Opening: Today's Fraud Picture

• How many payment transactions does your organization process per month? Options: Under 100k, 100k-500k, 500k-2M, 2M-10M, Over 10M
• What percent of recent monthly transactions are flagged by your current fraud system? Options: Under 0.1%, 0.1%-0.5%, 0.5%-1%, 1%-2%, Over 2%
• Who on your team routinely reviews authorization declines and investigates suspected account takeover alerts? Options: Fraud Ops, Payments Ops, Risk/AML, Customer Experience, Legal/Compliance, Other
• When was the last quarterly fraud loss report that caused leadership concern, and what changed compared to the prior quarter?
• On average, how many Reg E dispute cases does your team open per month? Options: Under 100, 100-500, 500-2,000, 2,000-10,000, Over 10,000

Where the Current System Breaks First

• If a single detection failure could trigger a regulatory exam next quarter, which type of failure would that be? Options: Missed account takeover, Synthetic identity bypass, High-value fraud slipping through, Dispute processing timeline failure, Other
• Describe the most recent account takeover case that bypassed your rules, including the first missed signal and the customer impact.
• Walk me through the last week of false-positive declines that generated customer complaints, and the downstream cost in calls, refunds, or lost customers.
• Which metrics show the biggest pain for you right now, fraud loss dollars, false-positive rate, dispute processing cost, or customer churn? Options: Fraud loss dollars, False-positive rate, Dispute processing cost, Customer churn, Regulatory risk
• Estimate the operational hours per week your team spends reconciling disputed transactions under the current workflow. Options: Under 10 hrs/week, 10-40 hrs/week, 40-160 hrs/week, Over 160 hrs/week
• What single technical or data gap in your environment would cause you to pause any vendor trial? Options: Missing transaction-level history, No API access to authorizations, Unreachable data owners, Regulatory or legal hold, None; we can work around gaps

Who Moves When Fraud Lands on Their Desk

• Who feels the most pressure when a spike in account takeover complaints appears, and what do they typically do first? Options: Head of Fraud Ops, VP Payments/Risk, Investigation Manager, CTO/Head of Engineering, Customer Experience Lead, Other
• Which teams must be involved for a scoring integration and who typically owns the API changes? Options: Fraud Ops, Engineering, Platform/Integrations, Product, Compliance/Legal
• List the roles that must sign off on a pilot and the usual order they review new vendors. Options: Fraud Ops Manager, Risk Director, Head of Payments, Procurement, Legal
• Could you name the person who would clear access to transaction history for model training and indicate their likely availability window?
• If the person who controls data access is unavailable for more than four weeks, does that stop the project? Options: Yes, it stops the project, No, we have a backup plan, It depends on mitigation and leadership signoff

Money Talk — What a Win Really Looks Like

• To move to production this quarter, how large a reduction in fraud losses or false positives would you require? Options: 10% reduction, 15-20% reduction, 25-40% reduction, Over 40% reduction
• How many dollars per month in reduced chargeback expense or operational cost would offset vendor fees for you? Options: Under $10k, $10k-$50k, $50k-$200k, Over $200k
• Name the KPIs your leadership will expect at pilot completion, for example detection lift, false-positive delta, or Reg E compliance. Options: Fraud dollars saved, False-positive rate, Time to provisional credit, Operational cost per case, Customer complaints
• Assuming the trial meets those KPIs, could your procurement team sign within 30 days and who would need to approve? Options: Yes, No, Needs legal review, Depends on pricing
• Identify the single contractual or budgetary blocker that would stop procurement from signing immediately if the pilot shows the promised results. Options: Contract terms, Budget constraints, Legal clauses, Data access limitations, Other

