Starting a new dashboard, it is nice to have a general idea of what kind of numbers you expect to see on your dashboard from the provided data. So, where to start if the numbers you get are not what you are expecting?
- Check Data Types
- One step at a time
1. Data Types
In the same way as Alteryx, Tableau might infer data types from the data it receives. The data may change, also, the inference itself can be not perfect. In the worst probable scenario, you can see the concatenation of numbers as strings. In a more typical case – zeros instead of decimals because your fields are set up as integers instead of decimals. And it is not actually easy to spot the difference just by looking at the screen.
Tableau will graciously provide you with default aggregations for any numeric field, which it thinks are right for you. Will those actually be working is a totally different matter. For instance, if you are using some types of joins (i.e. joining data tables with a different level of granularity) you need to deal with some duplications in your data source, which means you might need to use AVG or a combination of AVG and SUM in LOD instead of a simple SUM to get totals. You probably will need to build an even more complex aggregated calculation for an average (probably, using COUNTD).
Always remember the level of granularity of your data, mostly to understand how your aggregations work (see above), but also to understand if you need to pre-process your data for better performance and easier querying.
Depending on how you merge your data in Tableau, the results of calculations may be different. The main difference comes from using joins and relationships because joins might result in duplications in your data (and you need to correct your aggregations, as above), while relationships are supposed to take care of the duplications (but if it did work in your case is always a matter for investigation). If you are confused about your calcs and if there is duplicated rows in your combined data source, try troubleshooting your calculations step-by-step as I recommend below.
5. Step-by-step checks
If your calculation results in an expression that doesn’t look right or doesn’t make sense, try checking it one function at a time. Create a cross-table with the exact dimensions you want to use in your visualisations, the original measures you use in your calculation, and add one function at a time to your calculation. If you have a conditional expression, take the condition, the then-expression and the else-expression out on your column shelf separately and check if the results match your expectation.