January to December 2025, with 34,393 replies tracked, across its own client campaigns. Its average reply rate: 0.45%. Its definition is stated outright — unique replies divided by emails sent, excluding auto-replies and bounce notifications.

Instantly's 2026 benchmark report covers platform-wide data from 1 January to 18 December 2025, described as billions of interactions across thousands of workspaces. Its average reply rate: 3.43%. Its definition is also stated — all replies received, including follow-up responses, divided by total emails sent. Auto-replies are not excluded.

These are not rival estimates of one quantity. One counts distinct humans who answered; the other counts reply events, so an engaged prospect who sends three messages contributes three. One strips out-of-office noise; the other does not. Run 2,000 emails and the Belkins definition predicts about 9 unique human replies while the Instantly definition predicts about 69 counted replies. A founder who benchmarks their 0.6% against the 3.43% headline concludes their campaign is failing by a factor of six when it may be running ahead of the stricter measure.

Belkins is explicit about a second trap: its earlier studies used unique openers as the denominator, so its own figures across years are not directly comparable. The metric changed underneath the label.

Two more cautions on this material. Belkins reports data from campaigns it runs for clients, which makes it high-quality agency data rather than an industry sample. And Instantly publishes the benchmark while selling the platform the data comes from — as does almost every source in this category.

One concrete example of how fast this degrades second-hand: lemlist's widely-shared benchmark roundup lists a 27.7% average open rate attributed to Instantly 2026. Instantly's own report publishes no open rate figure at all. If a number matters to a decision, open the primary page.

The numbers somebody else computes for you

The reason open and reply rates drift is that you own both the numerator and the denominator, so the definition can move. The metrics that hold still are the ones computed by a party with no stake in flattering you — the mailbox provider.

Gmail's sender guidelines set a hard line: keep spam rates reported in Postmaster Tools below 0.3%, with the stated guideline being below 0.10% and never reaching 0.30%. Google says the rate is calculated daily. Senders above 0.3% become ineligible for mitigation and regain eligibility after staying below it for seven consecutive days. The bulk-sender requirements — including one-click unsubscribe for marketing and subscribed messages — apply above 5,000 messages a day to Gmail accounts and have been in force since 1 February 2024.

That gives you a short list of numbers with externally fixed meaning:

• Spam complaint rate in Postmaster Tools. Defined and calculated by Google, not by your tool. Target under 0.10%.
• Hard bounce rate. The receiving server's verdict, not an inference.
• Meetings booked and meetings actually held. The gap between those two is where most outbound reporting quietly rounds up. We went into this in the piece on AI SDR meeting quality.
• Replies under a definition you wrote down. Unique or total, auto-replies in or out, measured per lead or per send. Any of these is fine. Changing it mid-quarter is not.

Notice that open rate is absent and that this list is shorter than your dashboard. That is the honest state of email measurement, not a gap in your stack. If the arrival side of this is your problem rather than the measurement side, our deliverability piece covers authentication and sending limits, and the one on collapsing reply rates covers why the underlying response is falling.

There is a deliverability angle to tracking itself. Belkins reports that it disabled open tracking across every campaign in the study period, because the tracking pixel "was hurting deliverability throughout the industry in 2024." That is one agency's call rather than a universal rule, but it is worth sitting with: an organisation sending 7.5 million cold emails a year decided the measurement was costing more than it was worth and switched it off.

Where the channel changes the question

Email open tracking is a workaround — a pixel bolted onto a protocol that never carried delivery receipts. Some channels report status natively.

WhatsApp's Business Platform sends message status webhooks for outgoing messages, including sent, delivered and read, with delivered described in Meta's reference as a message "delivered to a WhatsApp user's device." That is a protocol-level status from the platform rather than an image request you hope fires.

The honest limits matter here, and we checked rather than assumed. Meta's webhook reference does not define what triggers the read status and says nothing about recipients who turn read receipts off — so read is best treated as a signal that arrives sometimes, not a guaranteed per-message fact. And status does not mean interest; a delivered message to someone who never opted in is a compliance problem with a tidy green checkmark. BOSRAI runs outreach across email, WhatsApp and LinkedIn for exactly this kind of mixed-signal reality, and the measurement advantage on WhatsApp is real but narrower than vendors imply. Our WhatsApp B2B outreach piece covers the channel properly.

What to do before your next campaign

Four things, none of which require new software:

1. Delete open rate from any report that drives a decision. Keep it as a rough deliverability smoke alarm if you like — a sudden drop to near zero still tells you something. Do not let it grade copy or subject lines.
2. Write your reply-rate definition into the report itself. One line at the bottom: unique replies, auto-replies excluded, divided by emails sent. Then never change it without relabelling the chart.
3. Connect Postmaster Tools before you need it. The spam rate is the one number a mailbox provider will hold you to, and it is calculated daily whether or not you are looking.
4. Count meetings held, not meetings booked. It is the only number in the chain a customer also recognises. Pipeline math is where that turns into a forecast.

Where BOSRAI sits on this

BOSRAI is an AI sales platform for small teams — ICP work, lead sourcing, outreach across email, WhatsApp and LinkedIn, follow-up, and a CRM — with a human approving messages before they go out. Pricing runs from a free tier at $0 through $79.99, $199, $499 and $999 a month, discounted annually, on the pricing page.

On the subject of this article, we are in the same position as every other vendor: we cannot show you a reliable open rate, because nobody can. We have no published case studies, benchmark figures or customer results to point at, and we are not going to manufacture any. What we do believe is that keeping a human in the approval loop is worth more than a dashboard full of metrics with moving denominators, because a person reading the actual replies is still the most reliable measurement instrument in outbound.

If your reply rate looks bad against a published benchmark, check the denominator before you rewrite anything.

Sources

• Apple Support — Use Mail Privacy Protection on iPhone — Apple's own wording that MPP hides the sender's view of opens, and that it does not extend to links or attachments.
• Litmus — Email Client Market Share — September 2026 data from over 1.5 billion opens: Apple 76.56%, Gmail 16.10%, Outlook 3.48%; the note that MPP opens are not reliable opens, and the unattributed 55-60% estimate.
• Belkins — B2B Cold Email Response Rates (2026 study) — 7,530,489 emails sent Jan–Dec 2025, 34,393 replies, 0.45% reply rate with its denominator stated, open tracking disabled over deliverability, and the earlier-studies denominator change.
• Instantly — Cold Email Benchmark Report 2026 — 3.43% average reply rate, 1 Jan–18 Dec 2025, defined as all replies including follow-ups over total emails sent; no open rate figure published.
• Gmail Help — Email sender guidelines — spam rate below 0.3% required and below 0.10% recommended, calculated daily in Postmaster Tools, bulk-sender and one-click unsubscribe requirements from 1 February 2024.
• Gmail Help — Email sender guidelines FAQ — daily spam-rate calculation, mitigation ineligibility above 0.3% and the seven-consecutive-day recovery condition.
• Meta for Developers — WhatsApp webhooks messages reference — the sent, delivered and read statuses for outgoing messages; no definition of what triggers read and no mention of recipients disabling read receipts.
• lemlist — Cold Email Benchmarks 2026 — the roundup that places these figures side by side, and the source of the 27.7% open rate attributed to Instantly that Instantly's own report does not contain.