Skip to content
AI & Automation

3 Steps: Best Payment Stack for Reducing False Declines 2027

Nov 18, 2026

The short answer: start with the processor you already run

The best payment stack for reducing false declines, for most small ecommerce stores, is the processor you already use, its built-in fraud screen and retry tools, and a review queue for held orders. A paid decision layer such as Signifyd, Sift or PayPal Braintree's Fraud Protection Advanced earns its fee only after you can show which legitimate orders are being refused and what each one is worth.

Stripe, Adyen and PayPal Braintree each sell fraud tools alongside their processing. Signifyd and Sift are sold separately as decision layers, and Spreedly is an orchestration layer for businesses that run more than one processor, so it solves a routing problem rather than a fraud problem.

Pick the layer that matches the failure you can prove. A soft decline or a failed subscription renewal points to retries. A legitimate order blocked by a fraud rule points to screening controls, allowlists and review.

TL;DR: keep your processor, switch on its retries and fraud screen, read your decline reasons for one full billing cycle, and only then compare the paid layers in this guide.

Key Takeaways

  • Start with the processor and fraud tools you already pay for, and switch only when you can name the decline type you are fixing.

  • Retries recover soft declines. Stripe's documentation says hard decline codes such as stolen_card and transaction_not_allowed are not retried automatically.

  • Most decision vendors do not publish a rate card, so get the approved-order fee, the chargeback liability term and approval-rate data in writing before you compare totals.

  • Signifyd's pricing page describes its financial guarantee as separate from a performance SLA that promises a specific approval rate, so read the guarantee scope before you buy.

  • Orchestration with Spreedly earns its place with two or more processors. With one processor it adds a vendor without fixing a decline.

  • Build the flow yourself in Zapier, Make or n8n only if your team will own monitoring, retries and access controls.

Why false declines cost more than the fee

A false decline is a legitimate payment that a fraud filter rejects, and a payment stack is the set of processor, fraud and retry tools that decides whether that payment goes through. The cost is easy to underestimate because a blocked order rarely shows up as fraud in your reports.

Industry estimates are directional. Merchant estimate: up to 5% of legitimate orders declined according to PYMNTS (2026). That article, published March 11, 2026, attributes the figure to PYMNTS Intelligence's December Payments Orchestration Tracker and does not state a sample size, so use it as a reason to measure your own rate, not as a benchmark for your store.

How we evaluated these payment stacks

This guide scores each option on six weighted criteria. It uses only what each vendor's public pages state, and it treats anything unpublished as unconfirmed until the vendor answers in writing. The weights favor the two things that decide whether a false decline gets fixed: the controls you can tune and the recovery tools that act on soft declines. Price transparency and liability terms come next, because they decide whether a small team can budget an option and know who pays when a decision goes wrong.

CriterionWeightMax points out of 100Why it carries this weight
False-decline controls: tunable screening, allowlists, step-up instead of decline25%25A stack that cannot separate good orders from bad ones costs sales directly.
Recovery of soft declines: retries, account updates, network tokens20%20Retries act on soft declines and cannot rescue hard declines.
Published pricing and fee clarity15%15A small team needs a budget it can check before a sales call.
Chargeback or loss liability on approved orders15%15Shows who pays when a decision goes wrong.
Integration effort with your current processor15%15Migrations consume staff time and can create new failure points.
Review queue, logs and override controls10%10You need to see why an order was held or blocked.

Lines labeled "Published" in the profiles below describe what the vendor page says. Lines labeled "Our read" are our analysis.

Two vendors are not profiled. Riskified's pricing and product pages returned access errors during this research, so its terms could not be checked. Kount's pricing page redirects to Equifax, whose identity-fraud page does not describe Kount, and IBS Intelligence reported Equifax's agreement to acquire Kount in January 2021.

Feature matrix: what each option controls

VendorPricing statusDecision scopeReview and overrideRetry or recovery toolChargeback liability on the page
StripePublished for Radar tiersTransaction and account risk, by Radar tierManual reviews on Plus and ProSmart Retries for subscriptions and invoicesNot stated
AdyenIndicative card rates; custom pricing through salesPayments with the RevenueProtect risk layerConfigurable risk rulesAuto Rescue for shopper-not-present paymentsNot stated
PayPal BraintreeQuote-based for Fraud Protection AdvancedCard transactions scored on a 0 to 1000 scaleReview decision, allowlist, blocklistNot described on the feature pageNot stated
SignifydQuote-based; percentage of approved order totalApprove or decline per orderGuarantee scope set in consultationNot described100% financial guarantee against chargebacks
SiftQuote-basedReal-time scoring in under 150 millisecondsStop, step up, or route to reviewNot describedNot stated
SpreedlyVault published; Performance Optimization quote-basedOrchestration across gateways and fraud toolsProtect and Composer workflowsOptimize for routing and auth ratesNot stated

Stripe, Adyen and PayPal Braintree sell fraud tools next to their processing, so their screens are easiest to test first. Signifyd, Sift and Spreedly are separate purchases, so their terms need the closest reading. The "Not stated" entries mean the vendor page we read did not say who bears chargeback losses. That is a question to ask every vendor before signing.

