Date transformations and visualizations

Time Series Analysis in Tableau

Chris Hui

VP of Product, Tracked

What's seasonality?

Seasonality is when time series data experiences regular and predictable changes that recur every calendar year
Examples include tourism or fruiting seasons that have variable prices based off timing
Seasonal behaviour allows business to effectively plan around peaks and troughs to optimize their business

3 line charts depicting different price points for different avocado categories

Common methods to treat seasonality:

Moving averages
- Technique to smooth out short term fluctuations (peaks/troughs) in the data over a specific time window
- Used to filter out noise while preserving the underlying signal

A line chart showing the average sales compared against a moving average which helps to smooth the peaks in the data

Common methods to identify seasonality:

Seasonal boxplots (quarterly analysis)
- Segments the data quarterly to enable visualization of seasonal (quarterly) fluctuations
- Consistent volatility across quarters are an indicator of seasonal behavior

A box plot visualization showing varying interquartile ranges by quarter and year

Anomalous values (outliers), are values that deviate outside the normal distribution
Outliers can be considered to be:
- Any value outside +- 3 standard deviations away from the mean
- More than 1.5 x IQR below Q1 or more than 1.5 x IQR above Q3

A box plot showing the sales variance between seasons

(Standard deviation is a measure of how far any value is from the population mean)

The Z-score is the number of standard deviations a given data point lies above or below mean
Z-scores within +-3 means ~99.7% of the population values lies within this range
Subsequently, any Z-score outside the +-3 range can be considered an outlier

A normal distribution showing the range of Z-Scores from -3 to +3

Percentiles determine where a value stands relative to other values
- Median (50th Percentile)
- Upper & Lower Quartiles (75th / 25th percentiles)

A normal distribution visualizing split into below the 50th percentile and above the 50th percentile

Provides a flexible approach to outlier detection beyond the traditional Z-score methodology

A histogram showing the cut off threshold for whether a value is anomalous or not

Time Series Analysis in Tableau