E-commerce
June 28, 2026
"I want to return my order." The agent responds: "Which number?" Then: "Which product?" Then: "What is the reason?" Four messages later, the customer is annoyed and your FCR plummets. AI chatbot return prequalification solves this problem: collecting order, reason, eligibility, and intent before a human opens the ticket.
The NRF estimates the online return rate at 19.3% in 2025, with projections around 20.8% in 2026. Every poorly qualified return costs $10 to $65 to process (Alhena, returns and AI 2026). Lorikeet points out that returns and exchanges are the second most important intent to automate after WISMO (Lorikeet, e-commerce chatbot 2026).
This guide #138 focuses on customer service return prequalification, not on the return policy or the entire journey. Distinct from returns chatbot (#10) (exchange and refund guidance): here, the focus is on structured collection before handoff.
Summary
Why pre-qualify returns before human intervention?
The chatbot return prequalification structures the customer request upfront so that the first agent message is actionable, not interrogative.
Cost of CS back-and-forths
Messages per ticket: unqualified return = 4 to 7 exchanges vs 1 to 2 qualified
FCR: each "I'll get back to you" tanks the resolution rate
CSAT: the customer repeats their story, frustration rises
Agent cost: 8 min/ticket vs 3 min with a complete file
Errors: wrong order, wrong SKU, poorly applied policy
DTC mode example
Jeans brand, 680 return tickets/month, 5.2 median messages per ticket. Bot prequalifies: order lookup, reason (size/quality/defect), Loop eligibility, exchange vs refund preference. Gorgias handoff with structured ticket. Median messages 2.1, return FCR +16 points in 10 weeks.
How does it differ from the general returns chatbot (#10)?
The article returns chatbot (#10) covers the complete conversational journey: eligibility, exchange, refund, escalation. #138 zooms in on the upstream phase which reduces bounces.
Scope #138
Decision tree: ordered questions, no skipping
Handoff form: mandatory fields before escalation
Routing intent: exchange, refund, dispute, out of time
Proof collection: damage photos before agent
Back-and-forth KPIs: messages/ticket, 48h recontact
Completes the return reasons analysis (product diagnosis) and return policy (drafting), without duplicating them.
What information should the bot systematically collect?
A standard return prequalification form standardizes what the agent receives as soon as the ticket is opened.
Required fields
Order identity: email + Shopify order # (auto lookup)
SKU / line item concerned: if multi-product order
Return reason: closed list aligned with the Loop portal
Eligibility: within/past deadline, final sale, hygiene
Intent: size exchange, product exchange, refund, store credit
Product condition: new, worn, packaging, tags
Proof: photo upload if defective or wrong item
Useful optional fields
Desired exchange variant, customer's usual size, existing return tracking number, VIP/LTV for priority routing. Auto Gorgias tag: `return_prequalified`, `return_exchange_intent`, `return_litige`.
How should the prequalification decision tree be structured?
The bot return decision tree avoids dead ends and "come back tomorrow" loops.
Level 1: Type of request
Start return / return status in progress / refund pending / damaged product / wrong item / package not received.
Level 2: Auto-eligibility
Bot calculates days since delivery, final_sale product tags, excluded collection. Eligible → Loop portal or draft API. Not eligible → proposal for exception escalation with summary.
Level 3: Target resolution
Same SKU exchange: bot checks variant stock, portal link
Exchange other product: AI suggestions, handoff if cart > threshold
Refund: shipping fee conditions, announced refund delay
Dispute: photos + immediate escalation to senior queue
Anti-loop rule
Maximum 8 bot questions before action (portal link, escalation, or closure). Beyond that, handoff with full transcript rather than continuing to question.
How do I connect the bot to Loop, ReturnGO, or Shopify?
Prequalification becomes powerful when the bot reads real data, not a static policy.
Loop Returns
Self-service portal for standard cases. API Draft Returns for custom integrations: initialize → add items → finalize → set credit type → submit (Loop, Return Create API). The bot can create the draft return in-conversation before customer confirmation.
Native Shopify
Self-serve returns enabled on customer account. The bot sends the account link + policy reminder. Limited native exchanges: the bot needs to know when Loop is required.
Sync policy
Return window, exclusions, fees: same source as policy page, Gorgias macros and bot corpus. One single truth, three surfaces.
Deployment phase
Week 1-2: lookup + read-only eligibility. Week 3-4: auto portal link. Week 5+: draft API if volume > 200 returns/month. Astucia recommends read-only first, write access after 2-3 weeks of validation (Astucia, chatbot returns 2026).
How do you direct customers toward exchanges versus refunds starting from the pre-qualification stage?
Exchange pre-qualification protects margins: a customer guided early towards an exchange costs less than a full refund.
Priority exchange signals
Size reason: offer adjacent variant if in stock
Color/shade reason: suggest close catalog match
Recent order: hot customer, LTV preserved
Cart > €80: Loop credit bonus if configured
Exchange-first bot script
"Your size M jeans can be exchanged for an L (in stock). The exchange ship within 24 hours, refund within 10 days after warehouse receipt. Would you prefer to exchange or be refunded?"
