What version of Tableau Desktop and Server are you using?
I agree with Tracy, the first thing to check is your versions. Tableau evaluates table calculations differently in different versions. One common situation for me is when using a date data type pill as the compute using, and changing from v6 to v7 but there are a few other situations as well, and the soultion may be that you need to adjust your approach.
Can you recreate your situation with some sample data? If you can, alternate approaches can explored as needed.
I'm running v7. I've resolved the situation in the subject, but unfortunately I'm not sure how. I am still having difficulties achieving predictable (and mathematically correct) results when using table calcs with large blended data sets.
I can't post sample data as it will be far too complex. In a nutshell, here is what i am trying to accomplish:
1. I have two data connections both housed in a GP database.
2. The first contains demographic data related to individuals (1B + records). The individuals records contain many different types of groupings (state, age, gender, category1, subcategory, etc). Furthermore, the individual records contain dates that indicate activity status (active/inactive). Each individual is represented by a unique "individual id". The dataset is so large because an individuals activity status is logged daily.
3. The second is transactional level data where each record represents an activity with a result. For argument sake, let's say the categories are arrived, departed, and enroute. There are further sub-groupings (let's say: planes/trains/automobiles). Each transaction is represented by a unique transaction id.
4. I have connected to each of the data sources and created relationships between the dates and appropriate sub-groupings.
5. The transactional data serves as the primary data source. The individual data as the secondary. when i calculate the unique transaction records i have no issues. But when i calculate the unique individual records (based on a COUNTD of individual id) i am getting duplicate counts.
6. I can count the unique individual records separately in a worksheet when only going against the individual data set, but it breaks when combined with the transactional data.
I'm sure i haven't provided nearly enough info (and i know it would help to see it) but can anyone point me to an advanced reference for applying table calcs to complex data blending scenarios?
I know an alternative is to summarize both data sets and then apply the calculations (which i have done previously), but the intent is to create a living dashboard.
Any help is appreciated.
Thanks again to Joe who showed me how to solve this.
If you are using time dimensions on your reports... perhaps your server work in a different time zone