In short
  • Treat opens as a rough signal, not proof of reading. Apple Mail Privacy Protection loads them whether or not anyone reads.
  • Filter automated clicks before you report to sponsors or judge an issue, and keep the raw numbers alongside.
  • Use first-party click tracking so you can see timing, links and patterns per recipient.
  • Write down what an engaged reader is, based on clicks and replies, and count readers, not just issues.
  • Track cohorts by signup month. Retention tells you more than any single issue's rate.
  • Give sponsors filtered numbers with a one-line method note, and stop reporting metrics you cannot defend.

Where fake engagement comes from

  • Privacy features that preload images. Apple says Mail Privacy Protection downloads remote content in the background by default, regardless of whether you engage with the email. The tracking pixel is remote content, so it fires when the message arrives.
  • Image proxies. Gmail serves images through Google's proxy servers and stops senders from learning the reader's location from image loads. Apple hides the reader's IP address too. Location and device data from opens is weak at best.
  • Corporate security scanners. Business email systems often open messages and visit every link to check for threats before a person sees them. Each visit looks like a click.
  • Link previews and bots. Messaging apps, browser extensions and crawlers can fetch links too, often after a reader forwards the issue.

None of this is a bug in your platform. The pixel and the redirect work exactly as designed. They just cannot tell a person from a machine on their own.

Illustrative example. Figures are fictional.

Check your own numbers in an afternoon

Before you change anything, find out how much of your current data is machine activity. You need a click export with timestamps, or access to your platform's raw click log.

  1. Split opens by mail client. If a large share comes from Apple Mail, your open rate mostly measures how many readers use Apple Mail.
  2. Plot clicks by minutes after send. People trickle in over hours. A spike in the first minute or two, before anyone could have read past the headline, is a scanner.
  3. Count links clicked per recipient. A reader who clicked every link in a twelve-link issue, including the unsubscribe and the privacy policy, did not read it.
  4. Look at your top clicked link. If the footer or the logo outperforms the lead story, something other than readers is clicking.
  5. Compare by recipient domain. Personal addresses and corporate domains behave differently. If one company's staff click at five times your average, check its timestamps.
  6. Find your platform's filter. Some platforms already filter for you. Beehiiv, for example, can show Verified Clicks alongside the unfiltered totals in post analytics. Know which number you have been quoting.

Write down what you find. That gap between raw and filtered is the number you will explain to sponsors later.

Define an engaged reader

Most teams never write this down, so the definition changes with whoever builds the report. Pick one, document it, and use it everywhere: sunset rules, sponsor reports, the dashboard.

A good definition has three parts:

  • A signal. A human click, a reply, a survey answer, a referral or a purchase. Not an open on its own.
  • A window. Long enough to cover a few issues. A daily newsletter can use a shorter window than a monthly one.
  • A threshold. One signal in the window, or two, depending on how strict you want to be.

Here is a fictional example for a newsletter that sends three times a week: an engaged reader has at least one filtered click or reply in the last 30 days. An at-risk reader had one in the last 90 days but not the last 30. Everyone else is inactive and goes into the sunset flow.

Decide what happens to Apple Mail readers who never click. Some of them read every issue in the preview and never need a link. If you sunset on clicks alone, send a plain re-engagement email first and let a reply count.

Pick metrics you can defend

MetricWhat it tells youHow much to trust itUse it for
Open rateThat the pixel loadedLow for any single issueLong-run trends and segment comparisons
Raw click rateThat a link was requestedLow for business audiencesSpotting scanner problems
Filtered unique click rateShare of recipients with a human clickGood, if you know the rulesJudging issues and sponsor reports
Reply rateThat someone read and caredHighEditorial direction and engagement definitions
Engaged readersHow many people are actually thereHigh, with a written definitionThe headline number for the business
Cohort retentionWhether new readers stayHighJudging acquisition sources
Downstream conversionsWhat readers did after the clickHigh, with clean UTMsProving sponsor and product results

Issue-level and reader-level metrics

Issue-level metrics answer one question: how did Tuesday's issue do? They are noisy. A holiday, a scanner rollout at one big company, or a single viral link can swing them.

Reader-level metrics answer a better question: how many people are reading, and is that number growing? They change slowly, which is the point.

Use both, for different jobs. Issue metrics help the editor compare subject lines, story order and formats. Reader metrics belong in the monthly review and in conversations about the business.

One rule keeps them honest: never average issue rates to describe your audience. Ten issues at a 3% click rate might be the same 300 people every time, or 3,000 different people. Those are very different newsletters.

Clicks and replies, done properly

Clicks

Report unique human clickers divided by delivered emails. Total clicks reward one enthusiastic reader, or one scanner, clicking ten times.

Track clicks per story, not just per issue. The story that earned the clicks is the one your editor needs to see.

Replies

Replies are the hardest signal to fake and the easiest to ignore. They land in an inbox nobody counts.

  • Send from an address that can receive mail, and route replies somewhere you can count them.
  • Exclude auto-replies and out-of-office messages before counting.
  • Ask for replies on purpose: a one-line question at the end of an issue works better than a generic "hit reply".
  • Feed a reply into the reader's engagement record, so a person who replies but never clicks still counts as engaged.

Cohort retention

Group readers by the month they signed up. For each cohort, track what share is still engaged after one, three and six months. This is the clearest view you will get of whether growth is real.

Signup cohortSourceEngaged after 1 monthEngaged after 3 months
MarchOrganic and referrals68%55%
AprilPaid social lead ads41%22%
MayGiveaway partnership30%11%

Illustrative example. Figures are fictional.

Read a table like this one by source. In this fictional case, April and May added the most subscribers and the fewest readers. The list grew. The audience barely did.

