Docs

Tracking Engagement in watchOS 27's Conversational Siri

watchOS 27's on-wrist AI Siri has no screen navigation to track — instrument conversation turns and outcomes before this engagement goes invisible.

Engagement

watchOS 27, rolling out through September 2026, brings a redesigned Siri that can hold multi-turn conversations directly on the wrist — answering open-ended questions, brainstorming, and continuing a thread across several responses without the user ever opening an app screen. For a platform where every previous analytics approach has assumed a session means someone looking at a screen and tapping something, this is a genuinely new kind of engagement: a user can have a meaningful, extended interaction with an app’s underlying capability entirely through voice, and it will produce close to zero of the screen-view or tap events most watchOS instrumentation is built around.

That’s a real gap, not a theoretical one. If an app exposes any functionality through Siri intents or App Intents that the new conversational Siri can invoke — checking an order status, logging a workout detail, asking a follow-up about a previous answer — every one of those interactions currently has a good chance of landing nowhere in an analytics pipeline built for taps and screen transitions. A team could ship a feature that gets meaningfully used through voice on the wrist and have their dashboards report it as dead, simply because the instrumentation was never designed to catch an interaction with no visual step in it. That’s the opposite problem from most tracking gaps: it’s not that the data is wrong, it’s that an entire category of real usage doesn’t generate any data at all.

Data Points to Track

  • Conversation turn count per Siri session, capturing how many back-and-forth exchanges happen in a single interaction, since a single question-and-answer and a five-turn brainstorm represent very different depths of engagement
  • Intent recognised and fulfilled vs. intent recognised but unresolved, logged separately so a failed or misunderstood request doesn’t get miscounted as successful engagement
  • App Intents or Siri-invoked actions attributed back to the originating app, tagged with a distinct event type from any equivalent action taken through the app’s own UI
  • Follow-through actions after a conversation, such as a deep link opened or a task completed in-app shortly after a Siri exchange, to connect voice engagement to a measurable outcome
  • Abandonment point in multi-turn conversations, marking where a user stops responding or exits, as the closest voice-interface equivalent to a funnel drop-off

Setup Steps

  1. Register App Intents with distinct logging for any capability exposed to Siri, so an invocation through voice is never silently merged with the same action taken through the app’s own screen.
  2. Add a conversation-level identifier that ties together every turn in a single Siri exchange, since without it each turn looks like an isolated, unrelated event.
  3. Instrument fulfilment status explicitly — recognised-and-completed, recognised-but-failed, or not recognised — rather than only logging that an intent was triggered.
  4. Track any in-app action taken within a short window after a Siri conversation ends, to build an attribution link between voice engagement and concrete outcomes.
  5. Review this data on its own dashboard initially, rather than trying to blend it into existing session-based watchOS reporting, since the two measure fundamentally different kinds of interaction.

Actionable Insights

A high rate of recognised-but-unresolved intents is the first thing worth investigating — it usually means the App Intent surface doesn’t cover a phrasing or scenario users actually try, which is a fixable integration gap rather than a sign the feature isn’t wanted. A feature with strong conversation-turn depth but weak follow-through action is being used for information, not action, which is a legitimate outcome but a different one from what a screen-based version of the same feature would suggest. And if this data shows meaningful usage volume at all in the first weeks after watchOS 27 ships broadly, that alone is the more important finding — it confirms voice-only engagement is real and growing, and that a Watch companion app’s usage was very likely undercounted before this instrumentation existed.

Expert help

Need help tracking this in your app?

Our team sets up analytics pipelines for mobile and web teams every day. Talk to us and get your first events flowing in under an hour.

Talk to an expert