Video |

Tableau | Data visualisation | Tool strategy

5 additional checks in Tableau Workbook Optimiser - New in Tableau 2022.2

Tableau 2022.2 quietly adds five more performance checks to the Workbook Optimizer, and a couple of them I'm surprised weren't there from the start.

Tim Ngwena •4 July 2022 •1 min read •6 min watch •923

Part ofWhat's new in Tableau 2022.2

Watch on YouTube

Tableau 2022.2 expands the Workbook Optimizer from 12 to 17 tests, adding five new performance checks
Minimise date calculations and conversions in Tableau, use DATEPARSE and built-in functions like DATEDIFF, and prefer relative date filters over discrete ones
Set a fixed dashboard size so Tableau Server or Cloud can cache the rendered workbook rather than re-rendering it for every screen size
Replace live data connections with extracts where possible, as extracts are almost always the more performant data source
Reduce data blending in favour of relationships, and use a calculated field with a CASE statement instead of native grouping for better performance

0:00In 22.1, Tableau introduced something

0:02called the Workbook Optimizer, essentially

0:04a series

0:05of tests and checks that Tableau does to

0:07make sure your workbook is optimized for

0:09production.

0:11In 22.2, they're adding five additional

0:14tests to that feature.

0:15Let's get started.

0:16So I've just recorded a video about the

0:18extension gallery inside of the product,

0:20and I was looking

0:21at an accelerator built on my colleague

0:22about Tableau server.

0:24That's what's open here.

0:25And I've actually run the Workbook Optim

0:27izer, and you can see here that there are 17

0:29tests.

0:30If I switch to a previous version of Table

0:32au, you can see there's actually 12 in the

0:34past.

0:34So there's five additional tests.

0:35Now, I didn't know a way to easily figure

0:37out what these changes are.

0:39I'm not looking at the release notes.

0:40So what I did is I essentially took a

0:42screenshot of all the features, and I put

0:44them on one

0:45page.

0:46And so if we actually go through this, it's

0:47much easier to see what's going on.

0:49In essence, the first screen gap is from

0:50the newest version of Tableau, 22.2.

0:53The second one is from 22.1.

0:55And so we can go through each of these

0:57sections for the same workbook and see

0:58which tests

0:59are new.

1:00So let's first start with the very top

1:03section here.

1:04There's obviously nothing new in terms of

1:05things that have failed in this workbook.

1:07So there's no new check that's coming up

1:09with a red issue here.

1:10However, if I go to the next section, you

1:12can see that the very top we have four,

1:14whereas

1:15before we had three, which means there's

1:17one new test.

1:18And if we just go down the list, you can

1:20see that the newest option here is a test

1:21that

1:22checks whether it's used a date calculation

1:24or not.

1:25If I go back to the products and I go back

1:27to the new test here, I can actually expand

1:29this and look at that piece of text.

1:31And it says here, the date source new audit

1:33tables contains nine day calculations.

1:36Date logic can be complicated, minimize the

1:38amount of day calculations and conversion

1:40you have to do in Tableau.

1:42If necessary, use date pass and make date

1:44before other methods and try and use built

1:46in functions like date diff when possible.

1:48If filtering on date, use relative date

1:50filters as a continuous date filter instead

1:53of a discrete

1:53filter.

1:54Essentially, those perform a little bit

1:55better.

1:56So there's a little bit more context to how

1:58this all works.

1:59And again, if I click on the learn more

2:01capability, it takes me to this page, which

2:03Tableau have

2:03sort of been updating over time.

2:05So I think they're going to add more and

2:06more of these tests.

2:07You can see it's still got the

2:09documentation for the previous version of

2:10Tableau, but this

2:11will be updated for the newest version and

2:13should have the notes about the new test

2:15there

2:15as well.

2:16So the first test uses day calculations.

2:19Let's switch back to our list and let's try

2:21and find the next one.