Cohorts also catch slow problems that issue metrics hide. If each new cohort retains a little worse than the last, something upstream changed: a signup form, an ad audience, a welcome sequence. You can model what that means for acquisition cost per engaged reader with our newsletter calculator.

Set up tracking you control

1. Own your click tracking

Route newsletter links through a redirect on your own domain, so every click is logged with its timing, the link and the recipient. Platform dashboards rarely let you see the raw patterns you need to filter.

Many sending platforms give you this through webhooks. Resend, for example, sends an email.clicked event that includes the link, a timestamp, the IP address and the user agent, and supports a custom tracking subdomain. Open and click tracking are off by default there, so turn them on deliberately. See our Resend work for how that fits a newsletter stack.

2. Score clicks by behavior

Scanners behave differently from people. They click within seconds of delivery, often hit every link in the email at once, and arrive in bursts from the same network. People click later, choose a link or two, and spread out over hours.

Turn that into explicit rules, and give each click a reason code:

  • Clicked within a short window after delivery.
  • Clicked most or all links in the issue within the same few seconds.
  • Clicked a hidden link no person could see. Some teams add one deliberately as a trap.
  • Came from a data center or known security vendor network.
  • User agent is empty or identifies itself as a bot or preview service.

Store the raw click, the verdict and the reason. When a sponsor asks why a click was removed, you can answer in one line.

3. Tag every link with UTMs

Google's own guidance is to always set utm_source, utm_medium and utm_campaign. Values are case sensitive, so utm_source=Newsletter and utm_source=newsletter show up as two sources. Pick a lowercase convention and generate the tags in your template, never by hand.

A simple pattern: source is the newsletter name, medium is email, campaign is the issue date or number, and utm_content names the placement, such as the lead story or the primary sponsor slot.

4. Send key events server-side

Browser analytics misses people with blockers and strict privacy settings. For events that matter, such as a signup, a purchase or a sponsor lead, record them on your server too. Google Analytics offers the Measurement Protocol for this, which Google describes as a way to augment its tag, not replace it.

Make your event intake idempotent. Webhooks can be retried, and a large send arrives as a burst. Store each event once, keyed on its ID, and never let a slow database drop one.

What belongs on the dashboard

One screen. If it scrolls, nobody reads the bottom half.

  • Engaged readers, with the definition written underneath.
  • Filtered unique click rate for the last ten issues, with raw shown faintly beside it.
  • Replies per issue.
  • Cohort retention for the last six signup months.
  • Bounce and complaint rates, because engagement falls when deliverability does.
  • Share of clicks removed by the filter. If it jumps, a rule or a scanner changed.

Leave opens off the main view, or show them as a trend only.

Share numbers honestly with sponsors

It is the morning after a sponsored issue. The sponsor wants a report. Here is what to send:

  • Delivered emails and engaged readers.
  • Unique human clicks on their placement, and total filtered clicks.
  • A one-line method note: how you filter, and what share of raw clicks it removed.
  • Their UTM-tagged link, so they can match it to their own analytics.

Lower but honest numbers build more trust than inflated ones that never turn into results. Picture a fictional sponsor who sees 2,000 clicks in your report and 400 sessions in their own analytics. They stop buying. The one who sees 450 in both comes back.

If a sponsor asks for open rates, give them, with the context. Never quote a raw click rate for a business audience without saying it includes scanners.

What to stop reporting

  • Open rate as a headline. Keep it as a trend line.
  • Raw click rate. Report filtered, and keep raw for diagnosis.
  • Total subscribers as audience size. Report engaged readers.
  • Click-to-open rate. When opens are inflated, the denominator is wrong.
  • Location and device from opens. Image proxies and privacy relays make them unreliable.
  • Averages across issues as audience reach. Count unique people instead.

What most guides miss

Lessons from systems we have built and run, not the usual checklist.

The inflation is worst where sponsors pay most

B2B newsletters reach corporate inboxes, which are exactly where security scanners run. A business audience can show click rates several times higher than real reading, which means a sponsor sees great numbers and weak results.

Filters can be too aggressive

A filter that removes too much will hide real readers and, if it feeds list cleaning, delete them. Review what a new rule removes before trusting it, and keep a way to restore people.

Your event pipeline can drop data

When you collect events yourself, a large send arrives as a burst. Make sure your system stores every event under that load, or your filtered numbers will be wrong in ways that look plausible.

Platform migrations break your history

Every platform filters and counts a little differently. When you move, your rates can jump or drop overnight with no change in readers. Run both systems side by side for a few sends if you can, note the date on every chart, and never compare before and after without saying so.

A scanner can hide a real reader

A scanner click and a real click from the same person can land on the same link. Do not discard the recipient because their first click was a machine. Filter clicks, not people.

Opens still have a use

Opens are unreliable for judging a single issue, but trends over time and comparisons between segments are still useful.

How we do it

We build first-party click tracking, behavior-based filtering and reporting that separates real readers from automated activity as part of our growth systems. For a B2B trade newsletter that sends three times a week with one operator, that meant bot-click filtering on engagement data after a move from Beehiiv to Resend, feeding sponsor reports and list cleaning. Related: the deliverability checklist.

Not sure what your numbers are hiding? Start with a process audit.

Sources

  1. Apple: Mail Privacy Protection and privacy
  2. Google Workspace Admin Help: Set up an image URL proxy allowlist
  3. Apple iPhone User Guide: Use Mail Privacy Protection
  4. Beehiiv Help: Post analytics dashboard walkthrough
  5. Resend Docs: Webhook event types
  6. Resend Docs: Open and click tracking
  7. Google Analytics Help: Collect campaign data with custom URLs
  8. Google for Developers: Measurement Protocol for Google Analytics

Facts checked against these sources in October 2026.