Tableau Desktop and Web Authoring Help

What makes a viz a good candidate for Explain Data

Explain Data works best on visualizations that require deeper exploration and analysis, rather than infographic-style, descriptive vizzes that communicate summarized data.

• Row-level data is necessary for Explain Data to create models of your data and generate explanations. Vizzes with underlying, row-level data, where relationships might exist in unvisualized fields are good candidates for running Explain Data.
• Vizzes based on pre-aggregated data without access to row-level data are not ideal for the statistical analysis performed by Explain Data.

What data works best for Explain Data

When you are using Explain Data in a worksheet, remember that Explain Data works with:

• Single marks only—Explain Data analyzes single marks. Multiple mark analysis is not supported.

• Aggregated data—The view must contain one or more measures that are aggregated using SUM, AVG, COUNT, or COUNTD. At least one dimension must also be present in the view.

• Single data sources only—The data must be drawn from a single, primary data source. Explain Data does not work with blended or cube data sources.

When preparing a data source for a workbook, keep the following considerations in mind if you plan to use Explain Data during analysis.

• Use a data source with underlying data that is sufficiently wide. An ideal data set has at least 10-20 columns in addition to one (or more) aggregated measures to be explained.
• Give columns (fields) names that are easy to understand.
• Eliminate redundant columns and data prep artifacts. For more information, see Change fields used for statistical analysis.
• Don't discard unvisualized columns in the data source. Explain Data considers fields in the underlying data when is analyzes a mark.
• Low cardinality dimensions work better. The explanation of a categorical dimension is easier to interpret if its cardinality is not too high (< 20 categories). Dimensions with more than 500 unique values will not be considered for analysis.
• Don't pre-aggregate data as a general rule. But if the data source is massive, consider pre-aggregating the data to an appropriate level of detail.
• Use extracts over live data sources. Extracts run faster than live data sources. With live data sources, the process of creating explanations can create many queries (roughly one query per each candidate explanation), which can result in explanations taking longer to be generated.

Situations where Explain Data is not available

Sometimes Explain Data will not be available for a selected mark, depending on the characteristics of the data source or the view. If Explain Data cannot analyze the selected mark, the Explain Data icon and context menu command will not be available.

Explain Data can't be run in views that use:	Map coordinate filters Blended data sources Data sources with parameters Data sources that don't support COUNTD or COUNT(DISTINCT ...) syntax, such as Access. Filters on aggregate measures Analytics objects (any of the items listed in the Analytics pane(Link opens in a new window)) Disaggregated measures
Explain Data can't be run if you select:	Multiple marks Axis Legend Grand total Trend line or reference line A mark in a view that contains a very low number of marks
Explain Data can't be run when the measure to be used for an explanation:	Isn't aggregated using SUM, AVG, COUNT, COUNTD Is a table calculation Is used in measure values
Explain Data can't offer explanations for a dimension when it is:	A calculated field A parameter Used in Measure Names and Measure Values A field with more than 500 unique values. Dimensions with more than 500 unique values will not be considered for analysis.