Pricing: published figures and what they would cost

Pricing checked November 14, 2026.

Stripe lists Radar pay-as-you-go as a per-screen rate. Radar pay-as-you-go: $0.05 per screened transaction according to Stripe (2026). Radar Lite's fraud alerts are included at no additional charge for businesses on standard payments pricing, and Stripe's Radar documentation says Radar screens every attempt type, including declined ones, and charges only when it screens a transaction.

On the monthly tab, Radar Standard monthly: from $10 per month according to Stripe (2026), with Plus starting at $14 and Pro at $20 on the same page.

Adyen's pricing page lists an indicative Mastercard and Visa rate of $0.13 + Interchange++ + 0.60% according to Adyen (2026). The page says its fees are indicative and that custom pricing runs through sales. Its currency labeling is unclear, because the same $0.13 fixed fee also appears on euro-priced rows.

Spreedly publishes a starting price for its vault. Vault plan: starting at $750 per month according to Spreedly (2026). The Performance Optimization tier has no published price on that page and points to Contact Sales.

PayPal Braintree's pricing page redirected to a PayPal fee page that returned an access error during this research, so no base processing rate is cited here. The Fraud Protection Advanced feature table marks "Additional fees" as "Yes" without an amount, per PayPal Developer.

VendorPlan or productFixed feeVariable feeIllustrative volumeExample monthly costPricing checked
StripeRadar pay-as-you-goNone stated$0.05 per screened transaction10,000 screened transactions$500November 14, 2026
StripeRadar Standard, monthly$10 per month, startingNot stated10,000 transactionsNot computable from the pageNovember 14, 2026
AdyenCard processing, indicative$0.13 per transaction0.60% plus Interchange++10,000 transactionsNot computable without interchangeNovember 14, 2026
SpreedlyVault$750 per month, startingNot stated10,000 transactions$750November 14, 2026

The table's volume is an illustration, not a forecast. The Stripe and Adyen rows show that the per-transaction and fixed components cannot be totaled without your own order mix.

VendorProductPricing statusWhat the page publishesAsk for in writing
PayPal BraintreeFraud Protection AdvancedQuote-basedAdditional fees marked "Yes," no amountThe fee amount and its billing basis
SignifydOrder decisions with guaranteeQuote-basedPercentage of the approved order total; rate not publishedThe rate and the guarantee scope
SiftPayment ProtectionQuote-basedNo price and no liability terms on the pageThe price and who bears chargeback losses
SpreedlyPerformance OptimizationQuote-basedContact Sales; no price on the pageThe price and the routing scope
AdyenRevenueProtectQuote-basedNamed as a risk layer; legal terms say some features are premiumWhich features are premium and what they cost

Three steps to set up the stack

  1. Pull one full billing cycle of declines and held orders, and sort them into three groups: soft declines that a retry might recover, hard declines that need a new payment method, and legitimate orders held by a fraud rule.

  2. Turn on what your processor already offers: automatic retries for subscriptions and invoices, the fraud screen tier you can afford, and allowlists for repeat customers. If you use automatic retries on Adyen, set up webhooks first, because Adyen's Auto Rescue documentation says to do so before use.

  3. Add a paid decision layer only if the held-order value from step 1 is larger than the vendor's fee, and the vendor will confirm its approval-rate data and liability terms in writing.

Worked example: one month of held orders

Illustrative numbers, not vendor data: a store screens 10,000 transactions a month on Radar pay-as-you-go at the $0.05 rate cited above, which is $500 in screening fees. If its holds catch 200 legitimate orders averaging $80, that is $16,000 in sales at risk, or 32 times the screening fee. Suppose 40 subscription renewals fail in the same month and half are soft declines, which leaves 20 retry candidates; a team that logs each invoice.payment_failed event and tracks its attempt_count can see which renewals Stripe is still retrying and which need a customer to update a card, because retries stop helping on hard declines. The screening fee is rarely the first number to fix. The 200 held orders are the larger line item, so a decision layer should be judged on that number, not on the $500.

