ZeroGrad
Capabilities · Recommendations

Product Recommendations

Healthy
In-context predictionEndpoint product-recommendationsProduction v12Challenger v13

Experiment: Product Recommendations v12 vs v13

Split live traffic between versions and compare them on a business metric, with guardrails.

SimulatedRUNNING · DAY 4 OF 14

Results

Primary metric: Revenue per session

v12v13
Traffic80%20%
Requests2.5M625K
CTR8.7%9.1%
Revenue per session$1.84$1.93
P95 latency58ms61ms

Simulated results. Directional only — the platform reports confidence intervals once the minimum sample size is reached. Quality and business rows only count predictions whose outcomes have matured (median delay 1.6 days).

Guardrails

✓P95 latency < 100ms
✓Error rate < 0.1%
✓Catalog coverage > 60%

Revenue per session by day

day 4 of 14
$1.78$1.88$1.99d1d3d5d7d9d11d13
v12v13
Remaining days are not yet observed.