How accurate is the prey detection?

With our current machine learning model we prevent more than 99.6% of prey delivery attempts. With our ever increasing number of installations, the models are becoming better generalised to new environments and different cat markings. In the event that there are any detection errors due to a new, unseen environment, these can be quickly reported helping the system learn.

A detection result does not equal transit

Prey detection and physical access are different parts of the behaviour to review. Transit Approved means no prey was identified and access was allowed by prey detection; it does not by itself prove that a cat physically passed through. The padlock icon shows the current padlock status.

Prey was flagged, but there was none

Open the relevant event → Detection Status and flag the misdiagnosis. Our team will review it and add it to training data. A routine false-prey report can use this path without an operator handover or a required device ID. See Reporting a false-prey event.

Prey was missed, or the cat entered unexpectedly

Contact Support with your device ID, links to the relevant events, what you expected and what actually happened. Include entry versus exit and the active policy (shown with a blue background), if known. An operator will review the evidence; do not treat this solely as a routine false-positive report or begin another generic reset sequence.

Open each event in the app → Share Event Link → paste the link(s) here in the existing support conversation. Use event links rather than event screenshots. If links are unavailable, tell Support; you do not need to recreate an event to obtain evidence.

When creating a training policy, you can configure a prey-detection rule, but detection in that mode may be less reliable. Enabling training mode on an existing policy leaves the lock permanently unlocked. See Active door policy and rule order.

Updated on: 06/10/2026