ZeroGrad
Capabilities · Classification

Fraud Detection

Healthy
In-context predictionEndpoint fraud-detectionProduction v7Challenger v8

Real-time production health

Known immediately — no outcomes needed.

Last 24 hours
Predictions
1.82M
Availability
99.997%
P50 latency
19ms
P95 latency
43ms
Error rate
0.01%

Predictions / hour

47K75K104K00:0004:0008:0012:0016:0020:00

Latency

0ms54ms108msSLO P95 < 100ms00:0004:0008:0012:0016:0020:00
P50P95

Inputs, context & predictions

Schema
Healthy
Feature drift
Low
PSI 0.04 max (merchant_category)
Missing values
Normal
OOD events
0.31%

Input drift vs reference (PSI)

0.1 · 0.25
merchant_category
0.04
country
0.03
amount
0.02
account_age_days
0.02
device_type
0.01
hour_of_day
0.01

Prediction distribution

today vs reference · no shift
0Predicted fraud probability1
Live Reference examples

Routing

In-context prediction
v7 · Managed A100%
Backend
Managed by Zero Grad
Context
pinned to v7
Fallback on timeout
Managed C · 0.02% of requests

Matured outcome quality

Performance metrics are computed only for predictions with observed outcomes. Chargebacks and confirmed fraud reviews take up to ~30 days to arrive, so recent predictions aren't scored yet.

Evaluation window
Aug 1 – Aug 31
predictions made in this period
Outcome coverage
72%
have an observed outcome
Median outcome delay
28 days
chargebacks and confirmed fraud reviews
Last outcome received
Sep 30, 23:41
via outcomes API

Business impact

Aug 1 – Aug 31 · 72% coverage
Confirmed fraud caught
$4.18M
Confirmed false positives
8,750
Estimated loss avoided
$3.34M
after review cost & friction

Uses your configured value per case. Grows as the remaining 28% of August outcomes arrive.

Model quality

Aug 1 – Aug 31
AUC
.947
Precision
91%
Recall
79%
False positive rate
0.7%
Calibration
Healthy
ECE 0.011

AUC — matured outcomes

By prediction week. Recent cohorts are provisional until their outcomes arrive.

.937.947.958Jul 6Jul 20Aug 3Aug 17Aug 31Sep 14
AUC · maturedProvisional (< 60% coverage)
Outcome coverage by cohorttoday →

Outcomes

Ground truth flows back by prediction_id. It improves evaluation and routing — never model weights.

Send outcomes

Every prediction returns a prediction_id. Send the real outcome whenever you learn it.

# Later, when the real outcome is known
zg.capability("fraud-detection").send_outcomes([
    {"prediction_id": "prd_01J9X4K2M8", "outcome": "fraud",
     "observed_at": "2026-09-30T23:41:00Z"},
])

Outcomes never update model weights — Zero Grad does not train a customer-specific model. They are used to measure real production quality, compare foundation models and context-selection strategies, improve routing, and decide whether a challenger beats the current production configuration.

What Zero Grad learned from your outcomes

Simulated

No training doesn't mean no learning.

  1. Context strategy: recent-weighted examples beat uniform sampling by +.006 AUC on matured outcomes — now the default for this capability.
  2. Routing: TabFM wins on transactions over $1,000. A routing rule for that slice is queued for shadow evaluation.
  3. Calibration: matured outcomes moved the 1% FPR operating point from 0.71 to 0.69.
  4. v8 (new foundation model release) beats v7 on the August cohort: +.009 AUC at 72% outcome coverage.

What adapts: which foundation model serves which slice, how in-context examples are selected, and calibration. What doesn't: model weights.