Open Rate Is a Diagnostic, Not a Success Metric

Opens are inflated by privacy changes. Clicks are inflated by bots. The signal that survives is what happens after the email. Here is how to measure it.

Most email dashboards still lead with open rate, and most people reading them already suspect the number is soft. It is worse than soft. As a measure of whether a campaign worked, open rate is now unreliable enough that leading with it misleads the people reading the report. This isn’t an argument to throw the metric away. It is an argument to move it to the role it can still do honestly, and to build campaign reporting on what the email caused.

Why opens broke

The open rate is measured by a tracking pixel that loads when the email is rendered. Apple Mail Privacy Protection, introduced in 2021, pre-loads images for Apple Mail users who turn it on, whether or not a human looks at the message, which registers as an open that didn’t happen. Proxy image loading and security scanners do the same elsewhere. The inflation can’t be cleanly corrected after the fact.

Kept in the right place, opens still tell you something: trends over time, subject-line health, early deliverability warnings. Keep it as a diagnostic. Don’t use it to judge return, because it no longer measures attention.

Why clicks only half-work

Clicks are better, but not clean. Corporate email gateways, spam filters, and anti-phishing scanners click links to check whether they are safe, so raw click counts are inflated too, worst in exactly the audiences that matter most for B2B senders: inboxes behind security tooling. Clicks are a step up from opens, not a destination.

Measure what happens after the email

The signal that survives is whether the email produced a real session and a meaningful action. The reporting model is a chain:

Email sent → unique click → website or app session → engaged session → meaningful action → conversion

Each step is stronger evidence than the one before it. A click with no session is probably a scanner. A click followed by an engaged session, an app event, a login, or a form submission is almost certainly a person who cared.

Two decisions follow. First, which actions count: a key page viewed, a login, a form, a booking, a purchase, an upgrade, a reactivation. Pick them per campaign type and write them down. Second, how long after the email you keep crediting it. Don’t take a window from a blog post; measure your own click-to-conversion timing (see the test below) and set the window where your conversions stop arriving. Higher-consideration products need a longer assisted window than a flash sale.

Report in layers

The fix for a misleading dashboard isn’t to hide the soft metrics. It is to rank them, so nobody mistakes a diagnostic for a result.

• Primary (success): conversions, revenue, pipeline, qualified leads, purchases, signups, reactivations.

• Secondary (how it happened): verified sessions, engaged sessions, unique click rate, post-click behavior.

• Diagnostic (health): open rate, raw clicks, bounce rate, unsubscribe rate, spam complaint rate.

Management reporting leads with the primary layer. Opens and raw clicks stay on the page, in the diagnostic layer, where they help you spot problems without being mistaken for outcomes.

Fig 1 · Follow the whole chain · From sent to conversion, reported in three layers

Telling bots from humans

Most platforms offer some bot filtering, and none of it is perfect; use it to reduce noise, not to prove a click was human. The usual bot signals: clicks within seconds of delivery, one recipient clicking every link, clicks with no site or app session, no JavaScript session, data-center IPs, identical click patterns from one corporate domain. The human signal is the opposite: a click followed by a session with several page views, a form, a login, or a purchase.

We know of no public, standardized benchmark for bot-filter accuracy, because every platform uses its own detection and scanners change constantly. The only benchmark that matters is your own, so run a short standing test:

• Bot filtering on versus off, to see how much it removes.

• Click-to-session match rate.

• Instant-click rate within 10 seconds, 30 seconds, and 2 minutes of delivery.

• Every-link-click rate per recipient.

• Click-to-conversion timing, bucketed by hours and days, which is where your attribution window comes from.

• A holdout group that received nothing.

The holdout is the one most teams skip and the one that matters most, because it is the only method that measures what happened because of the email rather than after it. That is incrementality, and it is the difference between a campaign that looks effective and one that is.

The confidence ladder

Fig 2 · The confidence ladder · From opens at the bottom to incremental conversion against a holdout at the top

All of this reduces to one ranking of how much to trust a signal:

1. Lowest: opens.

2. Medium: raw clicks.

3. Better: filtered unique clicks.

4. Strong: a click plus a verified website or app session.

5. Very strong: a click plus an engaged session or app event.

6. Best: incremental conversion against a holdout.

Report from the highest rung you can defend.

The bottom line

Keep opens and raw clicks in the dashboard. Don’t lead with them. For campaign performance, report verified sessions, meaningful post-click actions, conversions, and incremental lift. The metric that feels most familiar is no longer the one that tells the truth, and the fix isn’t a better open rate. It is measuring what the email caused.

Read next

• Incrementality Measurement (glossary). Measuring what happened because of marketing, not just after it.

• Attribution (glossary). Assigning credit, and its biases.

• Campaign Evaluation: Don’t Be Scared, Be Smart. The evaluation habit this piece narrows to one metric question.