Newsletter Metrics That Matter: Opens, Clicks and Replies

In September 2021 Apple shipped Mail Privacy Protection with iOS 15, which lets Apple Mail users hide whether and when they opened a message. With the feature on, Apple Mail preloads images and content, including the tracking pixel, whether or not the person opened the email (Constant Contact). Apple's proxy servers fetch the pixel when the mailbox updates, and the platform books it as an open (Omeda).

The scale is not subtle. Omeda compared 60 days of data before the rollout with later periods and found open rates running at nearly double the pre-MPP level for its clients (Omeda). Constant Contact estimated that roughly half of a typical list could eventually sit behind MPP, based on Apple's share of opens and the opt-in rate seen for iOS app tracking (Constant Contact).

Mailchimp states the effect plainly: for contacts with MPP enabled, emails present in Apple Mail are reported as opened regardless of activity, producing inflated and inaccurate open rates, and any automation or A/B test triggered by opens becomes unreliable (Mailchimp).

Then there is the quieter half of the problem: image blocking everywhere else. Readers on corporate Outlook or with images off simply never fetch the pixel, so they count as unopened even if they read every word. One distortion inflates (MPP), the other deflates (blocked images). The net number is a blend of both, and you cannot tell which dominates on any given week.

The practical implication: open rate is a directional signal about subject lines and sender recognition, not a score of readership. Letterhead's guidance is to keep using it as a quick check on subject line appeal and deliverability, but to shift the real scorekeeping to intentional actions such as clicks (Letterhead).

Clicks and replies both require a person to do something. Clicks survive MPP because the reader has to tap a link in the rendered email; Mailchimp notes that click activity and click rate are still reported for contacts who enable Apple MPP (Mailchimp).

Clicks are not a perfect read of attention. A reader who saves the email and types the URL later never clicks. A reader who clicks the same link three times may count once if the platform tracks unique clicks. But the error runs in both directions and is small compared with the open-rate distortion.

For benchmark context, use numbers with their source and treat them as a compass, not a target. Mailchimp's all-user averages are 35.63% open rate, 2.62% click rate and 0.22% unsubscribe rate (Mailchimp). Benchmark Email's own customer data shows average click rates clustering around 2% to 5% depending on industry (Benchmark Email). A publisher reaching 2.62% on clicks is at the Mailchimp all-user average, and that is a reasonable starting checkpoint.

Replies are the most honest metric you have and the least automated. A reply proves the email rendered, was read, and produced enough reaction to type. It is also the only metric that gives you qualitative data: what people noticed, what confused them, what they want next. Writers who ask directly for replies report that personal, specific stories pull far more responses than generic calls to action (Quora discussion on newsletter engagement).

One caveat on click rate: it depends on there being something worth clicking. An issue with no link has a 0% click rate by definition. Reply rate has no such structural problem, which is why it suits essay-style newsletters and click rate suits link-roundup formats. Pick the one that matches your format rather than forcing both.

Unsubscribe rate is a health metric, not a growth metric. A rising unsubscribe rate usually means one of three things: the content drifted from the promise, the frequency changed without notice, or new subscribers arrived from a source that attracted the wrong people.

The bigger risk is not unsubscribes but spam complaints. Yahoo's sender hub asks bulk senders to keep spam complaint rates below 0.3% and to implement a working list-unsubscribe header with one-click unsubscribe (Yahoo Sender Hub). Gmail's sender guidelines set the same 0.3% ceiling, require SPF and DKIM for all senders, and require SPF, DKIM and DMARC plus one-click unsubscribe for anyone sending 5,000 or more messages per day to Gmail accounts (Gmail Help).

That 0.3% is a delivery cliff, not a vanity metric. AWS's summary of the Yahoo and Gmail changes groups the requirements into stricter domain authentication, easy unsubscribe, and monitoring spam complaint rates under that threshold (AWS Messaging Blog). A newsletter with an unsubscribe rate four times the Mailchimp average is telling you the same thing complaints do, one week earlier.

So read the two together. Unsubscribes are the polite exit; complaints are the impolite one. When both climb in the same month, stop testing subject lines and audit the offer.

Here is the decision rule that makes this manageable: watch one behavioral metric weekly (clicks per thousand delivered or replies per thousand delivered) against a four-week rolling median, and check opens only as a trend, never as a grade.

