ZeroGrad
Capabilities · Forecasting

Weekly Demand Forecast

Healthy
Zero-shot forecastingEndpoint weekly-demandProduction v4Challenger v5

Real-time production health

Known immediately — no outcomes needed.

Last 24 hours
Series forecast / week
12.5K
Availability
99.99%
P50 latency
140ms
P95 latency
380ms
Error rate
0.03%

Predictions / hour

31951971800:0004:0008:0012:0016:0020:00

Latency

0ms540ms1080msSLO P95 < 1000ms00:0004:0008:0012:0016:0020:00
P50P95

Inputs, context & predictions

Schema
Healthy
Late data
2 stores
sales feed delayed > 24h
Missing values
Normal
New series
214

Input drift vs reference (PSI)

0.1 · 0.25
promo_flag
0.07
price
0.05
store_id
0.02
week_of_year
0.01

Prediction distribution

today vs reference · no shift
0.5×Forecast ÷ trailing 4-week average1.5×
Live Reference examples

Routing

Zero-shot forecasting
v4 · Managed A100%
Backend
Managed by Zero Grad
Context
pinned to v4
Fallback on timeout
Managed C · 0.02% of requests

Matured outcome quality

Performance metrics are computed only for predictions with observed outcomes. Realized weekly sales take up to ~9 days to arrive, so recent predictions aren't scored yet.

Evaluation window
Aug 3 – Sep 20
predictions made in this period
Outcome coverage
100%
have an observed outcome
Median outcome delay
9 days
realized weekly sales
Last outcome received
Sep 29, 06:00
via outcomes API

Business impact

Aug 3 – Sep 20 · 100% coverage
Stockouts avoided (est.)
1,920
Excess inventory reduced
$96K
Planner hours saved
~60 / wk

Inventory effects estimated against the previous planning process over 8 backtest windows.

Model quality

Aug 3 – Sep 20
WAPE (realized)
13.1%
Bias
1.4%
Interval coverage
91%
Calibration
Healthy

WAPE — matured outcomes

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

12.7%12.8%12.9%Jul 6Jul 20Aug 3Aug 17Aug 31Sep 14
WAPE · 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("weekly-demand").send_outcomes([
    {"prediction_id": "pred_01J9X4S9VD", "outcome": {"units_sold": 418},
     "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. Routing: intermittent series sent to the statistical baseline cut their realized WAPE from 31% to 24% (v5).
  2. Covariates: promo_flag improves promo-week WAPE by 3.1 pts; price added no lift and was dropped from context.
  3. Late sales from 2 stores are excluded from evaluation until they arrive.

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