Measure Fraud Alert Effectiveness
“Did the SMS verification stop fraud — or would the transaction have failed anyway?”
Book a demo and get a free trial of the full platform: research agent, fine-tune capabilities, and forward-deployed engineer support.
By submitting, you accept the Terms and Privacy Policy.
Loved by data scientists, ML engineers & CXOs at
A real-world example
Did the SMS verification stop fraud — or would the transaction have failed anyway?
Fraud team claims "SMS verification prevented $14M in fraud last quarter." But how much was actually prevented vs. transactions that would have declined anyway? Measuring true causal impact lets you remove unnecessary friction for 70% of SMS sends while maintaining fraud prevention.
How KumoRFM solves this
Graph-powered fraud intelligence
Kumo’s ASSUMING clause runs a counterfactual prediction: "What would happen if SMS was sent?" vs. "What would happen without SMS?" By comparing both predictions, you measure per-cardholder uplift. CH002 shows −0.35 uplift — SMS actually prevents fraud here. CH001 shows only +0.02 — SMS is just friction.
From data to predictions
See the full pipeline in action
Connect your tables, write a PQL query, and get predictions with built-in explainability — all in minutes, not months.
Your data
The relational tables Kumo learns from
Cardholders
| cardholder_id | card_type | avg_monthly_spend | sms_eligible |
|---|---|---|---|
| CH001 | Platinum | 4,200 | 1 |
| CH002 | Gold | 1,800 | 1 |
| CH003 | Silver | 650 | 0 |
Transactions
| txn_id | cardholder_id | amount | status | timestamp |
|---|---|---|---|---|
| T001 | CH001 | 245.00 | approved | 2025-01-10 |
| T002 | CH002 | 6,800 | declined | 2025-01-18 |
SMS Verifications
| sms_id | cardholder_id | trigger_reason | timestamp |
|---|---|---|---|
| S001 | CH001 | high_amount | 2025-01-10 |
| S002 | CH002 | new_merchant | 2025-01-18 |
Write your PQL query
Describe what to predict in 2-3 lines — Kumo handles the rest
PREDICT COUNT(TRANSACTIONS.* WHERE TRANSACTIONS.STATUS = 'declined', 1, 7, days) > 0 FOR EACH CARDHOLDERS.CARDHOLDER_ID WHERE CARDHOLDERS.SMS_ELIGIBLE = 1 ASSUMING COUNT(SMS_VERIFICATIONS.*, 0, 1, days) > 0
Prediction output
Every entity gets a score, updated continuously
| CARDHOLDER_ID | True_PROB (with ASSUMING) | True_PROB (without) | Uplift |
|---|---|---|---|
| CH001 | 0.12 | 0.10 | +0.02 |
| CH002 | 0.08 | 0.43 | -0.35 |
| CH003 | 0.15 | 0.14 | +0.01 |
Understand why
Every prediction includes feature attributions — no black boxes
Cardholder CH002
Predicted: -0.35 uplift (SMS prevents fraud)
Top contributing features
Transaction amount
$6,800
40% attribution
SMS trigger reason
new_merchant
23% attribution
Card type
Gold
17% attribution
Avg monthly spend
$1,800
13% attribution
SMS eligible
Yes
7% attribution
Feature attributions are computed automatically for every prediction. No separate tooling required. Learn more about Kumo explainability
PQL Documentation
Learn the Predictive Query Language — SQL-like syntax for defining any prediction task in 2-3 lines.
Python SDK
Integrate Kumo predictions into your pipelines. Train, evaluate, and deploy models programmatically.
Explainability Docs
Understand feature attributions, model evaluation metrics, and how to build trust with stakeholders.
Bottom line: Identify which 30% of SMS verifications actually prevent fraud. Remove unnecessary friction for the other 70%. Reduce false declines by 25% while maintaining fraud prevention.
Related scenarios
Explore more fraud predictions
Topics covered