2:22Obviously, I've got nothing else in Amber,

2:25but if I go over to the green section, the

2:27new version, I got 11 that passed and in

2:29the previous version, I got seven.

2:30So that means there's four new tests in

2:32here that we need to find.

2:34So let's go ahead and try and look at these

2:35.

2:35The first one is number of data sources.

2:37That's exactly the same unused data sources

2:40.

2:40My hunch is actually these are all going to

2:42be at the bottom.

2:43If I go right to the bottom number of

2:45filters.

2:45You can see that all the checks just being

2:48added in a sort of sequential order.

2:51They're right here at the very bottom.

2:52So if actually I went to take a screenshot

2:54there, what I actually meant to do is annot

2:55ate

2:56these like this.

2:57So the four tests are users data blending,

2:59dashboard size not fixed, live data

3:01connection

3:02and users groups.

3:03These are actually really important ones.

3:05I'm surprised some of them didn't make it

3:06to the original release, but here they are.

3:09Let's go take a look and see what those

3:10tests actually do.

3:11Let's go to the right version of Tableau as

3:13well to do that.

3:14Let's open up the 11 that pass and go right

3:16to the bottom.

3:17Let's start by looking at users data

3:18blending.

3:19So if I open up that one, the workbook

3:20follows best practice for this guideline.

3:22So it's working perfectly fine.

3:24But the text reads this data blending

3:27performance is driven by the number of

3:29unique numbers

3:30in the linking fields.

3:31Consider using relationships when possible.

3:33If a blend is required, try to use low

3:35cardinality linking fields.

3:37That makes a lot of sense.

3:39Try reducing the amount of blending because

3:40it's a very resource intensive way of quer

3:42ying

3:42your data.

3:43Dashboard size not fixed.

3:46Now I come across dashboards that have

3:48changing parameters all the time.

3:50And I always tell clients, hey, you do know

3:52that you are incurring a performance

3:54penalty

3:54by doing this and they don't understand why

3:56.

3:56And essentially I have to start explaining

3:58really weird concepts about Tableau server

4:00and caching.

4:01But the short story is this.

4:03If your dashboard doesn't have a fixed size

4:05, Tableau can't create a cached copy of that

4:08workbook because it doesn't know what size

4:10screen the user is going to have.

4:11So what it does is it basically renders the

4:13visualization new every single time,

4:15stopping

4:16it from being fast and performant.

4:17If you go to a typical website, what

4:19websites typically do is actually store

4:21most of that

4:22information in cache.

4:23They don't render it new every single time

4:24someone visits because that takes a lot of

4:26computing power.

4:27Same here.

4:28If you don't have a fixed dashboard size,

4:30it's not going to allow you to use the

4:31caching

4:32capability that works on Tableau server or

4:34Tableau cloud.

4:35So that in turn leads to performance issues

4:38.

4:38And you can see here that Tableau have

4:39added it to the checklist.

4:40That's something really good to be aware of

4:42.

4:42Now it's a check in the optimizer.

4:44Live data connection.

4:45The workbook follows best practice for this

4:47guideline.

4:48Tableau extracts are designed to be as

4:50efficient as possible for use with

4:52analytics.

4:53Using extracts is one of the easiest ways

4:55to improve performance due to the data

4:57source.

4:57Extracts also have more native features for

4:59optimization.

5:00So essentially just hinting to people that,

5:02hey, you've got a live connection, try

5:03using

5:03an extract instead.

5:04And I also do tell people there's quite a

5:07lot.

5:07Extracts should normally always be the best

5:09performing data source.

5:11Even if you have a rip raw, fast database

5:13in very few circumstances, will an extract

5:16perform slower?

5:17So just bear that in mind.

5:20Users groupings.

5:21This is the final test.

5:22Tableau's native grouping functionality

5:24loads the entire domain of the field, which

5:26can

5:26have a performance impact.

