You are on page 1of 3

Performance Optimization

Slow Workbook?
S E R V E R A N D DATA S O U R C E S

Find the workbook(s) with the


Test the workbook on Server and Desktop
slowest load times.
to see if one is noticeably slower.
H E R E ' S H OW

Slower on Desktop Slow on both


or Server N/A Slower on Server

?
Tools that can help.
Check system requirements Slow on What type of
TABLEAU SERVER
and try it on faster machines. all machines data connection?
MANAGEMENT ADD-ON

Faster on Extract Shared (Tableau Server)


Live
other machines

Upgrade your Check that the extract is optimized: ?


computer • Hide unused fields Are all calcs defined in the
• Aggregate to visible dimensions
? • Roll up dates to the level needed
shared data source?
Need real-time • Filter unneeded values before extracting
updates? CLICK FOR MORE Yes No

Yes No

? Create a local copy SOLUTIONS:


Does the data source to troubleshoot, • P ush all calcs to the
Create an extract and published data source
use Custom SQL? schedule refreshes then follow the steps
• C
 reate an extract of the
(server only) for the underlying published data source
connection type
Yes No
Too big to download

Define joins using Work with a data Work with a smaller ?


Tableau instead. engineer to tune data sample to Still too slow?
WHY? the database troubleshoot.

Not possible Yes No


with this query
?
Work with a data engineer to Still too slow? No E N J OY T H E S P E E D !
optimize the query
Yes

GO TO WORKBOOK DESIGN SECTION

1 interworks.com | 405.624.3214
Performance Optimization
Slow Workbook?
WO R K B O O K D E S I G N

Run a performance analysis to find the


This section assumes you have slowest views (H E R E ' S H OW ), then follow all
ruled out issues caused by the paths that apply.
database or Tableau Server.
GO TO
Uses
calculations CALCULATIONS
SECTION

Dashboard with Drop-downs with Uses


countd() Large tabular view Uses data blending
10+ sheets 100s of options

?
High cardinality
More sheets = more render time. Hide large tables until
linking field (1000+
SOLUTIONS: other selections are made
unique values)?
• Split content into multiple dashboards that limit the rows in view.
• Combine multiple sheets into one
Yes No

? ?
Consider joining data
"Only Relevant Check for other
Millions of rows? instead of blending
Values" checked? issues (above)
Note: cross-data source joins
or level of detail (LOD) calcu-
lations may be required.
Yes No Yes No

?
SOLUTIONS:
Counting a primary key Check for other
• Use context filters and check
(unique for every row)? issues (above)
"all values in context"
• Try other filter types (next step)

Yes No

SOLUTIONS: Use SUM( [Number of Consider ways to restructure


• Use "Apply" button records] ) instead. the data so that countd() isn't
• Use "Multiple Values (Custom List) type required. It's an expensive
• Use action filters from a related sheet calculation at large volumes.

2 interworks.com | 405.624.3214
Performance Optimization
Slow Workbook?
C A LC U L AT I O N S

This section assumes you are using a


Choose the data source behind the
Tableau Data Extract to optimize per-
slowest views, inspect the calculations,
formance. If using a live connection,
then follow all paths that apply.
pre-compute as much as possible at
the datbase layer.

Formula uses Uses Level of Detail Uses table calculations


parameters (lod) expressions

? ? ?

Is it a row-level calculation? Does removing the calcu- Does removing the table
(evaluated before aggregating lation with the LOD make it calculation make it faster?
with sum, avg, etc) faster?

Yes No Yes No No Yes

SOLUTIONS:
If possible, aggregate Check other
• Convert to a table calculation
values before evaluating the calculated fields
• Reshape data to elminiate the
parameter. Otherwise, see
need for lod
solutions in the next step.
• Reduce row count.

SOLUTIONS: SOLUTIONS:
• Consider alternatives that eliminate the need for a • Reduce the number of marks
parameter • Convert logic so that table
• Split the calculation so parts can be pre-computed calculations are not needed.
in an extract before evaluating the parameter

3 interworks.com | 405.624.3214

You might also like