Time Series

Use me when time is the star of your story. I'm specifically built for data that changes over minutes, hours, days, months, or years. Watch sales grow, traffic spike, or temperatures fluctuate. I'll help you spot trends (are we going up?), seasonality (summer peaks?), and anomalies (what happened in March?). Your time-traveling companion for any data with a clock attached.

Overview

A time series plot displays data points at successive time intervals, revealing patterns, trends, and changes over time. It connects observations in chronological order, making it ideal for tracking metrics and identifying temporal patterns like seasonality, trends, and cycles.

Best used for:

• Tracking metrics over time (daily, monthly, yearly)
• Identifying trends and seasonal patterns
• Forecasting future values based on historical data
• Comparing multiple time-based metrics
• Monitoring KPIs and performance indicators
• Analyzing cyclical behavior and anomalies

Common Use Cases

Business Analytics

• Sales revenue over time
• Website traffic trends
• Conversion rates by month
• Customer acquisition growth
• Inventory levels tracking
• Stock prices and financial metrics

Operations & Performance

• System performance monitoring
• Server response times
• Error rates over time
• Production output trends
• Quality metrics tracking
• Resource utilization

Scientific & Research

• Temperature and climate data
• Population growth
• Experimental measurements over time
• Sensor readings and IoT data
• Medical vitals monitoring

Options

X-Axis

Required - Select the time/date column and optional time interval.

This determines what appears on the horizontal axis, typically a date, datetime, or numerical column representing time.

Time Column

Select the column containing time values.

Accepts: DATES, DATETIME, NUMERICAL columns

Time Interval (Bucketing)

Optional - Group data into time periods.

Options:

• Day - Aggregate by day
• Week - Aggregate by week
• Month - Aggregate by month
• Quarter - Aggregate by quarter
• Half Year - Aggregate by 6-month periods
• Year - Aggregate by year

Useful for reducing granularity and showing broader trends.

Y-Axis

Required - Select one or more numerical columns to plot over time.

You can add multiple Y-axis columns to compare different metrics on the same time scale. Each series will be shown in a different color.

Value Column

Select the numerical column to track over time.

Lower Bound (Optional)

Define a lower confidence bound or range.

Creates a shaded area between the main value and the lower bound, useful for showing confidence intervals or ranges.

Upper Bound (Optional)

Define an upper confidence bound or range.

Creates a shaded area between the main value and the upper bound.

Reference Line/Function (Optional)

Optional - Add reference lines or mathematical functions to the plot.

Enter mathematical expressions or constants to overlay reference lines or functions.

Examples:

• 100 - Horizontal line at y=100
• x**2 - Parabolic function
• 2*x + 5 - Linear function
• np.sin(x) - Sine wave

Use 'x' to represent the x-axis values. Supports NumPy functions (np.sin, np.cos, np.log, etc.).

Settings

Hide Empty Values

Optional - Exclude time periods with no data.

Removes gaps in the timeline where no data exists, creating a continuous line.

Show Trendline

Optional - Overlay a trend line showing overall direction.

Adds a linear regression line to show the general trend direction (increasing, decreasing, or flat).

Use Logarithmic Scale For Y Axis

Optional - Apply logarithmic scaling to Y-axis.

Useful when data spans several orders of magnitude or when you want to emphasize relative changes rather than absolute changes.

Show Gridlines

Optional - Display background grid lines.

Adds horizontal and vertical grid lines to make it easier to read precise values.

Tips for Effective Time Series Plots

1. Choose Appropriate Time Intervals:

   • Match interval to your analysis needs
   • Daily for short-term trends
   • Monthly/quarterly for seasonal patterns
   • Yearly for long-term trends

2. Handle Missing Data:

   • Decide whether to show gaps or interpolate
   • Use "Hide Empty Values" for discontinuous data
   • Consider forward-fill or interpolation for continuous processes

3. Compare Multiple Series:

   • Use consistent time periods across series
   • Consider normalizing if scales differ greatly
   • Use different colors and labels clearly

4. Add Context with Reference Lines:

   • Show targets or thresholds
   • Mark significant events
   • Display moving averages or benchmarks

5. Use Confidence Bounds:

   • Add upper/lower bounds for uncertainty
   • Show forecast confidence intervals
   • Visualize ranges or tolerances

6. Scale Appropriately:

   • Use log scale for exponential growth
   • Start Y-axis at zero for magnitude comparison
   • Adjust Y-axis range to emphasize important changes

Understanding Time Series Patterns

Trend

Long-term increase or decrease in the data. Use trendlines to visualize.

Seasonality

Regular patterns that repeat at fixed intervals (daily, weekly, monthly, yearly).

Cyclical Patterns

Fluctuations without fixed periods, often related to economic or business cycles.

Irregular/Random

Unpredictable variations that don't follow a pattern.

Example Scenarios

Sales Trend with Forecast

Multiple Metrics Comparison

Performance with Confidence Bounds

Troubleshooting

Issue: Trend is hard to see due to noise

• Solution: Apply moving average smoothing, use LOWESS/LOESS smoothing, or decompose series to isolate trend component.

Issue: Seasonality obscures other patterns

• Solution: Deseasonalize the data first, show decomposition plot separately, or use seasonal adjustment methods (e.g., X-13ARIMA-SEATS).

Issue: Multiple series have different scales

• Solution: Use dual Y-axes (one left, one right), normalize/standardize series to same scale, or show as percentage change from baseline.

Issue: Forecast seems unrealistic

• Solution: Check forecast model assumptions, verify training data quality, shorten forecast horizon, or show confidence intervals to communicate uncertainty.

Issue: Gaps in time series

• Solution: Interpolate missing values, forward/backward fill, explicitly show gaps, or aggregate to coarser time intervals that have complete data.

Issue: Too many data points make chart slow

• Solution: Aggregate to coarser intervals (hourly to daily), downsample data, use WebGL rendering, or show subset with zoom functionality.

Issue: Can't identify change points or anomalies

• Solution: Add reference lines at key dates, use anomaly detection algorithms, color-code anomalous points, or overlay events/annotations.

Issue: Autocorrelation is not visible

• Solution: Create ACF/PACF plots separately, add lagged versions of series, or use specialized time series decomposition tools.

Issue: Comparing multiple periods is difficult

• Solution: Overlay previous period with transparency, show percentage change vs absolute values, or use small multiples for year-over-year comparison.