Trial Design That Actually Proves Something

• Imagine running a 60-day parallel scoring test with the platform scoring every authorization in real time alongside your incumbent, what realistic failure modes do you expect?
• List the types of data feeds you could share for parallel scoring and the earliest you could deliver a representative 60-day history. Options: Full transaction history, Aggregated summaries only, Tokenized data, Processor-level logs, Other
• Describe the trial acceptance criteria you use today, including thresholds for detection lift, acceptable change in false positives, and Reg E timeline performance.
• Name the people who will define scoring thresholds and the person who can adjust them during the pilot. Options: Fraud Lead, Data Science Lead, Ops Manager, CTO, Other
• Outline how you will measure model adaptation speed when new fraud patterns appear and the timeline that matters to your operations team.
• Would inability to provide transaction level history with required fields within three weeks disqualify the pilot? Options: Yes, disqualify, No, we can extend timeline, We would negotiate reduced scope

Integration and Data Reality Check

• Call out the single integration gap that would derail a go-live in under two months.
• Please enumerate the core systems that must connect for full functionality, for example your authorization switch, card processor, and case management tool. Options: Authorization switch, Card processor, Case management, Data warehouse, Payments gateway, Other
• Do APIs exist for those systems and who owns them? Options: Yes, documented public APIs, Yes, private/internal APIs, No APIs, only batch extracts, Unknown
• Could your team deliver a sample of normalized transaction data with the required fields within 14 days? Options: Yes, No, Need approval, Requires anonymization
• Estimate how many full time engineers you can dedicate to integration during the pilot and their typical daily bandwidth. Options: None available, 1-2 engineers, 3-5 engineers, More than 5
• Provide the named owner accountable for delivery and indicate whether the project would stop if that owner is unavailable. Options: Named owner with backup, project continues, Named owner without backup, project delayed, No named owner, project would stop

Operational Resilience and Compliance Knockouts

• Assume the dispute automation missed Reg E provisional credit timelines for 5% of cases during cutover, what would that cost in regulatory exposure or fines?
• Do you currently track Reg E and Reg Z timelines in a case management system or is it manual today? Options: Automated in case management, Manual spreadsheets, Mixed automated and manual, Not tracked consistently
• Identify the compliance stakeholders who need to review pilot runbooks and who signs off on customer notification templates. Options: Chief Compliance Officer, Head of Legal, BSA Officer, Risk Committee, Other
• In the last 24 months, how many regulatory inquiries related to dispute timelines have you received? Options: 0, 1-2, 3-5, More than 5
• Could missing audit trails for automated dispute decisions block your compliance team from approving production? Options: Yes, No, Probably, Need to review specifics
• Would an inability to retain immutable audit trails for 12 months prevent compliance from approving production? Options: Yes, would block production, No, proceed with compensating controls, Conditional approval with additional monitoring

Competitive Landscape — Who Else Is on Your Shortlist

• Tell me which categories you are actively considering instead of engaging an external platform, such as incumbent vendor, homegrown rebuild, or processor built-in controls. Options: Incumbent vendor, Homegrown solution, Processor built-in controls, Another ML vendor, Undecided
• Provide the alternatives by category you have benchmarked in the past 12 months, and note any that performed well on detection or dispute automation. Options: Incumbent vendor, Homegrown, Processor capability, New third-party vendor, Other
• Under what circumstances would your current system be sufficient so you would not change vendors?
• Has anyone on your team proposed solving this internally without a vendor, and if so what scope and timeline were suggested? Options: Yes, full rebuild proposed, Yes, augment existing rules, No internal proposal, Pilot by another vendor suggested
• Compare the alternatives you are considering on three priorities: reduction in fraud dollars, false-positive impact, and dispute processing cost. Options: Prefer lower fraud losses even if higher cost, Prefer lower false positives even if slight fraud increase, Prefer lower operational cost, Prefer simplest integration
• Make a decision: would matching detection lift without dispute automation be sufficient for you to switch today? Options: Yes, would switch, No, dispute automation is required, Maybe, need more data, Undecided

