Capture control
Any event can be stopped in both directions: dropped at ingestion, and the SDK stops sending it at all.
Dashboard → Events → Manage. Press Stop recording… on a row, or on an event's settings, and confirm. This loses data from that moment, so it asks first and says how much the event has been sending. Start recording turns it back on straight away.
Once an event is stopped, two things happen:
The pipeline drops it at ingestion.
Any copy still arriving — from old builds, from your server — is discarded before it's stored. The ingest response's dropped field says how many; the rest of the batch is kept.
Devices stop sending it.
The event name joins the disabled_events list in the SDK config. On the next config refresh — launch, foreground, or within about five minutes — installed apps stop emitting it entirely, saving the network traffic too.
The kill list is also readable directly at GET /api/v1/events/config with your x-api-key — useful if your backend batches events and wants to skip disabled ones at the source. Fictura's side caches the list for about 30 seconds, so stopping bites at ingestion almost immediately; devices pick it up with their next config refresh (up to five minutes).
You can also add an event from Manage: name it before the code that sends it ships, give it a description, and the row sits in waiting for data until the first occurrence lands — which is also a cheap way to catch a typo'd event name on day one.
Sampling
For a very chatty event, the Sampling field in its settings keeps it on only a share of devices: Keep on 10% of devices. The rate goes to the SDK in its config as sampling: { "<event_name>": 0.1 }, and the SDK decides per device and event, so a device is either always in or always out and its funnels stay whole. Kept events carry $sample_rate. Event counts on the Events tab and the total count measure in Insights are scaled back up from it, so a kept event at 10% counts as 10, and shows with ≈ in front of it. Breakdowns on an event's page show what arrived, because shares don't change. Unique users, property sums, funnels and retention are not scaled: they show the sampled devices only.
Empty, or 100%, keeps every event. Keep all clears the rate.