Improving email click-through rate when opens no longer mean anything

Click-through rate is clicks divided by delivered messages — and the honest version of that calculation is harder than it was, because a meaningful share of recorded opens and clicks are machines. Apple Mail Privacy Protection pre-fetches remote images for Apple Mail users, inflating opens. Security gateways and some clients fetch the URLs in a message to check them, producing clicks that arrive within milliseconds of delivery from datacentre IP ranges. Before you try to improve the number, make it measure people.

Clean the measurement first

1. Use clicks over delivered, not clicks over opened. Click-to-open ratios inherit all the open-rate distortion, so they move when Apple changes a default rather than when your email improves.
2. Discard sub-second clicks. A click recorded before a human could have read the subject line is a scanner.
3. Discard datacentre-origin clicks where your tooling exposes the IP, and clicks on every link in a message from one recipient within the same second.
4. Discard the unsubscribe link from the numerator, and decide once whether tracking-pixel-only engagement counts as anything.
5. Compare like with like. Segment by provider; Apple-heavy and gateway-heavy segments distort differently.

After that filtering the number is usually lower than what your tool reported, and it is the one worth optimising.

Placement sets the ceiling

A message in Junk has a click rate near zero, and a message in Gmail's Promotions tab gets read on a different schedule to one in the primary inbox. If your click rate fell without a content change, check placement before rewriting anything. Measure it per provider from seed mailboxes you control, and keep a caveat in mind: the Promotions tab carries the INBOX label, so any label-derived number counts a Promotions landing as a clean inbox hit. Check the tab by eye on a seed.

The reputation side of placement is also worth a quick look, with one correction to older advice: Google Postmaster Tools retired its domain and IP reputation charts, so there is no grade to consult. What remains is Gmail spam rate — keep it under 0.3%, ideally under 0.1% — plus authentication and delivery-error panels.

The three changes that actually move clicks

Relevance, which is mostly segmentation. The largest click-rate gains come from sending less mail to fewer people. A message to the 10% of your list who did something related last month will outperform the same message to everybody, and it protects your reputation at the same time.

One action, stated plainly. Multiple competing calls to action split attention and add links, which also worsens filtering. Decide what the single next step is, say it in words that describe the destination, and put it above the fold in text rather than inside an image.

Honest, specific subject lines and preview text. A subject that over-promises buys an open and loses the click, and repeated over-promising earns complaints. The preview text is the second subject line; leaving it to be auto-filled with "View in browser" is wasted space.

Mechanics that quietly suppress clicks

• Tracking domain on a shared host. Shared link domains carry other senders' reputation and can be listed. Use a tracking subdomain on your own domain, and check link domains against Spamhaus DBL and SURBL.
• Link text that does not match the destination. It depresses trust and it is a filtering signal.
• Image-only emails. They fail for image-blocking recipients and they filter worse.
• No plain-text alternative. Multipart messages render better and look more like normal mail.
• Mobile layout. Most mail is read on a phone; a CTA that needs a pinch-zoom is not clicked.

What not to do

Do not optimise open rate. It is partly machine traffic and it is no longer a reliable proxy for anything. Do not raise clicks by generating them — pools of controlled mailboxes opening and clicking your mail is exactly the synthetic engagement Validity's Heatwave blocklist, launched 3 September 2026, was built to detect; it lists domains rather than IPs, launched with over a million domains listed, and publishes no thresholds, so no tool including ours can promise exemption. Do not use misleading subject lines to lift the top of the funnel; the complaints cost more than the clicks are worth. Do not add a second CTA because a stakeholder asked. Do not A/B test on a sample too small to distinguish from noise, and do not test two variables at once. And do not compare your click rate to a published industry benchmark — the segments, filtering and measurement are not comparable, and the number you would be chasing is someone else's unfiltered data.

Make the test loop real

Change one thing, send to a randomly split segment of adequate size, apply the same machine-traffic filtering to both arms, and look at clicks over delivered plus replies and downstream conversions. Replies are worth tracking alongside clicks: they are both a conversion signal and the strongest positive deliverability signal you can earn.

MailPilot covers the placement half of this: real per-provider inbox placement measured through a seed network, SPF, DKIM and DMARC checks per domain, and per-mailbox reputation recorded daily — so when clicks fall you can rule the inbox in or out before you rewrite the email.