Time-Series Generator
Build synthetic time-series rows with a fixed start, interval, trend, seasonality, noise, gaps, and outliers.
Also known as: timeseries · telemetry · signal
seeded · synthetic data
Output
About this tool, tips & examples
What it does
The Time-Series Generator builds realistic timestamped metrics with the components real signals have: a baseline level, a trend per step, seasonality (amplitude and period), Gaussian noise, plus deliberate missing values and outliers at rates you control. Up to 10,000 rows from a fixed start datetime at your chosen interval — clock-independent and exactly reproducible by seed.
Common use cases
- Chart prototypes — line charts that look like production metrics, seasonality and all.
- Monitoring demos — dashboards with plausible daily/weekly cycles and the occasional spike.
- Forecasting experiments — data with known trend and seasonality is the honest way to test a forecasting model — you know the right answer.
- Pipeline hardening — the missing-value and outlier injectors exercise gap handling and anomaly detection with labeled ground truth.
Settings
- Start datetime / Interval / Rows — the timeline: fixed start, regular spacing, up to 10,000 points.
- Baseline / Trend per step — level and drift of the signal.
- Seasonality amplitude / period — a repeating cycle (set the period to your interval’s day-length for daily patterns).
- Noise standard deviation — realistic jitter.
- Missing-value rate / Outlier rate — 0 to 1, controlled data imperfections.
- Seed — identical seed + settings = identical series (CSV, JSON, or NDJSON).
Privacy note
Series are generated locally in your browser and never uploaded. Gaps and spikes are simulated, not real outages — label them as synthetic wherever you show them.
FAQ
How do I fake a week of hourly metrics? Hourly interval, 168 rows, seasonality period 24 for the daily cycle, mild noise — instant realistic ops data.
Why is known-composition data better for forecasting tests? Because you can score the model against the true trend and seasonality you injected, instead of guessing whether residuals are model error or data mystery.
Wandering instead of trending? Random Walk and Brownian Motion generate drift-style series without seasonal structure; the IoT Sensor tool wraps time-series output in device semantics.