Review Intelligence

Overview

Review Intelligence is a closed-loop optimization system that transforms customer review data into actionable listing improvements. It analyzes patterns in negative reviews to identify specific listing weaknesses, generates improved copy using AI, submits the changes for review, and tracks whether the improvements reduce negative feedback over time.

This creates a self-improving product catalog: listings get better with every review cycle, negative review rates decrease, and conversion rates improve.

Key Concepts

Listing Weakness — A specific, identifiable problem with a product listing that is driving negative reviews. Examples:

• Title claims the product is waterproof but reviews say it is not
• Bullet points omit the dimensions, causing "not as expected" returns
• Images show a different color than what is shipped

Improvement Proposal — An AI-generated listing change proposal. Each proposal includes:

• The original text
• The proposed replacement text
• The reviews that motivated the change
• A confidence score
• The expected impact on negative review rate

Closed Loop — The cycle: reviews → analysis → proposal → listing update → new reviews → analysis again. NexusCommerce tracks the sentiment trajectory for each SKU after a listing update to measure whether the update helped.

Impact Score — Estimated reduction in negative review rate if the proposal is applied. Calculated by the AI worker using a causal model trained on historical listing-change → review-change data.

Getting Started

Navigate to Review Intelligence under Advanced in the left sidebar.

For Review Intelligence to generate proposals, you need:

• At least 10 reviews on the product (more reviews = better proposals)
• Sentiment analysis run at least once (or Auto-sentiment enabled in Settings)

Run Your First Analysis

1. Navigate to Review Intelligence
2. Click Analyze Catalog
3. Select scope: all products, a specific category, or specific SKUs
4. Click Start Analysis

The analysis job runs in the background. Results appear in the Proposals list within a few minutes.

Features

Proposals List

The Proposals list shows all pending listing improvement proposals:

Sort by Impact Score to prioritize the highest-value improvements.

Proposal Detail

Click a proposal to see the full detail:

Evidence Section:

• Representative negative reviews that motivated this proposal
• Extracted topics from those reviews (e.g., "waterproofing", "size accuracy")
• Sentiment score for the affected topic

Proposed Change:

• Side-by-side diff of current text vs. proposed text
• Explanation of why the change addresses the identified weakness

Impact Estimate:

• Projected negative review rate before and after change
• Confidence interval on the estimate
• Historical impact of similar changes on comparable products

Actions:

• Apply — Push the proposed change to the product listing and all marketplace listings
• Edit and Apply — Modify the proposed text before applying
• Dismiss — Reject the proposal with an optional reason
• Send for Approval — Route the proposal to a specific team member for approval

Auto-Apply Mode

Enable Auto-Apply in Settings > Review Intelligence to have high-confidence proposals applied automatically:

• Proposals with Impact Score > 70 and confidence > 0.85 are applied without human review
• Lower-confidence proposals are queued for manual review
• All auto-applied changes appear in the Monitoring tab

Impact Tracking

After a proposal is applied, Review Intelligence tracks the SKU's review sentiment for the following 60 days to measure impact:

• Negative review rate: before vs. after
• Topic sentiment: before vs. after for the affected topic
• Overall rating: before vs. after

If the change had a negative impact (rare, but possible), NexusCommerce flags it and can revert the listing to the previous version.

Multi-Language Support

Review Intelligence supports analyzing reviews in English, German, French, Spanish, Italian, Dutch, and Polish. Proposals are generated in the language of the listing's primary marketplace.

For a product sold on both Amazon UK (English) and Amazon DE (German), separate proposals are generated for each marketplace's listing.

Integration with AI Studio

Create a scheduled Review Intelligence flow in AI Studio:

Trigger: Weekly (Sunday 11 PM)
Step 1: review_sentiment (lookback: 7 days)
Step 2: listing_optimize (min_reviews: 5, min_impact: 60)
Step 3: notify_slack (message: "{{step2.proposal_count}} new listing proposals ready for review")

Configuration