Here's the uncomfortable truth about email analytics: most of the numbers on your outreach dashboard are lying to you. That 60% open rate? A large slice of it is machines. That contact who "clicked three times"? There's a real chance a security scanner clicked those links and the human never saw the email. If you're deciding who to follow up with based on opens and clicks, you're following ghosts.

We run outbound at volume, and the thing that actually moves the needle isn't collecting more signals — it's ranking the signals you already have by how much each one is worth. Not all engagement is equal. A reply is worth a hundred opens. A verified human landing on your site is worth a hundred clicks. And an open, on its own, is worth almost nothing.

The signal hierarchy, from gold to noise

Every email interaction produces a signal, but they are wildly different in reliability. Here's how we rank them, strongest first.

Signal	What it really means	Trust
Reply	A human read it and typed words back. There is no faking this. It's the only signal that's unambiguous.	Gold
Verified landing	A real person, on a residential/mobile network, loaded your page after the email. Strong intent — if you filtered the bots.	High
Click	A link was requested. Could be the human. Could be a security scanner pre-fetching every URL in the message.	Noisy
Open	A tracking pixel loaded. Apple and corporate mail clients pre-load pixels automatically, so a huge share of "opens" are machines.	Near-noise

Notice the pattern: the signals are ranked by how hard they are to fake. A reply requires a human to compose text — impossible to spoof. An open requires nothing but a pixel request, which any mail server can trigger. Trust the signal in proportion to how much human effort it takes to produce.

Why opens are almost worthless now

Open tracking works by embedding an invisible 1×1 pixel in the email. When the image loads, you record an "open." That logic broke the day Apple shipped Mail Privacy Protection, which pre-loads that pixel for every message whether or not the human ever looks at it. Corporate mail security does the same thing at scale — it fetches every asset to scan for threats.

The result: a meaningful fraction of your "opens" happened in a data center, not on a phone. An open-rate number is still directionally useful for comparing two subject lines across thousands of sends, but for deciding whether this specific lead is warm, it's close to useless. We log opens. We do not let them flag a lead as hot.

Clicks are the trap

Clicks feel trustworthy — surely someone had to actually click? — and that's exactly what makes them dangerous. The single biggest source of fake engagement we've measured is corporate link-scanning. When you send to a business on Microsoft 365, its Safe Links service can request every URL in the email before the human ever sees it, to check the destination for malware. That registers as a click. On some sends, we found roughly 40% of "clicks" traced back to Microsoft's own infrastructure, not the recipient.

So a raw click count is not a lead-quality signal until it's been cleaned. Which brings us to the one technique that separates real engagement from noise.

The fix: classify every hit by who owns the IP

The reliable way to tell a human from a bot is not timing, not user-agent strings, not how fast they clicked. It's who owns the IP address. Every IP on the internet belongs to an autonomous system — an ASN — and that ASN tells you whether the request came from a residential ISP, a mobile carrier, or a data center.

A click or landing from Comcast, Verizon Wireless, or a regional ISP is a human. A click from Microsoft, Google Cloud, Amazon AWS, or a hosting provider is a machine. We maintain a lookup of IP ranges to their owners and tag every engagement event with a single is_bot flag based on the ASN that owns the source IP. One flag, computed once, and every downstream report filters on it.

This is the layer we're wiring into the outreach system right now. Reply capture is already live and it's the backbone of hot-lead detection today. The click-and-landing truth layer — capturing the source IP on every tracked event and running it through the ASN classifier before anything gets called "engagement" — is in progress. The design is settled; it's the piece that turns a noisy click count into a trustworthy one.

What’s already running: reply capture

The part that's fully live is the highest-value signal anyway. Every outbound send is logged against a permanent lead record. When a reply comes back, an inbound mail watcher catches it, matches it to the original send, and does five things automatically:

Step	What fires on a reply
1	Marks the lead as replied and flags it hot on the record.
2	Appends the reply text as a timestamped note on the lead, so the whole thread lives in one place.
3	Cancels any scheduled follow-up emails — nothing worse than a "just checking in" auto-send after someone already replied.
4	Fires an instant notification so a human can respond while the lead is still warm.
5	Forwards the message to the account owner so it's never sitting unseen in a shared inbox.

That's the whole point of email analytics for hot leads: the reply is the signal that matters most, so the system is built to never miss one and to act on it in seconds. Everything else — opens, clicks, landings — is supporting evidence that helps you rank the leads who haven't replied yet.

The scoring model that falls out of this

Once every signal is ranked by trust, hot-lead scoring becomes simple arithmetic. A reply pushes a lead straight to the top. A verified, bot-filtered landing on a high-intent page (pricing, contact) is the next tier. A bot-filtered click is a mild warm signal. A raw open is logged but scores nothing on its own. Stack those with the right weights and the lead list sorts itself — the people worth calling today float to the top, and the spam filters that "engaged" sink to the bottom where they belong.

Email analytics done right isn't about tracking more. It's about believing less — and believing the right things. The reply is gold. The verified human is high. Everything else earns your trust only after it's proven it came from a person.