Good Practices for Data Visualization
Below you can find samples of Guidelines and Good Practices for Data Visualization (mostly with Tableau), which I used recently.
Some of this samples are Tableau-specific, but others (may be with modifications) can be reused for other Data Visualization Platform and tools. I will appreciate feedback, comments and suggestions.
Naming Convention for Tableau Objects
Use CamelCase Identifiers: Capitalize the 1st letter of each concatenated word
Use Suffix for Identifiers with preceding underscore to indicate the type (example: _tw for workbooks).
Workbook Sizing Guidelines
Use Less than 5 Charts per Dashboard, Minimize the number of Visible TABs/Worksheets
Move Calculations and Functions from Workbook to the Data.
Use less than 5000 Data-points per Chart/Dashboard to enable Client-side rendering.
To enable Shared Sessions, don't use filters and interactivity if it is not needed.
Guidelines for Colors, Fonts, Sizes
To express desirable/undesirable points, use green for good, red for bad, yellow for warning.
When you are not describing "Good-Bad situation" (thanks to feedback of visitor under alias "SF") , try to use pastel, neutral and blind colors, e.g. similar to "Color Blind 10" Palette from Tableau.
Use "web-safe" fonts, to approximate what users can see from Tableau Server.
Use either auto-resize or standard (target smaller screen) sizes for Dashboards
Data and Data Connections used with Tableau
Try to avoid pulling more than 15000 rows for Live Data Connections.
For Data Extract-based connections 10M rows is the recommended maximum.
For widely distributed Workbooks use of Application IDs instead of Personal Credentials.
Job failure due expired credentials leads to suspension from Schedule, so try to keep embedded credentials up to date
Tableau Data Extracts (TDE)
If Refresh of TDE takes more than 2 hours, consider to redesign it.
Reuse and share TDEs and Data Sources as much as possible.
Use of Incremental Data Refresh instead of Full Refresh when possible.
Designate Unique ID for each row when Incremental Data Refresh is used.
Try to use free Tableau Data Exract API instead of licensed Tableau Server to create Data Extracts
Scheduling of Background Tasks with Tableau
Serial Schedules is recommended; avoid the usage of hourly Schedules.
Avoid scheduling during peak hours (8am-6pm), consider weekly instead of daily schedules.
Optimize Schedule Size, group tasks related to the same project into one Schedule, if total tasks execution exceeds 8 hours, split Schedule on a few with similar Name but preferably with different starting time.
Maximize the usage of Monthly and Weekly Schedules (as oppose to Daily Schedules) and usage of weekends and nights.
Guidelines for using Charts
Use Bars to compare across categories, use Colors with Stacked or Side-by-Side Bars for deeper Analysis
Use Line for Viewing Trends over time, consider Area Charts for Multi-lines
Minimize the usage of Pie Charts; when appropriate - use it for showing proportions. It is recommended to limit pie wedges to six.
Use Map to show geocoded data, consider use maps as interactive filters
Use Scatter to analyze outliers, clusters and construct regressions
The following two points contradict each other. Also, the green, red, yellow paradigm will be lost on color blind viewers.
ReplyDeleteTo express desirable/undesirable points, use green for good, red for bad, yellow for warning.
Try to Use pastel, neutral and blind colors, e.g. similar to “Color Blind 10″ Palette from Tableau.