Return Analytics

Overview

Return Analytics goes beyond the basic Returns page to provide systematic analysis of return patterns across your entire catalog. The AI classification layer normalizes return reasons from all marketplaces into a consistent taxonomy, while the anomaly detection system flags unusual return spikes before they become expensive problems.

Key Concepts

AI Classification — The return_classify worker analyzes the customer's free-text return reason and classifies it into one of the standard reason codes, even when the marketplace provides a vague or non-standard reason. For example, "doesn't work as expected" is reclassified as defective or not_as_described based on the full review text.

Anomaly — A return rate or pattern that deviates significantly from the baseline for that SKU, category, or marketplace. Anomalies are detected using a statistical z-score model with a configurable sensitivity threshold.

Baseline — The expected return rate for a SKU or category, calculated from 60+ days of historical data. New SKUs without sufficient history inherit the category baseline.

Root Cause — The underlying operational or product issue driving elevated returns. NexusCommerce categorizes root causes as: product quality, listing inaccuracy, packaging, shipping damage, fulfillment error, or external (marketplace promotion spike creating unqualified buyers).

Getting Started

Navigate to Return Analytics under Advanced in the left sidebar.

Return Analytics requires at least 60 days of return data to establish meaningful baselines. For new accounts, the anomaly detection system becomes active after this initial period.

To trigger an immediate classification run on all unclassified returns:

1. Click Run Classification
2. Select the scope (all returns or a date range)
3. Click Start

Features

Return Analytics Dashboard

Header metrics:

• Overall return rate (selected period) vs. baseline
• Return rate trend chart (90 days)
• Returns by reason code (donut chart)
• Return rate by marketplace (bar chart)

Anomaly summary:

• Count of active anomalies by severity (high / medium / low)
• Most impacted SKUs

Reason Code Distribution

A breakdown of all returns by normalized reason code:

The +5pp notation means 5 percentage points above the baseline for that reason code.

Anomaly Detection

The Anomalies tab shows all active anomalies detected by the system:

Anomaly types:

• Rate spike — Return rate is significantly above baseline (default: >2 standard deviations)
• Reason shift — The distribution of return reasons has changed significantly (e.g., defective returns suddenly spike for a product that rarely had defects)
• Velocity spike — The raw number of returns per day has increased sharply, independent of sales volume

Click an anomaly to open the investigation view.

Anomaly Investigation View

The investigation view provides context for understanding and resolving an anomaly:

Timeline:

• Return rate chart with anomaly start date marked
• Overlaid events: price changes, listing changes, ad spend changes, new reviews

Return Sample:

• A sample of returns from the anomaly window with customer free-text reasons
• AI summary of the common theme across sampled returns

Proposed Root Causes: The AI generates a ranked list of possible root causes with supporting evidence:

1. Product quality issue — "24 of 31 returns in the anomaly window mention 'broken' or 'stopped working after a few days'"
2. Packaging damage — "12 returns mention 'arrived damaged' or 'package was crushed'"
3. Listing inaccuracy — "8 returns mention 'not what the pictures show'"

Actions:

• Mark as Investigating — Assign the anomaly to a team member
• Add Note — Record investigation notes
• Mark Resolved — Close the anomaly with a resolution reason
• Create AI Studio Flow — Automatically create a monitoring flow to track this SKU's return rate going forward

Return Cohort Analysis

Return Analytics includes a cohort view that groups returns by the order's fulfillment method, shipping carrier, or promotional source:

• FBA vs. FBM return rate comparison (product issues appear equally in both; shipping damage is higher in FBM)
• Carrier comparison (identifies which carriers generate more damage-related returns)
• Promo vs. non-promo return rate (promotions often attract buyers who return more frequently)

Root Cause Report

The Root Cause Report is a structured monthly report that summarizes:

• Top 5 SKUs by return cost
• Primary root cause per SKU
• Recommended actions per root cause
• Estimated revenue impact of reducing each SKU's return rate to category baseline

Export as PDF for sharing with operations and product teams.

API Access

Fetch return analytics programmatically:

GET /api/returns/analytics?sku=SKU-001&start_date=2026-01-01&end_date=2026-03-01
Authorization: Bearer <token>
X-Tenant-ID: <tenant-id>

Response includes reason code breakdown, anomaly flags, and root cause summary for the requested SKU and date range.

Configuration