Acceptance Criteria and The Fast Track Question

• Suppose the trial shows a 30 percent fraud loss reduction but a 10 percent increase in false positives, would you accept that tradeoff? Options: Acceptable tradeoff, Unacceptable, Need to review customer impact data, Depends on dispute automation SLA
• State the exact numerical thresholds for detection lift, acceptable change in false positives, and Reg E timeline performance that would count as success for you.
• Name who must sign the go or no go acceptance within your organization and the typical decision timeline once pilot results are shared. Options: Head of Fraud, CRO, Procurement, Legal, Other
• Once your acceptance criteria are met in the pilot, how soon would your team be ready to begin production integration work? Options: Immediately, Within 2 weeks, Within 1 month, Over 1 month
• Is there an internal procurement or legal condition that, even if all technical criteria are met, would prevent you from signing within 60 days? Options: Yes, major condition, No, nothing prevents signing, Maybe, requires exceptions

Next Steps and Shortlist Commitments

• Given resolution of your top two blockers within the first week of the pilot, would you commit to a formal evaluation cadence? Options: Yes, commit to cadence, Maybe, need leadership signoff, No, not yet
• Please supply the pilot owner's role and preferred contact method for day to day communication.
• State how frequently you want the seller to report pilot metrics and the preferred format for review. Options: Daily dashboard, Weekly summary, Biweekly review, Ad hoc on request
• Specify the top three risks you want the seller to monitor in real time during the trial.
• Finally, if the pilot delivers the agreed outcomes, what is the earliest date you would want to begin contract negotiations? Options: Immediately, Within 2 weeks, Within 1 month, Longer than 1 month, Unsure

• success_criteria
• stakeholders
• gaps
• current_state
• decision_readiness
• desired_state
• current_state
• success_criteria
• decision_readiness
• stakeholders
• desired_state
• gaps
• stakeholders
• current_state
• decision_readiness
• decision_readiness
• decision_readiness
• decision_readiness
• decision_readiness

• Deploy real-time authorization scoring API
• Train institution-specific ML fraud models
• Execute parallel scoring alongside legacy system
• Configure detection thresholds and rule calibration
• Integrate with authorization switch and processor
• Automate investigator workqueue and case prioritization
• Automate dispute case routing and escalation
• Automate Reg E and Reg Z timeline tracking
• Automate provisional credit calculation and tracking
• Generate compliant dispute notification letters
• Deploy customer-facing digital dispute portal
• Automate chargeback lifecycle management and remediation
• Reconcile dispute outcomes with settlement and ledger
• Provide staff training on platform operations and workflows
• Ongoing model performance monitoring and retraining

Deploy real-time authorization scoring API

• Do you require synchronous scoring in the authorization flow (decision returned before approval) or asynchronous scoring for post-authorization? Options: Synchronous (pre-authorization), Asynchronous (post-authorization), Both
• Which endpoints will the integration use for authorization requests (authorization API, switch host, webhook)? Options: Authorization API (real-time), Switch host (ISO 8583), Webhook for async scoring, Other
• How many transactions per second (TPS) does your authorization switch process at peak? Options: <100 TPS, 100-1,000 TPS, 1,000-5,000 TPS, 5,000+ TPS
• Who on your team will manage API key and credential rotation for the authorization endpoints? Options: Fraud operations lead, Platform engineering, Security team, Other
• When do you plan to cut over the authorization scoring from parallel to live traffic once SLA and accuracy checks are met? Options: After confirmed SLA tests, After compliance sign-off, After a successful canary window, Custom date

Train institution-specific ML fraud models

• Where is your historical transaction data stored for model training (on-prem data lake, cloud storage, or processor reports)? Options: On-prem data lake, Cloud storage bucket, Processor-provided reports, Hybrid
• Provide the lookback window you prefer for training and any blackout windows for seasonal anomalies (in months). Options: 3 months, 6 months, 12 months, 24 months, Custom
• List the fraud outcome labels available in your dataset (chargeback reason codes, internal confirmed fraud, dispute disposition). Options: Chargeback codes, Internal confirmed fraud tags, Customer dispute outcome, No labels / need assistance
• Describe how you currently validate model performance and which metrics you can provide (AUC, precision at threshold, false-positive rate).
• Specify the feature sets that must be excluded from training due to privacy or contractual limits (full PAN, raw PII, third-party enrichment). Options: Full PAN, Raw PII, Device fingerprint hashes only, Third-party enrichment excluded, No exclusions

