Skip to main content

Vendor Selection (Operator Field Manual)

TL;DR
  • Start with the fraud tools included in your processor plan. Inventory what's actually switched on and what costs extra (Stripe Radar; Adyen Protect, or RevenueProtect on some integrations). Only pay for a dedicated vendor once false positives or fraud losses cost more than the vendor does
  • Write down the problem before you take a single demo. "Get chargebacks from 0.8% to under 0.5%" can be evaluated. "Improve fraud" can't
  • Run a scoped A/B pilot with a clean control before you commit. A vendor's case studies were written about someone else's traffic
  • Contract for the exit: data export, token portability, sane notice periods. That's what protects you when the tool underperforms or the company gets acquired

Pick tools by problem and volume tier, not by brand fame. Start with what's built in, then add only what you can measure, because a tool you can't measure is a tool you'll never be able to cancel.

Looking for Fraud Vendors Specifically?

For fraud-specific vendor evaluation, see the Fraud Vendor Selection Guide. This page covers choosing fraud and chargeback tools.

What Matters​

  • Define the problem first. Fraud guarantees, identity checks (IDV), alerts and chargeback ops are four different purchases.
  • Match the tool model to your volume. Guarantees suit a team with no capacity; rules and decisioning suit a team that wants control.
  • Start with the processor tools your plan actually includes and tune them properly before you buy more.
  • Pilot on a clean A/B split. Measure auth rate, fraud, chargebacks and support tickets together, because a tool that fixes one and wrecks another isn't a win.
  • Contract for exit: data export, token portability, notice periods.

Minimum Viable Stack by Volume​

  • Under $100k/mo: a starting point, adjusted for your products, geography and threats: your processor's included fraud tools, AVS and CVV checks, 3DS wherever the rules or the issuer require it (risk-based beyond that), and enrichment through Order Insight or Consumer Clarity. That's a baseline to test against, not a promise you need nothing else.
  • $100k-$1M/mo: add alerts (RDR plus Ethoca or CDRN). Consider piloting a guarantee vendor on one segment. Start IDV on your highest-risk flows.
  • Over $1M/mo: a dedicated fraud tool with rules and machine learning (ML), automated alert handling, IDV orchestration, and a dispute vendor if your volume warrants one.

Tool Categories and When to Use​

  • Fraud guarantees (Signifyd, Forter, Riskified). Buy these when you haven't got the headcount and you want specified losses covered by contract. Read which transactions and reason codes are covered, what's excluded, and what evidence you still owe. A guarantee reimburses covered losses; your network duties and lost good sales stay yours. Watch what they do to your approval rate, because that's how they manage their own losses.
  • Rules and decisioning (Sift, Kount, decision APIs). Buy these when you want control and you've got an analyst to spend on tuning. Without that person, you've bought a dashboard.
  • Processor-native rules (Stripe Radar, Adyen, Braintree). Check which features your integration and plan actually include, and tune those before you buy anything else. Adyen calls its current risk product Protect; some integrations still document RevenueProtect.
  • IDV and KYC (know your customer checks: Persona, Sardine, Alloy). For account creation and account takeover problems. Ask whether the product covers your channel. An onboarding check isn't card-present protection unless the vendor says it covers that flow.
  • Alerts and enrichment (Verifi RDR and CDRN, Ethoca, Order Insight, Consumer Clarity). Different jobs: RDR resolves qualifying disputes automatically under your rules; alerts warn you; Order Insight and Consumer Clarity put richer purchase detail in the bank's channels. None of that changes your billing descriptor. Set the descriptor itself with your processor.
  • Chargeback management vendors. For high dispute volume. Make them show you win rate broken out by reason code, and confirm they handle Visa Compelling Evidence 3.0 (CE3.0).

Pilot Plan​

Start in shadow mode: the tool scores orders but doesn't change any payment decision. If you run a live split pilot, write down the risk owner, eligible orders, loss cap, rollback, stop rule and outcome window first. Two to four weeks is a planning example. Call auth and support results only on an adequate, comparable sample, and wait for the predeclared outcome window before judging fraud or chargebacks.

  • Scope: one segment, whether that's a country, a payment method or a traffic slice. Keep a clean control group.
  • Metrics: auth rate, fraud rate, chargeback ratio, false positives, customer support tickets.
  • Success: net revenue lift, meaning approved-good revenue minus fraud and fees, with customer experience holding steady (tickets flat or down).
  • Exit: a documented rollback, with your original routing ready to switch back on.

Ask Before You Buy​

  • "How is this priced? Per transaction, percent of volume (GMV), or percent of what you approve or cover? Are there minimums?"
  • "Can we export our decisions, scores and device data? All of it?"
  • "How do we turn this off, and how fast? What does it cost us to leave?"
  • "Which brands and regions do you cover? Do you support 3DS? Do you support CE3.0 evidence?"
  • "What are your uptime, response latency and support service levels (SLAs), in writing?"

Where This Breaks​

  • No control group. You can't prove lift, so you'll renew forever on vibes.
  • Blended traffic in a "pilot." It can hide false positives and selection bias. Require stable assignment and joined decision logs.
  • Guarantee vendors quietly tightening. Lower approvals can cut both the vendor's covered losses and your good revenue. Score both.
  • Overlapping tools. Processor rules plus vendor rules can block the same good order twice. Join both tools' decision logs so you can see it.

Next Steps​

Picking your first vendor?

  1. Check minimum stack by volume - What you actually need
  2. Understand tool categories - Guarantees vs rules vs alerts
  3. Plan your pilot - Scoped test with control

Evaluating a specific tool?

  1. Ask before you buy - Pricing, data, portability
  2. Watch for pitfalls - No control group, double-blocking
  3. Run clean A/B - Measure auth, fraud, chargebacks

Need detailed evaluation?

  1. Review selection guide - Full process
  2. Explore fraud vendors - The options and their differences
  3. Set up experimentation - How to test a change safely

Something wrong on this page? Email matt at paymentsandrisk.com and say what needs fixing.