Returns Abuse Report 2026 Get the report →

Deciding trust, in real time.

Every claim, delivery and refund scored the moment it arrives — so good customers clear instantly and the rest never reach your margin.

Book a demo Explore the platform

Live decisions in under 400ms. Ready for Shopify and Magento.

Behavioural alertUnusual return pattern detected
Claim received Refund requested
Cloud Runner Trainers Cloud Runner Trainers UK 6 · Qty 1 Order #ES-28471 $135.00Order total
Flora Allen Returning customer Repeat return pattern · 3 prior claims ReasonSize didn't fit
Signals analysed
Athena Hannibal Medusa Hamilcar
Prior claims3
Linked accounts4
Cross-retailer patternMatch found
Dispatch weightMismatch
Risk assessed
0/100 Risk score High-risk pattern detected Top 2% of claims this month Manual review required
95
LowReviewHigh risk
Action confirmed
Risk score 95 High risk
Investigation opened
Refund placed on hold
Case sent for manual review

Backed by

Initiative RéseauGoogleSTATION FbpifranceEDHEC EntrepreneursCAMPUS CYBERRégion Hauts-de-FranceInitiative RéseauGoogleSTATION FbpifranceEDHEC EntrepreneursCAMPUS CYBERRégion Hauts-de-France
Decisions Overturns Deliveries Chargebacks Network New rules Thresholds Weights Corrections Versions

The rules rewrite themselves.

Every decision feeds back. The model writes the next rule.

Every change is versioned, visible and reversible.

Book a demo
Learns from
Decisions
Overturns
Deliveries
Chargebacks
Network
Changes
New rules
Thresholds
Weights
Corrections
Versions
Alyssa, the E-Sentinel platform

Everything your fraud team already opens, in one place.

Behavioural graph, delivery evidence, session trail and network signals — all reading from the same score, so nobody has to reconcile four tools by hand.

Alyssa
Search…
7 9 AF
Dashboard
Filter Select Date
Claims Management
Open Cases 2 Closed Cases 5
Reviewed Claims+
202 50%▲ +32.40
Claims Refunded+
179 32%▼ -18.45
Fraudulent Claims+
23 17%▲ +20.34
Total Revenue Saved
FranceUK
Orders Status
Problematic
Refund 62%
Problematic
Return 74%
Errors 30%
Red Flag 20%
020406080
Last Problematic Transactions
Search by anything… Select created… Select status
Launch analysisUser infoAmountDate Shipping countryBilling countryStatusRisk score
Flora Allen9855228 $135.00 Sep 10, 20263:22 PM UKUK Return Fraud
95%8% ↘
 +
Belle Lee9855229 $400.00 Sep 9, 20263:22 PM USUS Refund
12%5% ↗
 +
Regino Reese9855231 $200.00 Sep 9, 20263:22 PM UKHong Kong Red Flag
100%7% ↗
 +
Mohamed Pham9855232 $460.00 Sep 10, 20263:22 PM UKUK Return
36%9% ↘
 +
Jonas Walls9855230 $200.00 Sep 9, 20263:22 PM FranceFrance Return Fraud
85%8% ↗
 +
Rows per page: 5 123

Customer Behavioural Analysis

Alyssa decodes risky patterns (identity misuse, payment disputes, returns abuse) and segments them, so harmful behaviour is blocked without slowing everyone else down.

Risk Score

Behavioural modelling produces a dynamic score for every order, enabling real-time risk assessment while keeping false positives down.

Operational Efficiency

By removing manual bottlenecks, Alyssa optimises decision workflows — so fraud and risk teams spend their time on the cases that actually need them.

Intelligent Sales & Marketing

Alyssa identifies high-value customers from behaviour, then recommends and automates targeted campaigns.

A score tells you what. The recommendation tells you why.

Alyssa doesn't hand your team a number and walk away. Every transaction comes back with the reasoning written out — what triggered it, where the risk originates, and what to do next.

Mr M

Return fraud detected on transaction 9855228. Risk score 95%. Evidence pack drafted.

Read 4 signalsMatched 2 accounts0.3s
Mr M — Flora Allen case1:24

General recommendation

The plain-English verdict an analyst can act on without opening anything else.

  • Actionable guidance for the case in front of you
  • Points out the specific risks that drove the score
  • Context-rich, not just a threshold

Signal-level recommendation

The layer underneath: every trigger, broken out and explained.

  • In-depth view of triggers and patterns
  • Precise description of the suspicious activity
  • Immediate preventive measures
  • Automatic escalation where it's warranted
See a sample recommendation report (PDF) →

Returns stopped being a cost. They became a channel.

Organised refunders, wardrobers and empty-box claims now move faster than the teams reviewing them.

$13.70

Lost to refund abuse for every $100 of merchandise returned

Source: NRF returns survey

Six problems. One behavioural layer.

