Endpoint Reliability Baseline (Onboarding)

Endpoint Reliability Baseline (Onboarding)

1. 01

   Inventory every inbound and outbound event stream the client integration depends on

   List each source system, the event types it emits, and the downstream destinations that must receive them. A single consolidated endpoint such as EventSend can fan one payload out to Slack, Discord, Telegram, Pushover, or a custom webhook, so map which of those destinations are contractual and which are nice-to-have before you touch configuration.

2. 02

   Confirm the client's tolerance for duplicate deliveries in writing

   Ask whether a repeated event is harmless or billable. Deduplication via unique keys is a configuration choice, not a default guarantee, and the answer determines how much idempotency logic your delivery team must build around the endpoint.

3. 03

   Record the retry window and backoff behavior for each destination

   Document how many attempts occur, over what interval, and what happens after exhaustion. Silent drops usually surface here: a destination that fails quietly for six hours looks identical to a quiet client until someone checks.

4. 04

   Set a delivery-latency budget per event class and note it in the statement of work

   A lead-routing event and a nightly digest have different acceptable delays. Writing the number down converts an argument during an incident into a threshold you can monitor against.

5. 05

   Establish a dead-letter destination the agency controls

   Route failed deliveries to a channel your team owns rather than the client's. This keeps retainer work diagnosable without waiting on client-side access requests during an outage.

6. 06

   Run a synthetic event through the full path and timestamp each hop

   Fire one test payload from source to final destination and record arrival times. This baseline becomes the reference point when a client reports that something 'stopped working' three weeks later.

7. 07

   Log credentials, endpoint URLs, and rotation owners in the agency handover file

   Webhook secrets expire and staff rotate. A single page naming who holds each key prevents the integration from dying with a departed contractor.