5:28Using a calculated field with a case

5:29statement to group instead can have better

5:31performance.

5:32So we're not using grouping here, but

5:34essentially it's saying there's a couple of

5:35capabilities

5:36here in SQL that perform a bit better than

5:38the way the native grouping functionality

5:41works.

5:42So you might be getting a few performance

5:44benefits by doing grouping in a specific

5:45way.

5:46Again, once the documentation is out on

5:47this, you'll get a lot more detail when you

5:49click

5:49on that learn more option.

5:51So that's all the new optimizer checks that

5:53are being done in 22.2.

5:55Let me know what you think of those checks.

5:57What checks are still missing?

5:58Let me know in the comments below and I'll

5:59catch you in the next video.

6:01Thanks for watching.

6:01[BLANK_AUDIO]

Rollup

Tableau 2022.2 adds five new checks to the Workbook Optimizer, taking it from 12 to 17 tests, and each new check flags a specific performance habit worth fixing in your own workbooks.

The Workbook Optimizer runs a series of pass/fail tests against a workbook to flag performance issues. Comparing the same workbook's results between 2022.1 and 2022.2 reveals which checks are new.

The Breakdown

Compare optimizer results across versions 0:20

If you want to spot what's changed in a Tableau feature like this, one approach is to run the same workbook through the optimizer on both versions and screenshot the results side by side, making new checks easy to eyeball.
New check: excessive date calculations 1:20

The optimizer now flags workbooks with too many date calculations or conversions. Minimise these where you can, use DATEPARSE or MAKEDATE for parsing, prefer built-in functions like DATEDIFF over manual logic, and use relative (continuous) date filters rather than discrete ones.
New check: data blending 3:02

Blending performance depends on the number of unique values in the linking fields. Use relationships instead of blends where possible, and if a blend is unavoidable, blend on low-cardinality fields.
New check: dashboard size not fixed 3:41

An unfixed dashboard size stops Tableau Server or Cloud from caching a rendered copy, since it can't predict the viewer's screen size, so it re-renders the view every time. Setting a fixed size lets the server cache the workbook like a website caches pages, which is a real performance win.
New check: live data connection 4:45

Extracts are almost always faster than live connections, even against a fast raw database, because they carry extra native optimisation features. Where performance matters, use an extract instead of a live connection.
New check: native groupings 5:21

Tableau's built-in grouping feature loads the entire domain of a field, which can hurt performance. Using a calculated field with a CASE statement to do the grouping instead can perform better.

Worth Knowing

The Learn More links in the optimizer point to Tableau's documentation page, which was still showing 2022.1 content at the time of recording but is expected to be updated with details on the new checks.
The five new tests all appeared at the bottom of the results list in the same order as the old ones, which made them straightforward to locate once you know to look there.

Use It When

Run this comparison mentality whenever you're auditing a workbook for performance before pushing it to production, especially if you're relying on blends, live connections, native groups or flexible dashboard sizing without realising the cost.

How this Rollup was made provenance & method

A Rollup is drafted by AI from the video's transcript, then reviewed and edited by Tim. Everything used to produce this one is listed below — the model, the exact prompt, and the source video — so the process is transparent and reproducible.

Transcription: On-device — NVIDIA Parakeet v3 for recent videos, OpenAI Whisper large-v3 for earlier ones. The transcript never leaves the machine or gets published.
Drafting: Claude Sonnet 5 in the cloud, from that transcript.
Prompt: The exact Rollup prompt (v2) — the full system prompt, unedited.
Source video: Watch on YouTube
Drafted: 5 July 2026 at 09:38
Reviewed & edited: 5 July 2026 at 09:41 · by Tim Ngwena

Model + prompt + video is everything you'd need to recreate a Rollup like this yourself. The one thing we don't share is the transcript.

Rights. The video and its transcript are the property of TN Media Ltd. Unauthorised use or download is prohibited. © TN Media Ltd.