Each module solves a specific failure point. They share one score, so a signal caught in delivery still counts at the next checkout.

Athena

Judge the claim, not the customer

Every refund request is classified against behavioural and delivery evidence, so only the genuinely odd ones reach an analyst.

  • Auto-clears low-risk claims so analysts stop touching them
  • Flags repeat abuse across accounts, addresses and devices
  • Protects good customers from blanket policy tightening
Damaged on arrival · first claimauto-refund
Wrong size · returned in windowauto-refund
Not received · 3rd in 21 daysreview
Explore Athena →
Hamilcar

Prove what actually arrived

Weight, scan timing and delivery confirmation checked against the claim — so a carrier dispute starts with evidence instead of a guess.

  • Detects dispatch mismatches in weight and scan timing
  • Cuts carrier disputes by settling the record up front
  • Backs the refund decision with something defensible
Dispatch weight0.4 kg
Expected SKU weight1.3 kg
Delivery scanmismatch
Explore Hamilcar →
Hannibal

Know who's behind the account

Email age, IP posture and digital footprint checked at signup and at checkout, with no extra steps for real shoppers.

  • Verifies real identities in milliseconds
  • Surfaces masked signals like throwaway domains, VPNs and geo mismatch
  • Keeps onboarding frictionless for trusted users
email age 4dVPNgeo mismatch 4 linked accountsno social footprint
Explore Hannibal →
Medusa

Fraud rarely hits one retailer

Encrypted risk signals shared across the network, so an abuser burned somewhere else doesn't arrive at your checkout unknown.

  • Cross-retailer patterns without exposing personal data
  • Anonymised exchange that is GDPR-aligned by design
  • Stronger scoring from external context
Retailers in network34
Signals exchanged / day1.2M
Personal data sharednone
Explore Medusa →
Atlas

See the pattern, not the case

Every score, override and outcome lands in one place — so the team can show what fraud cost this quarter, and what it didn't.

  • Breaks results down by market, channel and claim reason
  • Compares against the baseline you had before deployment
  • Exports for finance and risk without a manual pull
Reporting dimensionsmarket · channel · reason
Baseline comparisonpre-deployment
Scheduled exportsweekly
Explore Atlas →
Gaia

Turn return data into CO₂ savings

Every blocked fraudulent return becomes a measurable CO₂e figure — auditable, and ready for the annual report.

  • Calculates avoided emissions from returns never shipped
  • Tracks waste reduction alongside fraud outcomes
  • Published methodology so the number survives scrutiny
Returns blocked184,220
Shipping legs avoided368,440
CO₂e avoided2,847 t
Explore Gaia →

Every return you don't ship is carbon you don't burn.

Few fraud tools measure this. Gaia turns blocked returns into a figure your sustainability lead can defend in a disclosure.

See the methodology

Impact Report

0 t
across participating retailers
Fraudulent returns blocked184,220
Shipping legs avoided368,440
Units diverted from waste91,700
Calculated with DEFRA 2025 conversion factors.

Plugs into what you already run.

Alyssa connects to your storefront, your order management and your carriers — no replatforming, no data migration.

Example merchant, reading live from 3 connected systems
ShopifyOrdersDHLCarrier scansStripeChargebacksAdobe Commerce (Magento)OrdersRoyal MailCarrier scansWooCommerceOrdersPayPalPaymentsEvriCarrier scansSalesforceCRMNetSuiteERPShopifyOrdersDHLCarrier scansStripeChargebacksAdobe Commerce (Magento)OrdersRoyal MailCarrier scansWooCommerceOrdersPayPalPaymentsEvriCarrier scansSalesforceCRMNetSuiteERP
BigCommerceOrdersZendeskCS ticketsSAPERPKlarnaPaymentsDPDCarrier scansPrestaShopOrdersAdyenPaymentsUPSCarrier scansREST APIAnything elseBigCommerceOrdersZendeskCS ticketsSAPERPKlarnaPaymentsDPDCarrier scansPrestaShopOrdersAdyenPaymentsUPSCarrier scansREST APIAnything else
N Northgate Apparelnorthgate.co.uk
Orders & customersShopify · 42,180 orders / month Connectedsynced 40s ago
Delivery scans & weightsDHL · Royal Mail Connectedsynced 2m ago
Payments & chargebacksStripe Connectedsynced 15s ago
Contact us about another integration →

Proven in production.

Measured against the pre-deployment baseline.

3.1% Annual revenue recoveredAnnualised
72h Maximum claim resolution timeInstant for low-risk claims 6-month pilot, 3 merchants, 2026
55% Attempted fraud prevented

“Engineered for scalability — the algorithm implements ML models in a way a fraud team can actually read and act on.”

See it score your own claims.

Send a sample month of order history. We'll show you what Alyssa would have flagged — and what it would have cleared.

Book a demo Speak with an expert

The audit is free.