Turning returns into the right pair
A fit-related return can become a lost sale, an additional handling cost and a missed opportunity to understand the customer. Our team explored how agentic AI could help customers find the right pair while protecting the retailer’s margin.
Perfect Pair brings customer context, product reviews, fit signals, return policy and available stock into one decision flow. It compares exchanges, credit and refunds, explains the trade-offs and routes judgment calls to people. A complementary prevention workflow identifies recurring return causes and proposes product-information improvements or quality follow-ups.
Early offline prototype using simulated scenarios
Three returns, three different decisions. Follow the evidence from customer need to resolution.
Illustrative walkthrough · Select a scenario and enlarge its screenshot.
A recommendation must solve the customer's problem and remain eligible under the policy before retained value becomes useful. The workflow checks reviews and stock, identifies who caused the return, and compares the costs of eligible options.
Size 8 wide addresses the narrow fit. A suitable pair must be available and reserved.
Store credit with suitable alternative picks when eligibility checks pass.
Return the payment when the customer prefers a refund.
Expected retained value = continuing-sale or credit value + returned-item recovery − logistics, handling, incentives and repeat-return risk.
The recommendation keeps customer choice visible. Priya’s damage case excludes Smart Credit; Jordan’s conflicting evidence pauses settlement.
Associates compare the physical condition with the photo and record the grade.
An acceptable check releases the held resolution. Conflicting wear or insufficient evidence pauses settlement for review; managers handle suspected abuse, exceptions and disputed condition.
Separate where the customer returns the pair from where it can best recover value.
Compare condition, size-level demand, stock and transfer costs. Inspection confirms the route: store resale, transfer, outlet or quarantine for a defect claim.
Log store grades against photo assessments. The competition design uses repeated mismatches to adjust photo confidence and focus inspection.
Recurring fit evidence becomes a product-page draft for merchandiser approval; batch defects become quality tickets. Simulated follow-up supports testing the proposed changes.
Our team used real ALDO data provided for the competition to investigate return patterns and estimate the potential value of converting suitable refunds into exchanges. I contributed to the business-value analysis and solution discussions. The team combined this business case with an agentic return workflow and completed AWS integration for the competition.
Compare transaction, return and store-exchange records across channels. Identify fit-related refunds and multi-size purchasing patterns, then use comments and product evidence to add context to broad reason codes.
Identify customers who still want the product but need a better size, width or replacement. Check fit evidence, eligibility and stock while preserving refund choice and human review.
Apply alternative exchange-conversion assumptions to suitable fit returns. Estimate retained margin and refunds avoided, using store exchange patterns as a reference while allowing for differences online.
The conversion, prevention and resale scenarios capture different sources of value and should not simply be added together.
Track whether suitable returns become exchanges that resolve the customer’s issue.
Assess value after shipping, handling, incentives and item recovery.
Measure time to completion, including approval and inspection waits.
Evaluate whether the outcome meets customer needs. The case uses a feature-based estimate; operational assessment would require customer feedback.
Our team won the Product Fit Award and received a $750 prize.
The team delivered a final pitch, a business case informed by real ALDO data and an AWS-connected competition solution. Scenario analysis estimated the potential business value of converting suitable refunds into exchanges; these projections were not measured operating results.
Perfect Pair connects customer service with inventory, merchandising and financial decisions—helping customers find a suitable resolution while turning return feedback into opportunities for improvement.
About this walkthrough: screenshots show an early offline prototype using simulated scenarios. This illustrative walkthrough is separate from our team’s later AWS-connected competition version, whose environment has since been cleared.