Keep your current stack
Supply the agreed events and context through a mapped integration. SENTR processes and stores the evaluation data under the agreed data-handling terms; your existing controls keep making production decisions.
SENTR · SHADOW MODE
How shadow mode actually works
One of us has something to lose. It's not you.
For 50 days, SENTR runs alongside your existing stack — read-only, silent, parallel. You get a full intelligence report. We get the chance to earn your confidence. No commitment changes that math.
Read-only connection. Read-only beside your stack.
From first call to live intelligence
Three steps. No production commitment required to evaluate.
Live architecture map
Step 1
First contact & Architecture Session
Deep dive into your way of working.
We map your current fraud stack and risk architecture. No pitch deck — a working session. We define the metrics we will evaluate, confirm the scope for Shadow Mode, and build your 50-day plan before we connect a single line of code.
- Map your current fraud stack and risk architecture
- Define success metrics and detection benchmarks
- Build your personalised 50-day shadow mode roadmap
- Confirm integration requirements with your technical lead
- No commitment at this stage — walk away if it's not a fit
Parallel evaluation Read-only · zero production writes
Shadow day 12 / 50
Mirroring production traffic
-
TXN-99201CNP · €428 · NL→EE IP Approve Decline Velocity + geo mismatch — shadow would block -
TXN-99177Retry cluster · same device Approve Review Escalated in shadow — no customer impact -
TXN-99162Known good cohort Approve Approve Aligned -
TXN-99154Blocked list hit Decline Decline Aligned
Illustrative sample — not live data.
Step 2
50-day shadow mode
SENTR running next to you.
Securely connected to compatible mapped events from supported sources, SENTR runs silently alongside your existing stack — scoring and explaining in real time while the incumbent remains live. Every week you see decision differences, linked activity and review effort on your own data. After 50 days you have comparison evidence before any production change.
- Read-only for production enforcement — we do not write decisions into live flows
- Your existing platform keeps making every live fraud decision
- Weekly intelligence: disagreements, reviewed findings and modelled impact
- Decision evidence baseline building in the background for deployer assessment
- Walk away at Day 50 with the comparison — no production invoice while read-only
Live operations console Primary routing Hypercare
Production decisions flowing through SENTR · your stack unchanged downstream.
Illustrative console — not live data.
Step 3
Go-live
Operational integration — after proven results.
After the data proves the case, we move to full integration into your risk operations workflow. Your team hits the ground running with training and documentation. Our team provides ongoing support from day one.
- Full integration into your live risk operations workflow
- Team training and runbook documentation
- Dedicated integration engineer throughout
- Support and incident coverage agreed before go-live
- You decide the timeline — no pressure before you're ready
The read-only connection
Integration with a clear plan
We work with your technical contact on access, mapping and validation. Agree scope, responsibilities and the integration timeline during the Architecture Session—before the evaluation starts.
Read-only — no production write-back during evaluation
We observe only. We write nothing to your production environment. Your fraud platform keeps making every live decision throughout.
The 50-day parallel run
-
Day 0 — Integration live.
SENTR connected to your transaction stream. Your existing platform unchanged. Nothing in your live environment is touched.
-
Days 1–49 — SENTR runs. You watch.
SENTR scores the agreed event stream without enforcing decisions in production. Your existing queues keep running. Review comparison reports with us, investigate differences and record which outcomes are confirmed or still open.
-
Day 50 — The data is yours.
Full intelligence report delivered. Your fraud data, analyzed. Side-by-side comparison with your current platform's output. Then you decide.
- Agreements where SENTR and the incumbent aligned
- Disagreements for independent review — not automatic labels
- Findings supported by investigation or confirmed outcomes
- Modelled financial impact with stated assumptions
- Decision evidence for deployer assessment — not a legal certificate
What the Day 50 debrief covers
-
Side-by-side decision comparison
We show every event where SENTR and your incumbent disagreed — SENTR flag / incumbent pass, and the reverse. Those are disagreements for review, not automatic “missed fraud” or “false positive” labels. Independently reviewed findings, unresolved differences and modelled impact are reported separately.
