800 Returns a Week.
Handled in 6 Hours, Not 4 Days.
Returns, catalogue management, and customer queries — three high-volume, rule-driven processes that grow with your revenue. When headcount is the only way to scale them, margins compress under growth. We build the automation that handles volume without adding people, and does it with consistent decisions that your returns policy actually intended.
as share of total sales
(NRF 2024 data)
on our e-commerce project
returns processing project
Growth That Makes Operations Worse, Not Better
E-commerce volume is a lever that amplifies. When your operations are manual, more orders means more returns, more queries, and more inconsistencies — at scale.
At 15% return rates, a €10M revenue business processes 1,500+ returns for every €1M in sales. Four people to process that manually, assessing condition, applying policy, issuing refunds or exchanges — it's a department that grows linearly with revenue.
The policy says 30 days, full refund on unused items. Person A interprets that differently from Person B. Edge cases — worn once, missing tag, different size claim — are decided by whoever is handling the queue that day. Customers get different answers for the same situation.
67% of shoppers check a retailer's returns policy before making a purchase (NRF). A slow resolution — even when it's eventually correct — erodes trust. The customer who waited 4 days for a refund decision doesn't come back next month.
Manual review can't catch patterns across thousands of transactions. The customer who's filed 7 returns in 60 days doesn't stand out in a queue. Automated detection catches the pattern — and flags it before the refund goes out.
What We Actually Built
800 Returns a Week — Without 4 Full-Time Staff
An online fashion retailer processing 800+ returns per week with four full-time staff. Average resolution SLA was 4 business days. Policy was applied inconsistently, and the team had no visibility into fraud patterns. We built a returns intelligence system — image-based condition assessment, automated policy application, a fraud-detection layer, and an exception queue for genuine edge cases. One person now manages what four used to handle, with faster resolutions and consistent decisions.
3 FTEs reallocated (€38K × 3 = €114K) + estimated fraud/chargeback reduction (€41K). Payback: ~10 weeks.
See full case study library →Consistent Decisions, Faster Resolutions, No Extra Headcount
The goal isn't removing humans from hard decisions. It's removing humans from the 90% of decisions that aren't actually hard.
Customer-submitted photos assessed for condition: unused, worn, damaged, missing tags. GPT-4 Vision classifies the item against your policy criteria. The decision is logged with the image, the classification, and the policy rule applied.
Your returns policy expressed as configurable rules — not code that needs a developer to change. Refund, exchange, store credit, reject — each outcome mapped to policy conditions, applied consistently across every return.
Patterns flagged automatically: high return frequency, serial returners, claim inconsistencies, address-email mismatches. Flagged before the refund runs. Exceptions route to a human with full context attached.
Genuine edge cases — unusual damage, customer dispute, policy ambiguity — routed to a single reviewer with all context: images, history, policy match rationale. One person handles everything the system couldn't resolve with confidence.
What We Build On
How many returns are you processing a week?
Share your returns volume, your current team, and your main SLA challenge. We'll map what's automatable, what needs human review, and what payback looks like for your numbers.
Get a scoped estimate