# Data Visualization Best Practices

Turning numbers into knowledge is to not only to separate noise from the data, but also to present it appropriately. Adaptive Discovery is here to make it easy, powerful and fast!

**Data Representation Methods:**

There are 4 different basic data representation methods, each requiring a different way to represent the data:

- Comparison
- Composition
- Distribution
- Relationship

However, which chart type is most appropriate? A screwdriver should never be used to hammer a nail and the same is true for data visualization, therefore, we should address a few questions before proceeding:

- How many variables should be present?
- How many data points do we have per variable?
- What is/are the categories that we want to break the data down by?
- If the category is time, what timeline do you want to display these values over?
- If the amount of data points is too large, do we need to sort & limit to display only a certain portion?

It is important to recognize that there are 2 main vehicles for data analysis: tables and charts, each focusing on different outcome:

**Tables:**

- Tables are usually used for few metrics/planning accounts for comparison and analysis.
- Tables are also used to look up a specific value in a metric/category intersection.
- Tables are also used when precise values are necessary to avoid confusion and for when a different unit of measure is needed.

A **Scorecard **should be used when comparing a few metrics across multiple units of measure including a variance in values & percents.

A **Crosstab** behaves a lot like a pivot table in excel and serves as a report where accounts/metrics are broken down by either time or a dimension with precise values for further drill down.

**Charts:**

Charts serve as a more visual representation of data in Discovery and are generally easier to read and draw conclusions.

**Column Chart**:

- Most used and easiest to create.
- Could be used for comparison within different categories (dimensions or time periods).
- It is best to keep the number of categories to no more than 5.
- Try and use column charts if there is a reasonably low number of data points and if every data point has a clear value.
- Histograms also fall into this category and can also be set up in Adaptive if necessary.
- A slightly more advanced use of Column charts are the stacked columns and is typically used to show composition.
- Simplify your column chart if possible after initial creation. less is more.

**Bar Charts**:

- Similar to column charts and can be substituted under certain conditions.
- Long category names could benefit from this.
- If the number of categories is more than say 5 but no more than 15, it is a better option to use bar charts over column charts.
- Bar charts can also be stacked for a more compact view and when there is more of the emphasis on the composition than the comparison. Both need to be present still.

**Line Charts**:

- Use a line chart when you are working with a continuous data set.
- A much better alternative to a column if a chart needs to be small to fit on a dashboard.
- Needs to represent the flow of data/trend with elements of single-point comparison included.
- Use line when the number of data points is high or higher than would be appropriate for a column chart.

**Area Charts**:

- Useful when showing flow & trend of an accumulated series or account.
- Avoid using area charts for fluctuating amounts.

**Pie/Donut Charts**:

- Pies/donuts are often misused so it’s always a good idea to know the basics behind it.
- Use pie charts to represent a metric or dimension that has a small number of categories.
- If you need to point out a clear winner in a category, use this chart.
- Avoid using a pie chart if every category for comparison are similar or the same in size/portion.

**Gauge Charts**:

- Used for displaying a KPI or one very important metric.
- This adds to the simplicity of a dashboard to be easier to interpret.
- These fit well when you wish to display overall performance of a metric in relation to a goal.
- Single key metric that can be quickly scanned and understood.
- KPI charts are also a good example of this. Both should enable you to start asking questions as quickly as possible.

**Waterfall Charts**:

- Allows you to plot sequential change in a metric given different categories.
- Provides storytelling via a Discovery chart of a metric's performance.
- Has a starting point, what positive/negative values affected the metric, and its final value.
- The category could be anything, time or dimension.
- Begging and ending headcount plot is a perfect example of a waterfall chart.

### **Overall advice**:

- Use horizontal axis for time if displaying a metric over time.
- Less is more; try to simplify your charts after creation by removing redundant lines, colors and/or texts.
- Use the sort & filter function on charts.
- Use a legend whenever possible.
- If data labels overlap, you have too many categories for a bar chart.
- If a chart turns out to be more complex than originally thought, think about splitting the chart into multiple charts on a single dashboard or spread them across different dashboards in Discovery.