Where the pain lives: renewals and held orders

A proposed, configurable US Tech Automations workflow for the renewal case would start when a failed-invoice webhook arrives, read attempt_count from the payload, and leave soft declines on the processor's retry schedule while logging each attempt to a shared sheet. For hard declines, which Stripe does not retry without a new payment method, the workflow would look up the decline code on the related payment record and create a card-update task for a named staff member. A person would approve the outreach email before it is sent. The prerequisites are a webhook endpoint the store controls, read access to invoice data, and a sheet or CRM the team already uses. This is a design to confirm against your own processor account, not a live deployment.

A second proposed US Tech Automations workflow covers held orders from a fraud tool. The trigger is a hold decision from Braintree's Review decision, or a decline decision in a Signifyd or Sift export. The action pulls the order value, the customer's order history from the store export, and the rule that fired. The output is a daily review queue sorted by order value. The prerequisites are decision-export or API access from the fraud vendor and an order-history export from the store platform. The human review points are that a person releases or cancels every held order, and no automation overrides a fraud decision.

Vendor profiles

Stripe

Best fit: Stores already on Stripe that sell subscriptions or invoices, where automatic retries and the failed-invoice webhook matter most.

Published: Radar comes in Lite, Standard, Plus and Pro tiers. Radar Plus adds custom rules, backtesting and manual reviews. Smart Retries is enabled by default for new Stripe accounts and applies to subscription and invoice payments.

Limitations: Smart Retries does not cover one-time checkouts. Stripe does not retry payments when the issuer returns a hard decline code such as stolen_card or transaction_not_allowed, so a card update is the only fix for those. Radar's per-screen fee applies to declined attempts as well, so screening every attempt adds cost even when nothing is approved.

Implementation: Turn Smart Retries on or set a custom schedule in the Dashboard, then subscribe to invoice.payment_failed and track attempt_count to see how many retries have run. The Smart Retries documentation lists the non-retryable cases.

Our read: For a Stripe merchant, reading decline codes first costs less than adding a second fraud vendor.

Adyen

Best fit: Merchants selling in several markets or through many payment methods who want a sales-led contract that bundles risk tools with processing.

Published: Adyen's pricing page describes a fixed processing fee plus a fee set by payment method. Its legal terms say some additional services are standard and others premium, but they do not say which is which. RevenueProtect is described there as a risk management solution with a configurable risk rules component.

Limitations: Because pricing is indicative and sales-led, a small store cannot budget from the public page alone. Auto Rescue covers shopper-not-present payments such as subscription renewals, and Adyen's documentation says payments declined permanently, such as those on a closed account, are not retried.

Implementation: Set up webhooks before enabling Auto Rescue, because they report each retry, whether another retry is scheduled, and the end of the rescue process. The Auto Rescue documentation covers the eligibility rules.

Our read: Ask Adyen which risk features are premium before you compare its card rate with anyone else's.

PayPal Braintree

Best fit: Merchants already on PayPal Braintree who want a machine learning risk score, a review queue and allowlists without moving processors.

Published: Fraud Protection Advanced scores each card transaction on a scale from 0 to 1000 according to PayPal Developer (2026), where higher means riskier, and rejects transactions above the filter's threshold. It supports allowlists and blocklists, rules built from more than 200 built-in data features, and a Review decision that sends a payment to manual review.

Limitations: A decision of Not Evaluated sends the payment to the processor by default, so a timeout or error passes the payment without a fraud check. The fee amount is not published on the pages we could read.

Implementation: Filters and rules are configured in the Control Panel, and webhooks can report whether a reviewed transaction was approved or rejected.

Our read: A good fit when the review queue matters more than the fee, provided the fee is confirmed in writing first.

Signifyd

Best fit: Merchants who want the vendor to make approve-or-decline decisions and to carry fraud chargebacks.

Published: Signifyd's pricing page describes a 100% financial guarantee against chargebacks according to Signifyd (2026), configurable from fraud-only coverage to full liability shift. The page charges a percentage of the approved order total, charges nothing for orders declined as fraud, and says the rate depends on product type, vertical, volume and ticket size. It separates the financial guarantee from a performance SLA that promises a specific approval rate. Its customer case studies report approval rates, which are vendor-reported.

Limitations: No rate is published, so total cost is unknown until a consultation. A guarantee shifts chargeback losses to the vendor.

Implementation: Integration is followed by a consultation to set approval-rate targets and the guarantee's scope.

Our read: A vendor that absorbs chargeback losses has a reason to be conservative on borderline orders, so ask how it sets that threshold before you rely on its approval claims.

Sift

Best fit: Merchants who want a real-time risk score and their own decision rules, and who are prepared to carry chargeback losses themselves.

Published: Vendor-labeled acceptance rate: 99.4% according to Sift (2026), with no timeframe or method on the page. Risk scoring runs in under 150 milliseconds according to Sift (2026), and suspicious transactions can be stopped, sent for extra verification, or routed to review.