-
Chargeback attribution analysis
For any chargebacks that landed during the 50-day window, we map them back to the detection decision at transaction time. You see how often your current model was right, wrong, and why.
-
Investigation effort audit
Where review time is captured, compare the effort to reach a supported decision. Separate observed handling time from estimates of capacity that clearer explanations or connected context could release.
-
Decision-evidence review
Inspect recorded attribution, human overrides and exportable evidence from the evaluation. Identify evidence gaps for your governance review; legal classification and compliance assessment remain separate.
-
Go/no-go framing
We state our view directly: whether the data supports going live with SENTR, and on what timeline. If the data doesn't make the case, we say so. The report is yours either way.
The Day 50 intelligence report
Exactly what you receive.
Fraud matched
Events both systems flagged. This shows decision alignment; confirmed fraud still requires a supported outcome.
Decision disagreements to review
Events where SENTR and your incumbent differ, queued for independent review and outcome tracking.
Potential false positives
Candidate cases where legitimate customers may have been blocked, pending review confirmation.
Modelled chargeback impact
Modelled impact from reviewed findings, with assumptions shown separately from confirmed outcomes.
Decision explainability
Per-decision reasoning log. Evidence for deployer assessment — not automatic legal compliance.
This report is the evidence you use to decide whether to proceed—or walk away. The qualified read-only evaluation is free; paid production is a separate agreement.
Integration architecture
We connect to where your fraud decisions are made — not to your payment processor.
SENTR integrates read-only to the risk data you can provide: internal event streams, mapped provider outputs, and chargeback data layers. Your existing payment and fraud controls can stay in place while comparison runs.
Where decisions are made
Transaction event stream
We connect via read-only API to your live transaction data — your processing core, your event log, or your internal reporting layer. Your stack doesn't change.
Running alongside your current tool
Fraud & risk engine output
SENTR observes your existing fraud decisions and runs its own logic in parallel. At Day 50, you see both outputs side by side. No replacement required before you decide.
Full-loop visibility
Chargeback data layer
We ingest your dispute and chargeback data to map the full fraud-to-resolution picture. Shadow mode shows you both the fraud your platform caught and the losses it didn't.
Questions about integration for your specific stack? Contact our integration team — hello@sentr.io
Decision evidence you can inspect
Building your audit trail in shadow mode.
Sample explainability output
DECLINE Model confidence 94.2%
- Evaluation mode
- Shadow · read-only
- Risk signals
- Velocity · Geo mismatch · CNP · MCC risk
Rationale (human-readable)
- Velocity anomaly: three successful authorisations from the same device fingerprint within four minutes — exceeds your configured burst threshold for this segment.
- Billing country (DE) inconsistent with IP geolocation (high-risk jurisdiction).
- Card-not-present with elevated merchant category risk score.
Human-readable explainability log
SENTR stores attribution for evaluated events and renders readable explanations for relevant rule and anomaly findings. Human overrides keep their written reason. During Shadow Mode, inspect this evidence while your existing controls remain live—without treating the evaluation as a legal compliance assessment.
- Per-decision explainability — why we approved or declined
- Exportable evidence when the deployment supports it
- Supports building an audit trail baseline for deployer assessment
- Explainability evidence to support your legal and compliance review
SENTR.Tower early access
SENTR.Tower’s free 14-day guided start
Lean teams on SENTR.Tower still need a clear path — without a full SENTR.Citadel operator programme. After guided fit, the initial cohort can begin with a free 14-day guided start: setup, presets, alerts, explanations, review/override and feedback. Production price is disclosed before that work begins. It is not freemium self-serve and not a substitute for long-lag fraud proof. SENTR.Citadel’s qualified 50-day Shadow Mode remains the deep read-only comparison.
Check SENTR.Tower fit → Explore SENTR.Tower →Bring us the events you already run.
Architecture Session maps your stack and whether SENTR.Citadel Shadow Mode or SENTR.Tower’s guided start is the honest next step — or whether it is not a fit yet.
Read-only evaluation. Production is a separate decision. No pitch deck required to start the conversation.