Analytics reference
What the sender sees, precisely
Deal intelligence is only useful if the numbers are honest. This page defines every signal — including the traffic that deliberately never counts.
What counts as reading
Engaged time accrues only while the page is visible, the window is focused, and the reader has interacted within the last 30 seconds (scroll, cursor, keys, touch). An open tab in the background counts nothing. A single session's counted time is capped at 60 minutes, so a laptop left open doesn't fabricate a devoted reader.
"Opened 3× · 11m engaged" therefore means eleven minutes of an actual human actually reading, counted per stakeholder link.
The signals
| Signal | Meaning |
|---|---|
opens / sessions | Distinct visits on a stakeholder's link, bot-filtered. |
engaged time | Seconds of real reading per session and per section (see above). |
section dwell | Time per tagged section — "read pricing twice, 6 minutes" not "viewed page." |
returning visits | The same viewer coming back — often the strongest buying signal on the list. |
new_viewer | Someone who isn't the original recipient opened a stakeholder's link — a probable forward, scored with confidence and evidence (below). |
comments | Feedback buyers leave anchored to specific blocks of the page, with open/resolved state. |
automated | Traffic classified as scanners, link-preview bots, or datacenter fetches — reported in its own bucket, never mixed into the human numbers. |
New-viewer confidence, and the language contract
Forward detection is probabilistic by design — a phone and a laptop look like two viewers; VPNs move cities. So every new_viewer event carries a confidence score with its evidence:
| Confidence | Typical evidence | Read it as |
|---|---|---|
| ≥ 0.7 | New device, new city or country | Very likely a genuinely new person — the link traveled. |
| 0.4 – 0.7 | New device, same city | Probably new — could be a colleague on the same network. |
| < 0.4 | Same network, or same city + different device class | Plausibly the same human on another device. Don't over-read it. |
The contract your agent is taught: say "someone new opened Sarah's link" — never "Sarah forwarded it," and never guess a name for an unknown viewer. Dealpages reports facts and confidence; identity claims are not facts.
What never counts
- Email security scanners (SafeLinks, Barracuda, Mimecast and kin) that open every link you send — detected and bucketed as automated.
- Link-preview bots (Slack, Teams, iMessage unfurlers) — served a title card, never the page.
- Your own visits — the preview URL is untracked, and opening it once marks that browser as yours, so your later visits to any of your own stakeholder links ("did Sarah's link work?") count nothing either.
- Killed links — a link you revoked answers with an unavailable screen; nothing on it counts after that, and its history is kept.
- Idle time — visible-but-untouched tabs stop accruing after 30 seconds.
The automated bucket stays visible in reports rather than being silently deleted — so when your raw email tool says "5 clicks" and Dealpages says "1 human read it," you can see exactly where the other four went.
Section tagging
Section-level dwell comes from a simple convention: the publishing agent wraps major blocks in data-section attributes (recap, pricing, next-steps…). The templates do this automatically; pages without tags still get page-level engagement, just not the per-section story.
What buyers are told, and what they're never asked
Every tracked page carries a visible "engagement analytics" notice. Buyers never log in, never enter an email to view a page. IP addresses are truncated before storage (city-level geo only). Browsers sending the Global Privacy Control signal get no persistent identifier at all. Raw request logs are pruned after 90 days, fine-grained engagement events after 180. The FAQ answers the rest bluntly.