Limitations: The page publishes no price and does not say who bears chargeback losses. It never uses the term "false positive," so the false-decline rate has to be requested directly.

Implementation: The page's calls to action are "Get started" and "request a demo," so expect a sales-led onboarding.

Our read: Sift fits a store that wants control over its own rules and can absorb chargebacks. It is a weaker fit for a store that wants a guarantee.

Spreedly

Best fit: Merchants with two or more processors, or a plan to switch processors, who want card data held in a vault they can move.

Published: Vault plan: starting at $750 per month according to Spreedly (2026). The vault page describes payment token ownership and protection against vendor lock-in. The Optimize product is described as workflow-driven payments intelligence for smarter routing and higher auth rates.

Limitations: Orchestration adds a vendor and an integration to maintain. With one processor and a simple checkout, it does not fix a fraud rule or a soft decline by itself.

Implementation: Move card tokens into the vault, connect gateways, then add fraud tools one at a time.

Our read: Look at Spreedly after you have a second processor, not before.

Who this is for

This guide is for an owner or operations lead at a small ecommerce business who can name the decline type they are fixing and who has someone able to read processor webhooks or exports each day.

Red flags: you cannot export declines by reason, you need a published rate card before any sales call, or nobody on the team will review held orders.

Choose Stripe if you already run on it for subscriptions or invoices. Choose PayPal Braintree if you already run on it and want a review queue. Choose Adyen only if you have a multi-market plan and accept a sales-led contract. Choose Signifyd if fraud chargebacks are your main loss and you want the vendor to carry them. Choose Sift if you want real-time scores and your own rules. Choose Spreedly if you already run two or more processors.

Common mistakes

  • Buying a fraud vendor before reading a full billing cycle of decline codes.

  • Treating retries as a fix for hard declines.

  • Comparing only the per-transaction fee and ignoring approved-order fees, screening fees on declined attempts and liability terms.

  • Using a vendor's case-study approval rate as the rate you should expect.

  • Turning on automatic retries before webhooks are set up, which Adyen's Auto Rescue documentation says to do first.

  • Assuming the vendor carries chargeback losses without seeing that term in writing.

Build it yourself or buy the stack?

Zapier, Make and n8n can run the flows described above. When they are configured, they can keep run histories, retries, error branches and audit evidence. The buyer still has to design and own observability, idempotency so that a repeated webhook does not create two review tasks, escalation paths, access controls and ongoing maintenance. A proposed US Tech Automations build would configure those same pieces, but it needs the same prerequisites: webhook or export access, a named owner for each queue, and a person who approves every outreach email and every release.

Questions buyers ask

When NOT to use US Tech Automations: a simpler setup wins in three cases. If your processor's built-in screen and retries already cover the declines you measured, buy nothing new. If your processor or fraud vendor cannot export or send webhook data, the workflows above cannot start, and a manual spreadsheet review is the honest option. If your real need is a chargeback guarantee rather than better routing or retries, a guarantee product such as Signifyd fits that job more directly than any workflow.

Which payment stack should I choose first?

Start with the processor and fraud tools you already run, because switching processors only fixes a decline problem when the processor is the cause. Measure one full billing cycle before you add anything.

Do retries fix false declines?

Retries fix some soft declines and cannot fix a hard decline, because Stripe does not retry hard decline codes such as stolen_card until a new payment method is supplied. Retries are a recovery tool, not a fraud-screening tool.

Is a chargeback guarantee worth the fee?

It can be, but only if the guarantee's scope covers the losses you actually have. Signifyd's guarantee is configurable from fraud-only coverage to full liability shift, so check which one you are buying before you compare its approved-order fee.

Why don't most decision vendors publish prices?

Vendors that charge per approved order or by volume usually quote a rate after they see your order mix. Signifyd says its rate depends on product type, vertical, volume and ticket size, so a public number would not match most stores.

Do I need orchestration with one processor?

Usually not. Spreedly's value is routing and vaulting across several processors, and its Performance Optimization tier has no published price, so a single-processor store pays for a layer it may not need.

What should I ask a vendor in writing?

Ask for the approved-order or per-transaction fee, who pays chargebacks on orders the vendor approved, and approval-rate data for merchants with an order mix like yours. A vendor that will not answer those three questions in writing is not ready to be compared.

Bottom line

The best payment stack for reducing false declines for most small stores starts with the processor and fraud tools already in place. Measure a full billing cycle, then compare the paid layers in this guide against your own held-order numbers rather than against a vendor's case study. If the measured losses justify a review queue or a renewal workflow, see how US Tech Automations configures this against your own processor and fraud-tool exports before you commit.

About the Author

Garrett Mullins
Garrett Mullins
Workflow Specialist

Helping businesses leverage automation for operational efficiency.