Wow. These are deeper waters than I knew:
I employed the method in the workbook posted by Yuriy Fal and excellently explained by him here:
This got the correct rolling customer count.
I duplicated the method to get the rolling sales, changing the [In Window] to be :
Please see the workbook v10.3 attached in the Forum Thread.
279697rolling.twbx 46.8 KB
Thank you! This is awesome.
But now it gets deeper. I manage to implement this technique in my dashboard by following your advice. But I forgot to mention the size of the real dataset. So, the other challenge is handling a 23,485,632 rows view with this technique. I tried to do reduce the data by grouping before loading to Tableau and following the technique but It did not improve the dashboard substantially.
This ocean is deep and wide.
I can imagine that the rolling calculations are bogging down performance.
But I'm not sure if there is another method that can avoid table calculations, and yet still get the desired result.
1. Please describe what each row in your true dataset represents: is it every customer transaction?
Then in your aggregation, does each row look like what you have in your sample:
for each month the total of sales by customer?
2. Are you using Tableau Prep for the aggregation? Can you post a screenshot of your flow scheme?
3. Would it be permissible for you to post a screenshot of your workspace, including the pills
and their locations on the filter shelf, row/column shelves, etc.? If the fieldnames themselves
are sensitive, you can black them out.
4. You may already be familiar with these, but if not, here are some links to general tips on improving performance: