Engagement-seconds: how long users actually spend in your email

The web learned this lesson a decade ago. Bounce rate alone is meaningless — you need time-on-page. Email is still catching up. A message that is “opened” for 0.3 seconds before being deleted is not engaged with. A message that is read for 45 seconds is a buying signal. Engagement-seconds is a single, cheap, mostly-honest quality metric that most teams do not track at all.

What engagement-seconds measures

The implementation trick: embed two (or more) tracking pixels in the message, spaced by DOM position. The first fires when the email is opened. Subsequent pixels fire as the client requests images further down the body. Sum the deltas and you have a rough time-on-email figure.

<!-- top of email -->
<img src="https://t.example.com/o?id=X&p=1" alt="" />

<!-- middle, ~40% down -->
<img src="https://t.example.com/o?id=X&p=2" alt="" />

<!-- bottom -->
<img src="https://t.example.com/o?id=X&p=3" alt="" />

Most clients lazy-load images as they enter the viewport. Apple MPP pre-fetches everything at once (rendering the metric invalid for MPP-enabled recipients, which you must exclude or segment separately). Gmail caches but serves on viewport. Outlook Consumer blocks by default.

Why the metric is useful despite the noise

Three reasons engagement-seconds is worth tracking even with the caveats:

1. Relative, not absolute. You are comparing campaigns to each other, not to a universal benchmark. MPP noise is roughly constant across campaigns.
2. Distribution shape matters. A campaign with a fat right tail (many readers > 30s) is qualitatively different from one with a spike at <2s.
3. It catches hollow wins. Clickbait subject lines inflate opens; they do not inflate read time. Engagement-seconds will out them.

Always segment by provider

Do not report a single engagement-seconds number. Report one per provider:

• Gmail Webmail — cleanest signal, decent accuracy
• Gmail Mobile — viewport scrolling is real, pixel delta reflects scroll
• Apple Mail (MPP) — segment separately or exclude; MPP pre-fetches all pixels at once
• Outlook Consumer — images blocked by default; only trust results from recipients who clicked “Show images”
• M365 / Outlook.com — Safe Links can interfere; budget 1–2s of noise

What the numbers look like

From our corpus of opt-in B2B newsletters in 2026:

• Median engagement-seconds (non-MPP): 7–12s
• P75: 18–25s
• P95: 45–90s
• Bot-cluster (excluded): 0–1s burst, all pixels in same 200ms window

Cold outbound skews lower (most messages get scanned, not read). Transactional skews much lower (receipt checked and closed). Long-form newsletter content skews higher.

Filtering out the bots

You will see three bot signatures in raw data:

1. Burst pattern. All pixels fire in <500ms from a datacenter IP. Almost certainly a security scanner.
2. Single-pixel-only. Only the first pixel fires, ever. Imaging disabled or early-exit scanner.
3. Re-open pattern. Same recipient opens the same message every 5 minutes for days. Usually an indexing bot.

Flag and drop these before computing medians. Failing to do so will make your engagement-seconds look artificially low.

How to use engagement-seconds in practice

1. A/B test subject lines. If variant B has higher opens but lower engagement-seconds, variant B is clickbait. Ship variant A.
2. Newsletter section optimisation. Pixel-per-section gives you per-section dwell time. Shrink sections that nobody reads.
3. List health check. A list whose engagement-seconds is drifting down month-over-month is a list that needs pruning or re-engagement.

Limitations to be honest about

• MPP recipients are ~35–55% of typical consumer lists. Segment them out; do not pretend the metric works for them.
• Outlook Consumer blocks images. Most of your data will come from Gmail + mobile.
• A recipient who opens, leaves the tab open for an hour, and never scrolls will report a huge engagement number. Cap at 5 minutes.

FAQ