LIVE INTERACTIVE ARTIFACT

Ad Performance & Forecast Showcase

This interactive report showcases the deterministic analytics contract produced by PaceCast. Every prediction interval, KPI metric, backtest receipt, and anomaly diagnosis below is computed directly from actual daily ad records.

Portfolio Key Performance Indicators

Aggregated across Meta Ads and Google Ads campaigns with 30-day projected pacing.

Planned Spend (30d)
$36,285
GOOGLE + META across campaigns
Forecast Conversions
713
90% Range: [116 – 1,375]
Blended CPA
$50.88
Spend / forecasted conversions
Model vs. Baseline
+12.6%
WAPE reduction vs. Seasonal Naive
Calibrated 90% portfolio interval: 9 out of 10 periods land inside this range.

Calibrated 90% Prediction Intervals

Switch between portfolio totals and individual campaigns to inspect empirical prediction boundaries.

30-Day Forecast Horizon

Conversions & 90% Conformal Interval

Actual Conversions
Forecast (Point Estimate)
90% Prediction Interval
Cutoff: Dec 30, 2024
Loading chart visualization...
Showing last 60 days of actuals + 30-day forecast. Bands represent calibrated 90% conformal prediction intervals.Calibrated at account portfolio level (90% coverage).

Model Backtest Scoreboard

Every analysis runs cross-validation against baseline seasonal models to prove forecast accuracy.

Verifiable Receipts

Backtest Scoreboard

Every run executes rolling-origin cross-validation before printing predictions.

Model Delta
+12.6% vs. baseline
Evaluation StrategyWAPE (Daily)WAPE (Weekly)90% CoverageVerdict
Seasonal Naive Baseline38.3%30.6%—Reference
PaceCast Model33.5%26.8%90.0% (Target: 90%)Calibrated
Calculated via rolling-origin cross-validation against baseline.

Statistical Quality Guard

Low-volume or irregular campaigns are quarantined to prevent skewed predictions.

Data Integrity & Quality Guard

Excluded Campaigns

PaceCast filters campaigns that violate minimum statistical criteria to prevent noisy or uncalibrated forecasts.

2 Campaigns Excluded
CampaignReason for ExclusionPipeline Action
Meta Meta C104 Short
meta_meta_c104_short
Insufficient history: 13 days span (minimum 28 days required)Omitted from model fitting
Google Goog 804 Short
google_goog_804_short
Insufficient history: 8 days span (minimum 28 days required)Omitted from model fitting
Why exclusions matter: Campaigns with fewer than 28 days of history or zero recent conversions are too new to forecast with calibrated bounds — they will be included once attribution matures next cycle.

Reproducible Local Execution

Ingest Meta and Google Ads CSV files through reproducible analytical pipelines with zero cloud data retention.

bash — pacecast-pipeline
$python -m forecast --meta meta_ads.csv --google google_ads.csv --horizon 30
[1/4] Parsing inputs...
✓ Meta Ads: 90 days loaded(3 active campaigns, 1 excluded: history < 28d)
✓ Google Ads: 90 days loaded (3 active campaigns, 1 excluded: 0 spend in 14d)
ℹ Attribution maturity trim: Dropped last 7 days of Meta data to avoid immature conversion lag.
[2/4] Rolling-origin backtest (5 folds, horizon=30d)...
Fold 1 [origin 2024-10-01]: Naive WAPE 39.4% | Model WAPE 34.1% | 90% Cov: 91.2% Fold 2 [origin 2024-10-20]: Naive WAPE 38.1% | Model WAPE 33.6% | 90% Cov: 89.8% Fold 3 [origin 2024-11-09]: Naive WAPE 37.8% | Model WAPE 32.9% | 90% Cov: 90.4% Fold 4 [origin 2024-11-29]: Naive WAPE 38.7% | Model WAPE 33.8% | 90% Cov: 88.9% Fold 5 [origin 2024-12-19]: Naive WAPE 37.6% | Model WAPE 33.1% | 90% Cov: 90.1%
[3/4] Fitting forecast model (horizon=30d)...
✓ Conformal intervals calibrated at α=0.10 (90% nominal coverage achieved).
[4/4] Writing report receipts...
→ results/forecast.csv [30 rows, per-campaign spend & pred [lo, hi]]
→ results/backtest.txt [pooled & campaign cross-validation scorecards]
→ results/plot.png [publication-ready interval band chart]
→ results/report.json [canonical schema v1.0 data contract]
✔ Complete in 5.84s. Model outperformed seasonal naive baseline by +12.6%.

Forecast your own advertising campaigns

Connect your Google Ads account or drop in Meta and Google Ads daily exports. Get deterministic anomaly detection and 30-day forecast certainty today.

50 free monthly AI credits·Raw CSVs processed in memory