Execute parallel scoring alongside legacy system

• Confirm if you can provide a real-time mirrored authorization feed from your switch for the parallel scoring window. Options: Yes, 100% mirror available, Yes, sampled mirror available, No, only batch extracts, We need assistance to mirror
• Are there any transaction types to exclude from the parallel run (recurring billing, internal transfers, BIN test transactions)? Options: Recurring, Internal transfers, BIN test transactions, None
• Will you accept the parallel scoring trial if it meets pre-agreed detection uplift, false-positive delta, and Reg E timeline parity over the defined 60-day window? Options: Yes, No, Need to define thresholds
• Do you plan to log both incumbent and new platform decisions for each mirrored transaction for side-by-side audit? Options: Yes, full decision and metadata, Yes, summary decisions only, No, metadata only
• Is there a preferred sampling rate if you cannot mirror 100% of authorizations (for example 100%, 50%, stratified by amount)? Options: 100% mirror, 50% stratified by amount, Sampled by merchant category, Custom sampling

Configure detection thresholds and rule calibration

• Which environment will you use for threshold tuning (development sandbox, QA mirror, production-safe canary)? Options: Development sandbox, QA mirror, Production canary, Other
• Which processor or acquirer constraints affect rule calibration for specific MIDs or merchant categories? Options: Acquirer limits, Issuer constraints, Merchant agreements, None
• Which switch response codes should automatically override model decisions (for example, issuer declines, offline approvals)? Options: Issuer decline, Offline approval, Manual override codes, None
• Which message format should the platform use to signal review actions back to the switch (ISO 8583 response field modifications, JSON callback)? Options: ISO 8583 modifications, REST JSON callback, Webhook event, Other
• Which BIN ranges require bespoke calibration due to corporate card behaviors or high-volume issuers?

Integrate with authorization switch and processor

• Which settlement ledger entries should be annotated when scored authorizations flow through the integration (tags for provisional credit, disputed flag)? Options: Provisional credit tag, Disputed flag, None
• Which case types should the integration surface to the processor (authorization hold, routed review, dispute flag)? Options: Authorization hold, Routed review, Dispute flag, Other
• How often should the platform poll the switch versus receive push notifications for authorization events? Options: Push notifications preferred, Poll every second, Poll every minute, Batch updates
• How long should a persistent connection to the switch remain idle before re-authentication is required? Options: 30 seconds, 5 minutes, 30 minutes, Custom
• How do you prefer to manage certificate rotation and key exchange with the processor (scheduled rotation, automated rotation via PKI)? Options: Scheduled rotation, Automated PKI rotation, Manual on-demand, Other

Automate investigator workqueue and case prioritization

• How will you define prioritization weights in the investigator workqueue (score weight, transaction amount, repeat offender flag)?
• Which thresholds should trigger auto-escalation to a senior investigator (for example score > X and amount > Y)?
• Which escalation paths must be available from the investigator UI (team lead, compliance, legal, fraud analytics)? Options: Team lead, Compliance, Legal, Fraud analytics
• Which training materials should be embedded in the investigator UI for suggested actions (case playbooks, evidence checklist)? Options: Case playbooks, Evidence checklist, Model rationale snippets, Other
• Which roles need read-only versus edit access in the workqueue (junior investigator, senior investigator, compliance reviewer)? Options: Junior investigator (edit), Senior investigator (edit), Compliance (read-only), Other

Automate dispute case routing and escalation

• Which credentials will your routing system accept for automated hand-off (SAML assertion, API token, signed webhook)? Options: SAML assertion, API token, Signed webhook, Other
• Which endpoints should receive routed dispute cases (merchant acquirer endpoint, internal case API, third-party vendor)? Options: Merchant acquirer endpoint, Internal case API, Third-party vendor, Other
• Which response codes from the network should trigger immediate routing (chargeback notification, preliminary dispute)? Options: Chargeback notification, Preliminary dispute, Reversal notice, Other
• Which KPIs should be used to measure routing effectiveness (time-to-first-action, routing accuracy, SLA breaches)? Options: Time-to-first-action, Routing accuracy, SLA breaches, Other
• Which acceptance criteria will validate the dispute routing and escalation trial (routing accuracy %, time-to-first-action SLA, correct mapping of reason codes)?

