Step 1
Refund Check
Detects refunds already issued and existing recovery patterns before any rebuttal work begins.
How Victava works
From dispute to decision — you control what gets filed.
Five stages — gather, score, decide, approve, file — run as a durable, checkpointed process that never drops a dispute. Victava pulls the evidence Stripe can't reach, scores the odds, and tells you when a dispute isn't worth fighting. You see every factor and every source. Manual review is the default; hands-off filing happens only if you separately enable Auto-Pilot and define its limits.
The 5-stage pipeline
Gather → Score → Decide (EV) → Review → File. Manual review is the default; Auto-Pilot changes only the review decision after explicit merchant opt-in and within configured limits.
- 1
Gather
A new chargeback hits your Stripe account and Victava picks it up within seconds. It then pulls evidence from your whole commerce stack — the Stripe charge, card, and 3DS/AVS/CVC signals, your Shopify orders, fulfillment, and tracking, the customer's own words in your helpdesk, and signed carrier delivery proof. Stripe's own tool only sees what's in your Stripe account; we connect the rest.
Implementation: Sources: Stripe charge + Radar · Shopify orders/fulfillment/tracking · Zendesk/Gorgias/Gmail · FedEx/UPS/USPS — every field stamped with its API and timestamp
- 2
Score
A factor model computes your win probability and shows its work. Card-network rules — Visa CE 3.0, VFMP 10.5, 3DS liability shift — are deterministic and labeled as fact; everything else is a calibrated estimate, labeled as an estimate. Every factor shows its weight and its source.
Implementation: Six specialist agents build the case — Refund Check, Order Provenance, Evidence Completeness, Fraud Signals, Rebuttal Draft, and Your Approval (see below).
- 3
Decide (EV)
Not every chargeback is worth contesting. The expected-value engine weighs the amount, the real win odds, and the fees, and tells you plainly when fighting would lose money — so you don't pay to chase a dispute that isn't worth it. Fewer pointless fights also keeps your dispute rate down.
Implementation: Expected value = amount × win probability − fees, computed per reason code before anything is drafted.
- 4
Review
By default, Victava assembles the defense and then stops. You read it, edit any line, and approve — usually about a minute's work. If you separately opt into Auto-Pilot, only cases inside your score, amount, timing, deadline, and validation limits can proceed without that click.
Implementation: Manual review is the default. Auto-Pilot is off by default, explicitly enabled by the merchant, and can be paused or disabled at any time.
- 5
File
After manual approval — or after an opted-in case passes every Auto-Pilot guardrail — Victava files the validated evidence with Stripe. The audit trail records the decision path, edits, evidence validation, and final factor score at submission.
Implementation: Submission to Stripe is the only step that touches the bank; every manual or automated path is validated and logged before it runs.
We pull data from Shopify (orders, fulfillment, tracking) to win your Stripe disputes today. Run on Shopify Payments instead of Stripe? It's not live yet — tell us and we'll build it; we ship processor support fast for our first merchants (support@victava.com).
Six specialists handle every dispute
Six clear steps run against every chargeback, and every finding and submission decision stays visible in the audit trail.
Step 2
Order Provenance
Pulls Stripe charge details, Shopify fulfillment, carrier delivery proof, and prior-purchase history — every field stamped with where it came from and when.
Step 3
Evidence Completeness
Flags the missing fields that would lift your win probability the most. No black-box scoring — every gap is named.
Step 4
Fraud Signals
3DS, AVS, CVC, CE 3.0, prior-purchase IP — each factor weighted from documented industry data, not vibes.
Step 5
Rebuttal Draft
Drafts a rebuttal that cites every evidence item by source. Editable line by line — never fire-and-forget.
Step 6
Submission Control
Manual mode stops for your approval. Auto-Pilot is a separate opt-in and proceeds only inside your configured limits.
Win probability — fraudulent dispute
A real factor stack from our live scoring model. Weights and names update automatically when the production model does.
Estimated win probability
95%= 11% base + 84pp factors
- 3DS Authenticated+35pp
- CE 3.0 Eligible+30pp
- CVC Match+10pp
- AVS Full Match+8pp
- Delivery Confirmed+6pp
- Customer IP Available+4pp
- Prior Purchase History+4pp
- Radar Low Risk+3pp
Base rate 11% (industry floor for fraudulent disputes) + 84pp from 8 active factors, capped at 95%. Negative signals (Radar high risk, CVC fail) subtract in production; omitted from this view to keep the visual scannable.
Why control is architectural — not cosmetic.
Manual review is the default: the pipeline stops before submission so you can inspect the evidence and edit the response. Auto-Pilot is a separate opt-in path, not a silent default. It can act only inside the score, amount, delay, deadline, risk, and packet-validation limits you configure, and every decision is recorded.
Nothing bypasses validation, and automatic dispute filing is separate from Early Fraud Warning refund rules. Merchants who want direct review keep manual mode; merchants who want hands-off operation can opt in without giving up caps, exclusions, or an audit trail.
Read the full security & trust page →See it run on your real disputes.
The Free Dispute Leak Audit shows you exactly what would have happened on the last 12 months of your chargebacks. 90 seconds, no card.