Why a rolling median rather than last week's number? Newsletter sends vary. A guest issue, a holiday week, or a topic that lands flatly all move the raw number, and a single week's dip means nothing. A median over four consecutive issues smooths the noise while still reacting within a month.

Why per thousand delivered rather than a percentage? Multiply by 1,000 and the number becomes a countable thing: 26 clicks per thousand delivered. If your list is small, percentages round to useless precision (1.8% versus 2.0% is one person on a 500-person list), while per-thousand counts stay legible. It also puts clicks and replies on the same scale, so you can compare them directly.

A worked example. Say you send to 8,000 delivered. Clicks: 176 this week, and 190, 154, 168, 201 over the previous four issues. Per thousand delivered, the four-week figures are 23.75, 19.25, 21.0 and 25.1, so the median is about 22.4. This week's 176 clicks is 22.0 per thousand. That is inside normal variation and requires no action. Now suppose this week brought 96 clicks, which is 12.0 per thousand. That is roughly half the median, which is a signal worth investigating before the next send.

delivered = 8000
this_week_clicks = 96
history_clicks = [190, 154, 168, 201]

per_k = [c / delivered * 1000 for c in history_clicks]
per_k.sort()
median = (per_k[1] + per_k[2]) / 2          # 4 points: mean of the middle two

this_week = this_week_clicks / delivered * 1000
ratio = this_week / median

print(f"this week {this_week:.1f} per 1,000, median {median:.1f}, ratio {ratio:.2f}")
# alert if ratio < 0.6, meaning clicks fell more than about 40% below the median

A threshold of 0.6 is deliberately loose. On lists of a few thousand, genuine one-week swings of 30% happen. You want to catch the halves and thirds, not the rounding.

The weekly checklist:

That last row is the one people skip. Without a note on what changed, four weeks of data tells you a number moved but not what moved it.

A weekly routine only earns its keep if a reading triggers an action. Three rules cover most cases.

Rule one: a click collapse is a relevance problem, not a subject line problem. Subject lines affect opens. Once someone has the email open, a lack of clicks means the issue did not deliver what the subject promised, or the links were buried. Fix the body and the link placement before touching the subject.

Rule two: flat clicks with a rising open rate is an MPP artifact, not progress. If opens climb while clicks hold steady, suspect that more of your list moved onto Apple devices with MPP enabled rather than more people reading. Mailchimp's own guidance is to reset success measurement around clicks, bounces, unsubscribes and conversions, and to watch spam complaints through Postmaster tools (Mailchimp).

Rule three: a reply rate you can count beats an open rate you cannot. If a newsletter consistently pulls 5 replies per thousand delivered and a new format drops it to 2, the format is the suspect, whatever the open rate says.

Consistency matters more than cleverness in generating these numbers at all. Practitioners who write about newsletter craft emphasize a fixed content schedule and a recognizable structure, because readers and measurement both depend on a predictable cadence (WeWork on newsletter consistency).

If you want a second pair of eyes on the routine, an expert can take your last quarter of sends, rebuild the per-thousand series and hand back a weekly template with thresholds set to your list size (https://quantaintelligence.ai/agents).

Do not optimize until you have four weeks of the chosen metric recorded. The first month is calibration, and its job is to tell you what normal looks like for your list.

During calibration, note three things besides the metric itself. First, your send cadence, because a weekly and a fortnightly newsletter have different noise levels. Second, the share of clicks (or replies) that come from your most engaged segment, if your platform segments engagement. Third, the unsubscribe and complaint trend, to confirm you are nowhere near the 0.3% complaint ceiling (Yahoo Sender Hub).

After calibration, change one thing at a time and let each change stand for at least three issues. A new subject-line style, a new link position, a new send day: one per month. Anything faster and the four-week median becomes a blur of overlapping changes.

Keep a simple log alongside the numbers: date, delivered, chosen metric per thousand, open rate, unsubscribes, complaints, and the one-line note. Twelve months of that log is a more useful asset than any benchmark table, because it is the only benchmark built from your own readers.

For the technical foundation, make sure authentication and unsubscribe handling are correct before you blame content for engagement dips. Gmail requires SPF or DKIM from all senders and one-click unsubscribe plus DMARC from bulk senders (Gmail Help), and Yahoo lays out the same unsubscribe and spam-rate expectations for senders (Yahoo Sender Hub).