Automate Reg E and Reg Z timeline tracking

• Which data extracts will you provide to support Reg E and Reg Z timeline automation (settlement files, dispute logs, customer correspondence)? Options: Settlement files, Dispute logs, Customer correspondence, All of the above
• How will evidence of provisional credit deadlines be captured for audits (timestamped notices, signed acknowledgments)? Options: Timestamped notices, Signed acknowledgments, System logs only, Other
• Provide sample Reg E or Reg Z notification scenarios that must be supported during automation (for example unauthorized ACH debit, card-present fraud).
• Do you require configurable calendar rules for calculating Reg E and Reg Z deadlines (banking days, weekends, jurisdictional holidays)? Options: Banking days only, Banking days + jurisdictional holidays, Custom calendar
• Which internal teams must receive automated timeline alerts (compliance, fraud operations, customer service)? Options: Compliance, Fraud operations, Customer service, All listed

Automate provisional credit calculation and tracking

• How many provisional-credit events do you average per month that will flow through automation? Options: <100, 100-1,000, 1,000-10,000, 10,000+
• Who on your team will reconcile provisional credit postings with the general ledger? Options: Finance lead, Reconciliation analyst, Shared responsibility, Other
• When do provisional credits typically post relative to transaction settlement in your current process (same day, next business day, after settlement)? Options: Same day, Next business day, After settlement, Varies
• Where is your ledger system that must receive provisional credit transactions (GL system name or export endpoint)?
• Provide the business rules you use today to cap provisional credit amounts, if any. Options: Full amount, Capped at average monthly debits, Capped per product, No cap

Generate compliant dispute notification letters

• List the notification templates you currently use for dispute communications (initial notice, provisional credit notice, final outcome). Options: Initial notice, Provisional credit, Final outcome, Other
• Describe how identity proofing is validated before sending dispute notifications (last four PAN, DOB, security question).
• Specify any regulatory language or jurisdiction-specific disclaimers that must be embedded in letters.
• Confirm if letters must be archived to your records retention system with immutable timestamps for audit. Options: Yes, No, Archive externally
• Are there accessibility or language requirements for mailed dispute notices in your customer base? Options: Yes, No

Deploy customer-facing digital dispute portal

• Will you allow customers to initiate disputes via the portal for all card products or restrict by product? Options: All products, Restricted by product, Pilot product only
• Do you plan to integrate the portal with your online banking single sign-on for customer identity and smooth navigation? Options: Yes, SSO integration, No, separate auth, Consider later
• Is there a maximum dollar amount for disputes that can be submitted through self-service versus requiring agent intervention? Options: No limit, Limit by product, All require agent review
• Which environment will host the customer portal (your public cloud tenant, a dedicated hosted instance, or embedded widget)? Options: Public cloud tenant, Dedicated hosted instance, Embedded widget
• Which processor branding or customer-facing messaging guidelines must the portal comply with?

Automate chargeback lifecycle management and remediation

• Which switch or network notifications indicate a chargeback has been posted and must trigger lifecycle automation? Options: Chargeback posted, Chargeback pre-notice, Reversal notice, Other
• Which message format will you use to exchange case evidence with merchants or external recovery vendors (REST JSON, SFTP batch, API payload)? Options: REST JSON, SFTP batch, API payload, Other
• Which BIN ranges or merchant categories have special remediation flows (for example high-volume merchants, corporate travel cards)?
• Which settlement ledger fields should be updated when remediation recovers funds (settlement ID, recovered amount, posting date)?
• Which case types require manual review before automated remediation steps (suspected false positive, identity theft)? Options: Suspected false positive, Identity theft, High-dollar cases, None

