Reliable Back in Stock Detection Design (0 -> 1)

Why Detection Design Matters

In Back in Stock systems, delivery failures are visible, but detection failures are more dangerous.
When inventory fluctuates rapidly, weak detection logic causes duplicate sends and poor customer experience.

As you know, these issues are easy to miss until traffic spikes.

This article focuses on building stable detection with minimal false positives.
Think of it as a design baseline you can implement without guesswork.

Trigger Rule: Only 0 -> 1+

Use inventory recovery as the only trigger:

trigger when prevStock = 0 and currentStock > 0
ignore transitions where stock remains above zero
evaluate by variant, not product-level aggregate

This one rule removes most duplicate scenarios.
Simple rules tend to be easier to operate and explain.

Core Detection Data Model

Use at least:

stock_snapshots: variantId, stock, observedAt
detection_events: variantId, transitionType, eventId, processedAt

detection_events is key for idempotency.
Store event IDs (or deterministic hashes) and skip already-processed events.

That single safeguard prevents many "why did we send this twice?" incidents.

Event-Driven Plus Polling Backup

Event-driven path

process inventory webhooks for low-latency detection

Polling path

periodic reconciliation to recover from webhook loss
nightly consistency checks to fix drift

Use event-driven as primary and polling as safety net.

Handling Inventory Flapping

Some variants oscillate between 0 and 1 within short windows.
Add debounce before enqueueing notification jobs:

set readyAt = detectedAt + 5 minutes
re-check inventory at readyAt
enqueue only if stock is still positive

Cooldown Policy

Prevent over-notification by variant/channel cooldown:

key: variantId + channel
duration: for example 24 hours
block repeated sends within cooldown window unless manually overridden

Detection Observability

Track:

detection count (0 -> 1 events)
enqueue rate
debounce drop rate
cooldown suppression rate
idempotency hit rate

Without these metrics, troubleshooting "too many" or "too few" notifications is guesswork.
And in operations, guesswork gets expensive very quickly.

Summary

A robust Back in Stock system starts with robust detection.
Combining transition-based triggers, idempotency, debounce, and cooldown gives you stable downstream delivery and cleaner operations.