Skip to content

To: marcus@reedco.com

“quick question” · Q3 Outbound · Step 1 of 4

Replied
  1. Sent

    Tue 9:02 — their timezone

  2. Delivered

    accepted by gmail.com

  3. Opened — verified human

    read for 64s · image re-fetched · not a proxy burst

  4. Replied

    “Sounds interesting — Tuesday at 10?”

2 machine opens filtered from this email — not shown, not counted

Honest analytics · included in every plan

Analytics that only count humans

Every open and click is verified the moment it happens — Apple Mail auto-fetches, Gmail proxies and security scanners get filtered, real people get counted. Smaller numbers. True ones.

Replies can't be faked · Every number clicks through to its emails

Every event, graded before it touches your dashboard

  • open · verified human
  • open · Apple Mail auto-fetch — filtered
  • reply · real person
  • click · security scanner — filtered
  • open · likely human (Gmail, read 64s)
  • click · verified human
  • open · image proxy — filtered
  • reply · out of office — classified

The anatomy of a fake open

Your email gets “opened” before anyone sees it.

The moment your email lands, machines start touching it: Apple pre-fetches every image, Gmail caches them through a proxy, corporate security opens the message and clicks every link looking for malware.

Norbelys watches how each event happens — network, timing, fetch pattern — and grades it instead of guessing. Only then does it count.

Open · 0.4s after delivery, Apple network

Machine

Click · linkscanner UA, every URL in the email

Machine

Open · Gmail proxy, image re-fetched at 64s

Likely human

Reply · written by a person, stops the sequence

Human

Graded verdicts — machine, possible, likely human — because honesty includes admitting uncertainty.

100%

of events verified before counting

3

verdict grades — not a coin flip

0

vanity metrics on the dashboard

1

click from any number to its emails

The four views

From one campaign
to one email, no dead ends

The funnel that doesn't flatter you

Sent → delivered → opened by humans → replied → meetings. Each stage clicks through to the actual emails behind the number.

Sent 4,812
Delivered 4,704
Human opens 1,017
Replies 443
Meetings 23

Every event, searchable

A live feed of everything that happened — filter by campaign, sender, contact or event type, with the verdict on every row.

  • Marcus Reed replied Human
  • Lena Fischer opened Likely human
  • old@gone.com bounced Suppressed
  • Priya Sharma clicked Scanner — filtered

A/B tests no bot can rig

Variants compete on real replies — the metric machines can't inflate. The winner takes over automatically.

A · “Quick question” 38 real replies
B · “Idea for you” 21 real replies

A wins on humans — now sending to everyone

Deliverability, watched daily

Bounces and failures per day, per sender — the early-warning view that feeds the automatic reputation protection.

Thursday's bounce spike slowed that mailbox automatically — you saw it here, your domain never felt it.

Lower numbers.
Better decisions.

Every plan ships with the full honest-analytics suite — verified events, funnels, traces and reply-first A/B. From $29/mo.

Questions, answered honestly

Why are Norbelys numbers lower than my old tool's?

Because roughly half of all email “opens” are machines — Apple Mail pre-fetches, Gmail proxies images, corporate scanners click every link. Most tools count all of it. Norbelys verifies each event as it happens and counts only the humans, so the number is smaller and true. Replies can't be faked, which is why everything important here is reply-first.

How does the classification actually work?

Each open and click carries signals: the fetching network, the user agent, the timing relative to delivery, whether an image was re-fetched long after the proxy cache. The classifier grades each event — machine, possible, likely human — instead of pretending certainty. You see the verified number first, with the filtered noise listed right next to it.

Can I see the story of a single email?

Yes — every email has a trace: sent, delivered, opened (with the verdict), clicked, replied, on one timeline, with the exact message that went out preserved. When a prospect says “I never got it”, you'll know.

Does the A/B testing use these honest numbers?

It's built on them. Variants are compared on real replies — the one metric no bot can inflate — and the winner takes over automatically. A subject line that only robots loved will never win your test.