• Master Services Agreement (MSA)
• Statement of Work (SOW)
• Subscription Agreement
• Order Form (Pricing & Fees)
• Service Level Agreement (SLA)
• Data Processing Agreement (DPA)
• Data Access Authorization
• Go/No-Go Acceptance Criteria
• Security & Compliance Addendum (SOC 2 / Data Residency)
• Change Order Agreement
• Confidentiality & Non-Disclosure Agreement (NDA)

• Go-live Health Check
• First Measurement Review
• Acceptance Gate Review
• Quarterly Success Review
• Annual Realization and Risk Review

• Produce a quarterly metrics packet showing trend lines, segment-level drivers, and SLA adherence.
• Execute the remediation plan for any failed criteria with defined milestones and verification steps.
• Perform the incumbent system wind-down tasks or final-read-only configuration and archive migrated data according to the agreed schedule.
• Trend review of key outcome metrics
• Confirm whether fraud loss ($ per month) and false-positive rate (%) are stable or improving toward targets recorded in Solution Evaluation and identify any regressions.
• Clear the top operational blockers and commit concrete verification dates for closure.
• Prioritize submitted enhancements that materially affect the named metrics and place them in the operational roadmap for the quarter.
• Re-confirm scope and acceptance criteria
• Close or reassign the top 5 operational tickets and document verification steps to confirm closure.
• Document enhancement requests with expected metric impact and schedule items into the quarterly work plan.
• Annual performance vs targets
• Validate whether year-over-year fraud loss reduction (%) and per-case dispute processing cost ($) meet the realization expectations recorded in Solution Evaluation.
• Confirm compliance posture for Reg E/Reg Z timelines with supporting artifacts and identify any remediation required.
• Set the annually recurring model retraining and validation cadence to mitigate model drift risk.
• Run a full compliance audit package for Reg E and assemble required artifacts for internal records.
• Schedule the annual model retraining and validation windows and document expected data inputs.
• Produce an operational risk register with mitigation owners and review cadence for the coming year.
• Confirm the deployment completed and all critical integrations are functionally live.
• Identify and document the top 3 go-live blockers with remediation actions and target dates.
• Verify initial operator onboarding and confirm any immediate training gaps to be closed in the next 14 days.
• Publish a deployment health summary with API connectivity and test transaction results.
• Provide the requested system logs and sampled test extracts for unresolved integration errors.
• Open remediation tickets for each blocker with target resolution dates and expected verification steps.
• Present first measurement dataset
• Determine whether fraud loss ($ per month) and false-positive rate (%) are trending toward the targets recorded in Solution Evaluation and identify the primary drivers of variance.
• Agree a prioritized corrective action plan with timelines sufficient to reach the acceptance gate.
• Confirm the required evidence package and data exports needed at the acceptance gate.
• Produce a sliced dataset showing fraud losses and false-positive rates by customer segment, channel, and scoring band.
• Implement agreed scoring threshold adjustments and document expected impact and verification tests.
• Schedule model retraining run and provide timeline for validation in the next measurement window.
• Restate acceptance criteria and targets
• Produce a documented pass/fail result for each numeric acceptance criterion as recorded in Solution Evaluation and capture the acceptance decision.
• If any criterion is unmet, agree remediation tasks and timelines sufficient to close the gap before a secondary acceptance checkpoint.
• Confirm the incumbent system is either decommissioned or formally retained-read-only, its data archived or migrated complete, and the team's fallback habit is closed.
• Publish the acceptance decision record and attach the evidence package used for the evaluation.
• Deployment and integration validation
• Operational issues and ticket burn-down
• Compliance and audit artifacts
• Present outcome data against each criterion
• Diagnose root causes for any metric gaps
• Agree corrective actions and timeline
• Document pass or fail per criterion and acceptance decision
• Model risk and data quality assessment
• Early adoption signals and usage patterns
• Enhancement requests and impact assessment
• Open issues and blockers
• Quarter action plan and verification criteria
• Year ahead operational improvements
• Confirm path and timeline to acceptance gate
• Remediation plan and incumbent system wind-down
• Agree immediate remediation actions