The Returns ML Lab for E-commerce
Predict Returns. Keep Profit.
One API call at checkout. ReturnML scores the return risk of every cart in under 50 ms and tells you the most profitable action — so you stop losing €8 on orders that come straight back.
Fashion e-com loses $850B/year to returns
Most brands fight it with two bad tools.
❌ Blanket free returns
Lose €8+ on every returnable order. You train customers to buy five, keep one.
❌ Charge everyone
Kill conversion. Good customers leave. Return rate barely budges.
✓ ReturnML
Score the cart before checkout. Charge, coupon, or do nothing — based on actual expected profit.
How it works
Three steps from zero data to live profit decisions.
- 01
Connect
Drop in a 1-line SDK or fire a single API call when a user hits the cart page. No warehouse, no ETL, no sprint required.
- 02
Score in <50 ms
ReturnML returns a return risk score and the incremental profit for each available action — charge delivery, offer a coupon, or do nothing.
- 03
Decide
You choose the action. ReturnML explains why with SHAP: "3 similar items in cart · 62% hist. return rate · COD payment."
Live demo
See it in action
Add items to the cart below. When you hit checkout, ReturnML scores the cart in real time and recommends the most profitable action.
Modanova
Shopify PlusNew Arrivals
Silk Wrap Dress — Red, M
DRESS-M-RED
High-Rise Jeans — W32
JEANS-32
Oversized Blazer — Black, L
BLAZER-L-BLK
Linen Shorts — Sand, S
SHORTS-S-SND
Ribbed Knit Top — Cream, XS
KNIT-XS-CRM
Your Cart (0 items)
Add items from the store to get started
One call. Full picture.
Risk score + expected profit per action, straight from the response body.
// POST /v1/score
{
"session_id": "usr_8f2a",
"cart": [
{ "sku": "DRESS-M-RED", "qty": 2 },
{ "sku": "JEANS-32", "qty": 1 }
]
}
// Response (47 ms)
{
"return_risk": 0.83,
"actions": [
{ "type": "charge_delivery", "delta_profit": 2.40 },
{ "type": "coupon_5pct", "delta_profit": 4.10 },
{ "type": "do_nothing", "delta_profit": 0.00 }
],
"explanation": [
"3 similar items in cart",
"62% historical return rate",
"COD payment method"
]
}Not just prediction. Uplift.
Most tools say "this order is risky". ReturnML says "if you charge delivery, you'll make €2.40 more than if you do nothing." That's causal ML, not correlation.
Day-1 accuracy
Cross-merchant embeddings mean the model is useful from your very first order — no cold-start waiting period.
<50 ms p99 latency
Built for checkout speed. ReturnML sits in your critical path and still returns before the page finishes loading.
Guardrails included
VIPs and first-time buyers are automatically protected. Never charge a top customer by accident.
Explainable by default
Every score comes with SHAP-based reasons your ops team can actually read — and your legal team can defend.
The outcome
3–6% fewer returns.
Zero conversion loss.
For a €100M fashion brand, that's €2–3M in net margin per year. You keep more sales and keep the right customers.
Works with Shopify Plus, Salesforce Commerce Cloud, and custom stacks.
Ready to stop leaving profit in the returns bin?
ReturnML is in early access. Request a spot and we'll set you up with a sandbox environment and a pilot plan.
Request early access →