Benchmark
Loop cites a 73% conversion rate of return requests into exchanges for optimized brands vs ~32% for the industry (D2C Times, Loop exchanges 2026). See store credit vs refund, size support (#128).
Which cases should be escalated immediately without lengthy pre-qualification?
Prequalification speeds up simple cases; certain intents require rapid human escalation.
Immediate escalation (max 2 bot questions)
Dangerous product / allergic reaction
Announced chargeback
VIP customer LTV > threshold
Package not received + tracking delivered
Suspected fraud: multi-returns same SKU
Very negative sentiment: threat of public review
Mandatory enriched handoff
Bot transcript + Shopify order + photos + intent tags + calculated eligibility. The agent does not ask any question that has already been answered. See chatbot limitations, VIP escalation, order fraud.
How do I write bot scripts by return reason?
Prequalification scripts by reason reduce unmatched cases and standardize the tone.
Reason: incorrect size
Bot: confirms SKU, offers 2 stock variants, links to PDP measurement guide, CTA for exchange portal. If out of stock: refund or restock alert.
Reason: defective product
Bot: requests 2 photos (product + packaging), confirms eligibility for free replacement, escalates if value > €150 or if repeat customer for defect on same SKU.
Reason: does not meet expectations
Bot: reformulates expectation vs PDP description, suggests catalog exchange, reminds about return fees if required by policy. Adds `return_expectation_gap` tag for merchandising.
Reason: pending refund
Bot: looks up Loop/Shopify status, announces warehouse receipt date + credit card processing time of 5-10 days. No escalation if info is available. See support templates, answers database (#102).
How do you measure the reduction in after-sales service back-and-forths?
Return pre-qualification KPIs prove ROI beyond the deflection rate.
Leading metrics
Median messages/return ticket: target < 2.5 post-bot
% pre-qualified tickets: complete fields at handoff
48h re-contact same intent: < 8%
Return FCR: target 75-88% (Bookbag)
Exchange vs refund rate: monthly trend
Lagging metrics
Return intent CSAT, agent/ticket time, return customer service cost, blocked abusive return rate. Astucia cites -40 to -60% return ticket volume within 30 days with a well-integrated bot.
Monthly dashboard
Bot vs human return volume, top 5 reasons, top 3 unmatched, correlations reason × SKU. See FCR (#136), ticket taxonomy (#135), response quality (#116).
How to train agents and align macros on prequalification?
Bot prequalification fails if agents ignore the handoff card and ask the same questions again.
Team Rule
Ticket tagged `return_prequalified`: the agent reads the card before responding. Prohibited: "What is your order number?" if already in the handoff. Allowed: 1-sentence confirmation + action (link, replacement, exception).
Aligned Macros
RET-PREQUAL-EXCHANGE: acknowledges receipt of the card + personalized portal link. RET-PREQUAL-REFUND: recall delay + RMA number if created. RET-PREQUAL-LITIGE: empathy + 2h senior SLA.
Feedback Loop
Agent flags "incomplete card" → bot review within 48h. Agent flags "bad bot eligibility" → fix corpus policy. Weekly review of 5 escalated tickets: could the bot have resolved them? See choosing questions to automate, DTC playbook.
How does Qstomy prequalify conversational feedback?
Qstomy automates the return prequalification with Shopify lookup, policy sync, and Gorgias structured handoff.
Features
Order lookup: email or # without reordering
Reason tree: aligned with Loop/portal reasons
Eligibility calculation: timeframe, exclusions, final sale
Exchange/refund routing: real-time stock
Handoff file: transcript + tags + photos
Unmatched return report: monthly corpus brief
Quantified DTC Scenario
Skincare brand, 520 return tickets/month, 6.1 median messages, return FCR 54%. Qstomy prequalifies 71% of return contacts (eligibility + reason + intent). After 12 weeks: median messages 2.3, return FCR 78%, exchange rate +19 points (Loop Shop Now), agent time/ticket -42%, return intent CSAT 4.6/5.
Explore AI support, Shopify, request a demo.
Which operational playbooks should be launched this week?
Playbook 1: audit of 30 unqualified return tickets
Export Gorgias return tag, last 30 tickets of 5+ messages. List repeated agent questions. Prioritize top 3 missing fields. Deadline: 2 hours.
Playbook 2: Notion decision tree
3-level diagram (request type → eligibility → resolution). Validate with support + ops. Implement in bot the following week.
Playbook 3: Gorgias handoff sheet
Internal note template: order, SKU, reason, eligibility, intent, photos, desired variant. Rule: tag return_prequalified if 6/6 fields completed.
Playbook 4: 14-day read-only phase
Bot lookup + eligibility + Loop portal link, zero write API. Measure messages/ticket and 48-hour recontact. Then activate draft API if KPI OK.
Playbook 5: agent handoff training
30 min: read sheet, do not re-ask questions, macros RET-PREQUAL-*. Quiz with 5 fictional scenarios. Target of 90% correct answers.
Useful links
Pre-qualifying is not about pushing the customer away: it is about saving them time and allowing your team to make a decision from the first human message.
Enzo
June 28, 2026
