Changelog

All notable changes to ReturnShield AI are documented here. Version numbers follow Semantic Versioning.

v1.1.0 — June 2025

Features

Manual outcome override period

Applied fixes can now have their outcome tracking period extended from 60 to 90 days. This is useful for products with low order velocity — on a product that ships 15 orders per month, a 60-day window produces only 30 data points, which may not be statistically conclusive. The 90-day option gives the comparison more signal before ReturnShield marks the outcome as final.

To extend: open any applied fix in the Applied tab, click the three-dot menu, and select "Extend tracking to 90 days". The extension can only be applied before the 60-day window closes.

Bulk reject

Previously, rejecting Action Queue items one by one was the only option. This release adds a multi-select checkbox to the queue and a "Reject selected" button. You can select up to 50 items at once and assign a shared rejection reason from a dropdown (Options: "Evidence insufficient", "Change not appropriate for this product", "Will handle manually", "Other"). Rejection reasons are logged and feed into future AI calibration — items rejected for "Evidence insufficient" are flagged for re-evaluation when more data accumulates.

Dashboard dark mode

ReturnShield's embedded dashboard now respects the Shopify admin's dark/light mode setting. Previously the dashboard rendered in light mode regardless of the admin preference. The transition between modes is animated and matches the Shopify admin's own transition behavior.

Segment threshold presets

Three threshold presets are now available under Settings → Customer Segments: "Conservative" (harder to reach At Risk or Serial Returner status), "Standard" (default values), and "Aggressive" (flags customers with lower return rates). One-click application replaces all four segment thresholds simultaneously.

When to use each preset: Conservative is appropriate after a category expansion where your new customers have naturally higher exploration-and-return behavior. Aggressive is appropriate for stores with wafer-thin margins where even moderate return rates are a significant cost driver.

Improvements

• Return rate trend chart now displays "Fix applied" markers at the dates when Action Queue fixes were applied. This makes before/after comparison visual — you can see whether a spike preceded the fix and whether the rate declined afterward.
• Action Queue confidence scores are now recalibrated monthly using actual outcome data from applied fixes. A fix type that consistently produces outcomes better than predicted will receive higher confidence scores; a type that underperforms will receive lower scores. This self-calibration improves prediction accuracy over time without any manual intervention.
• Customer segment export now includes a segment_assigned_date column alongside the existing segment name and risk score columns. This allows you to track how recently each customer was re-segmented.

Fixed

• Sizing widget incorrectly displayed "0% of customers kept" for products where exactly 10 customers had purchased and the 10th customer returned the item. The widget now correctly falls below the display threshold (80% keep rate) in this case rather than showing a misleading zero.
• Serial Returner alert fired multiple times for the same customer when multiple returns were processed in quick succession (e.g., a customer returning three items from one order that were processed individually). The alert is now debounced to fire once per 24-hour window per customer.
• Return rate on the trend chart was calculated incorrectly for the last calendar day of a month on stores with Pacific timezone settings. The off-by-one in timezone conversion caused a day's worth of returns to be attributed to the following month.

After Upgrading to v1.1.0

No configuration changes are required. The new segment threshold presets do not alter existing thresholds — they appear as options in the settings panel only. The "Fix applied" chart markers appear automatically for all previously applied fixes, retroactively.

v1.0.0 — March 2025

Initial release of ReturnShield AI.

Features

Return Intelligence Dashboard

90-day return rate trends compared against industry benchmarks by product category. True cost calculator that factors in refund amount, outbound shipping cost, return shipping cost, warehouse processing time, and restocking loss — showing the total cost of returns, which is typically 2–3x the refund value alone. Product risk scores from 0–100 ranked by return rate, return note pattern strength, and deviation from category benchmarks. Savings-to-date tracking for applied fixes.

AI Action Queue

Word-level diffs of proposed product description changes, each backed by an evidence summary drawn from customer return notes (e.g., "11 of 18 returns in the last 90 days mention 'tight shoulders'"). Each item shows a confidence score, expected monthly savings, and a data point count. Actions: approve and apply directly to Shopify, edit before applying, or reject with a reason. Optional auto-apply at a configurable confidence threshold for high-confidence items. 60-day outcome tracking per applied fix with predicted vs. actual return rate comparison.

Customer Risk Scoring

0–100 risk score per customer calculated from: return rate relative to store average, recency decay (recent returns weighted more), bracketing pattern detection, return reason quality (specific complaints weighted higher than "doesn't fit"), and order value vs. refund value ratio. Four segments: Champions (low return rate, high lifetime value), Standard (average behavior), At Risk (elevated return rate), Serial Returners (highest return rate, high fraud signal). Shopify metafield sync: returnshield.risk_tier written to each customer record for use in Klaviyo, Omnisend, and other email tools.

Shopify Flow Integration

Four triggers: High-Risk Order Placed, Return Spike Detected, Serial Returner Identified, Product Risk Threshold Exceeded. Two actions: Add Order Tag, Update Customer Metafield. Full payload schema documentation with ReturnShield-specific fields. Available on Growth plan and above; requires Shopify plan or higher for Flow access.

Sizing Social Proof Widget

"X% of customers kept size Y" widget rendered on product pages via Shopify metafields. Requires at least 10 purchases per size at 80% or higher keep rate before displaying. Customizable display text with variable interpolation. Automatically hidden for sizes with insufficient data.

Post-Purchase Emails

"How's it Fit?" email with configurable send delay from 3 to 14 days post-delivery, triggered by the carrier delivery date in Shopify. Post-return survey with conditional question routing to sizing, description quality, or product quality branches. Survey responses feed back into the AI analysis with 2x weighting compared to unstructured return notes.

Bracketing Detection

Flags customers ordering 3 or more variants of the same product in a single order. Cross-order bracketing pattern detection for customers who repeatedly bracket across multiple orders. Configurable threshold. Shopify Flow action for automated fulfillment holds.

Plan Availability

| Feature | Free | Starter ($29/mo) | Growth ($79/mo) | Scale ($199/mo) | |---|---|---|---|---| | Return Intelligence Dashboard | Up to 50 orders/mo | Unlimited | Unlimited | Unlimited | | AI Action Queue | — | Included | Included | Included | | Customer Risk Segments | — | Included | Included | Included | | Shopify Flow Integration | — | — | Included | Included | | Post-Purchase Emails | — | — | Included | Included | | Bracketing Detection | — | Included | Included | Included | | Sizing Widget | — | Included | Included | Included | | Data retention | 90 days | 12 months | 12 months | Unlimited | | Support | Community | Email (2 BD) | Email (1 BD) | Priority (same day) |

Requirements

• Shopify store on any paid plan (Basic or higher)
• Minimum 30 completed returns for meaningful AI analysis
• Shopify Flow available only on Shopify plan or higher (required for Flow integration features)

Next Steps

• Installation Guide — get started in under 15 minutes
• Dashboard Walkthrough — understand every metric on your dashboard
• AI Action Queue — start reducing returns with your first applied fix