Email and channel engagement metrics

What it does

Email insights is the per-message analytics view: delivery trend, delivery funnel, engagement trend, device and client breakdown, per-link performance and a click map over the email design. The aggregate channel analytics dashboard rolls the same interaction data up across messages and journeys, with published formulas for each KPI.

Key facts

Email insights surfaces delivery rate, blocks, delivery failures, open rate, click rate, click-to-open rate, spam reports and unsubscriptions, split between unique and total opens and clicks.
The delivery funnel runs sent → delivered → unique opens → unique clicks. It is a funnel of counts, not of distinct people across stages, so don't read it as a cohort.
Link insights (clicks, unique clicks, click rate per URL or link alias) is only supported for emails and journeys created after September 2024. Click map needs the October 2024 release. If a journey contains even one older email, the whole journey is treated as legacy and link tracking and the heatmap are disabled for every email in it.
The click map shows only the latest version of the email, not the version a given recipient received.
User agent data — email client, browser, OS, device type — can be partial or missing for journeys that were active before August 2024. Undetectable agents show as "Unknown", detected but uncategorised ones as "Other".
Opens are pixel-based, so they depend entirely on the client loading images. Apple Mail Privacy Protection and similar features break open detection outright, and most modern clients block images by default.
Bot protection filters non-human clicks on email, text, push and custom channels. It explicitly does not filter opens. It is not retroactive, and it can affect journey triggers and branching that use "email clicked".
Bot filtering works by routing clicks through an intermediate page where the check runs, then redirecting.
Tracked links are rewritten to a geo-specific dynamics.com/api/orgs/<hash>/r/<id> form. Links can be marked non-trackable per link in the editor. Since October 2024, links in messages sent more than six months earlier still redirect but no longer record tracking results.
Tracking consent is cached for 24 hours, so an interaction can be recorded non-anonymously up to a day after the customer opted out. Opted-out clicks record as anonymous interactions with no profile reference.
Channel analytics KPI formulas are documented: open rate uses unique opens ÷ delivered (email) or ÷ sent (push); click-through rate uses unique messages with at least one click over delivered/sent; click-to-open uses unique clicks ÷ unique opens. Send rate is sent ÷ inflow; delivery rate is delivered ÷ sent.
Push notifications are all treated as delivered and are excluded from delivery rate; they report no bounce rate. Text messages record no opens at all, so open rate and CTOR are N/A for SMS.
Bounces are split into hard and soft for email, each measured against total sent; text messages report single bounces.

When to use / skip

Email insights is where you take a client when they ask why a send underperformed. The device and client breakdown is genuinely useful for design arguments — it settles the "does anyone still open this in Outlook desktop" question with data rather than opinion. Link insights and the click map are the two things marketers actually enjoy, so demo them, provided the tenant is new enough to have them.

Aggregate channel analytics is the right surface for anyone comparing across sends. It is also the only place with published formulas, which matters when a client's agency reports a different open rate from their own ESP. Nine times out of ten the disagreement is a denominator: Microsoft divides open rate by delivered, several ESPs divide by sent.

Where I'd skip: treating open rate as a KPI at all. On a modern audience it is measuring image loading policy, not human attention. If the client's dashboard has an open rate target on it, that is a conversation to have during design, not after the first send misses it.

Configuration decisions

Whether bot protection is switched on, and when. Turning it on mid-programme creates a visible step change in click rates that you will have to explain.
Which links get marked non-trackable — typically legal, privacy and preference centre links, where a tracked click is noise or a consent problem.
Whether the reporting pack uses open-based metrics at all, and if not, what replaces them. Click-to-open and click rate are the usual substitutes.
Whether older emails get rebuilt so that link insights and the click map light up, given one legacy email poisons the whole journey.
Which denominator convention goes into the client's own reporting layer, if they are rebuilding these KPIs in Power BI rather than reading the dashboards.
How SMS is reported, since it has no opens and its click trend can legitimately exceed anything open-shaped.

Gotchas

One pre-September-2024 email in a journey disables link insights and the heatmap for every email in that journey. People rebuild the wrong email trying to fix this.
Bot protection filters clicks but not opens, so after enabling it your click rate drops and your open rate doesn't. That asymmetry reads as a data error to anyone who wasn't told.
Because bot protection also affects "email clicked" triggers, a journey branch that was working on inflated bot clicks will suddenly see fewer people take the clicked path.
The 24-hour consent cache means a just-unsubscribed contact can still generate an identified interaction. It looks like the opt-out failed. It didn't.
Links in emails sent more than six months ago stop producing tracking data while continuing to work. Long-lived evergreen sends quietly stop reporting clicks.
The click map reflects the current email design. Edit an email after sending and the heatmap overlays new content onto old click data.
Text message trend charts can show clicks above opens, because SMS opens are not recorded at all. That is not a rendering fault.

Consultant notes

Get the open-rate expectation reset in the first analytics workshop, with the Apple Mail point made explicitly. Doing it later sounds like an excuse.
Demo the click map on a real send with real clicks. It is the single most persuasive analytics screen in the product and costs nothing to show.
Before go-live, agree in writing which KPI definitions the client is signing up to, and put the formulas in the handover document. Denominator arguments with agencies are otherwise permanent.
Turn bot protection on before the first production send rather than after, so there is no discontinuity in the baseline.
Check whether the tenant's journeys pre-date the 2024 link tracking changes. On an upgraded estate they usually do, and the client will assume the missing heatmap is a licensing issue.

Worth another look if Microsoft extends bot filtering to opens, or changes the six-month link tracking cut-off.