Improve Tableau dashboard/reports performance.
Just Imagine of a day when you want to
quickly pull out some insights from your data to publish numbers to your
business stakeholder but your dashboard just keeps loading!!……Loading…!!..Loading!!!
Isn’t it a painful experience when you
have to wait for a while to see the numbers on the dashboard? I believe, most
of the business users have had experienced it. Seriously there is no use of
having a tool which doesn’t show up the numbers quickly but honestly speaking
the problem is not always with the tool instead it is with the way we
configured the tool.
Tableau is undoubtedly one of best
business reporting tools available in the market. The power of tableau can be
experienced only when it is used in the proper way, what I mean here is one
should know the tableau best practice and optimization methods to leverage its full
potential.
Tableau helps non-technical persons to
plug and play with data and draw some cool insights, here the users doesn’t
need to have technical knowledge of data models, data structures, SQL and data
types. Unlike Data guys business users and stakeholders usually don’t possess
this skill set all they know is business process and tableau helps them to
generate Ad-hoc reports instantly giving them the freedom to self-service. These
ad-hoc reports doesn’t serve over long run in the business and in turn becomes
a burden when published on Tableau server.
Let’s through shed some light over Tableau’s
best practices, I’m confident that by implementing the below recommendations
one can definitely minimize the tableau report loading/refresh time.
Well there are some areas which are
not in control of a report developer but has a significant impact on the
performance like network speed, number of concurrent users accessing the
tableau server OR database server, having a live connection to data source
etc., these are subjective so we aren’t going to focus on these things for now,
We focus on the topics where we have
full control, please be mindful of the below topics before you actually begin
the report building.
·
Custom SQL
·
Live/Extract
·
Adding Data source filters
·
Sorting
·
Avoid Table views
·
Filter exclude/include options
·
Count Distinct
·
Quick Filters
·
Dashboard layout(resolution)
Custom SQL:
Custom SQL is a great feature of tableau
but keep in mind that it will consume a lot of memory, custom SQL actually runs
in tableau memory this means every time you interact with your report/dashboard
tableau has to run the query over the network and get the results from database
this is load on server. I don’t recommend custom filters on a live connection
instead create a materialised view of and access it through tableau.
Live/Extract connection: Every tableau
users would be aware that extract connections perform better than live connection
but again this is very subjective and depends on individual business
requirements. I recommend to prefer extract connections where ever possible.
Data source filters: adding data
source filters will cut down the volume of data that is being loaded into tableau
memory. Example if we are building a monthly report then we don’t need daily
transactions in our data set, here rolling up the data to month by applying
aggregation option will help to minimize the records and improves the
performance. In some cases we would have a data set having data from all regions
over a vast time period but the report is intended to be used by just one
region/data for a just certain time period in this case applying a filter in
the report will fulfill the purpose but it doesn’t add any value to have the
data in tableau, instead of this we should use data source filters. The point
to focus here is to analysis the business requirements before you actually
start building your report, I recommend to catch-up with the business stakeholder
or end users to understand what exactly they want to see in the report and work
on to model your data.
Table views: One picture speaks 100
words. Tableau is a visualisation tool and should be used to aggregate data
values and report through visuals. Most
of the business users, particularly finance teams loves to see row level data
just like in an excel sheet, but be mindful that this is not what tableau supposed
to do. Reporting row level data in using tableau table view is not at all
recommended (simple and small table views to show the details is always
welcome) but avoid recommending Table views to your report users. I don’t mean one
shouldn’t use table views but keep in mind that tableau is a visualisation
tool.
Secondary Data source:
Data blending is a powerful feature of
tableau helping users to build reports that are not possible to join but try to
avoid creating calculated fields which refers data from secondary sources this
is again a memory killer.
Sorting: This is a
quite used feature of tableau and doesn’t make any sense if I ask anyone to
avoid using it but the fact is sorting is a memory consuming function. I would
recommend to avoid sort on dimensions which has a long text in it I know it is
not easy to avoid sort function but please put your best efforts to avoid where
ever possible. Clear sorting option where it is really not required.
Quick filter: Quick
filters are very handy for every report and users love to have filters all over
the report but be mindful that this is a memory and performance killer. Tableau
offers two ways of applying filters include values/Exclude values I recommend
to make the best use of these options. If you need to show just 2 values out of
25 then don’t deselect 23 values instead use exclude option and select deselect 2 values, this has a significant effect on tableau’s behavior. This approach
is usually come handy while avoiding null values in the report.
Count Distinct function:
Count Distinct is a memory killer. I
know it’s not possible to avoid the CountD function but please try to avoid
where ever possible.
Dashboard design and use of Action
filters
·
Use “Show relevant values” option to one of
two filters only, applying this on all quick filters will put a very big load
on tableau server.
·
Use dashboard action filters and engage users
to click and drill down the reports instead of showing all quick filters on
dashboard.
·
Give a type in feature for given quick filter
instead of multiple select values.
·
Prefer Range over automatic size option in
dashboard. Automatic is a very good option when your report is viewed across
various screen sizes but keep in mind that tableau need to serve every screen
size when a request is received this means extra load on tableau server. Always
create respective layouts like desktop, tablet and mobile instead of automatic(Exception
is when you are have less number of users hitting the tableau server then
automatic size is not a big load).
There are few more topics which have moderate
impact of performance and would discuss over those in some other article. To
sum up the above discussed concepts
Avoid
in the first place Data volume, Sorting, filter conditions,
Tabular data, custom SQL and showing up quick filters.
Significant
impact but out of control: Data types, Hardware, Load on server,
concurrent users and Extract/Live connections.
