Tableau troubleshooting 02: It opens, but you cannot change it
A difference in a chart catches your attention. Breaking down this category might explain it. Adding another year might reveal whether it is a temporary change.
Then the work stops. The Tableau workbook you received lacks the fields you need, and you have neither the source data nor the preparation steps. To investigate further, you must contact its creator, locate the files and perhaps prepare the data all over again.
Just as you begin to see what needs investigating, you cannot test it. The tool intended to support analysis is now stopping it.
When someone builds a Tableau tool for you, consider whether you can carry the analysis further as well as whether the finished views work. This article covers commissioning and handover whenever the people building and using a tool are different, whether they work in different companies or within the same organization.
1. Why a workbook can display charts but still ask for source data
The charts in a workbook you received appear correctly. You can change filters. But when you try to inspect or refresh a data source, Tableau may ask for the original Excel file or a connection you cannot access.
Tableau can convert all or part of source data into a format used for aggregation and display, then save it. Tableau calls the saved result an “extract.” Here we use “extracted data” to distinguish it from the operation of extracting it.
If you have the extracted data, the views may still display its stored values even when the original Excel file is unavailable. Incorporating corrections or additions from the source, however, requires access to the source file or database to refresh the extracted data. Tableau documentation: Refresh Extracts
For example, the creator may have generated extracted data from an Excel file on their own computer without handing that file over. The recipient can see the current charts, but cannot recreate the extracted data from that Excel file.
Having the data needed to display a view does not necessarily mean you have the material needed to update it.
A request for source data does not prove it has been lost. A file may simply have moved, or permission to access a connection may be missing. First inspect the message and connection destination, then establish where the required files are and whether you have access. Preserve the workbook as received and investigate using a copy.
2. What you need depends on what you want to change
Changing a Tableau tool can involve several different stages.
| Intended change | What is needed |
|---|---|
| Adjust colors, layout or filters using existing fields | An editable workbook and the data needed to display it |
| Correct source values or add another year’s data | The source data and instructions for refreshing the extracted data |
| Change classifications or the rows and columns selected | Original data, classification tables, preparation flows or code, and working instructions |
With suitable editing software and permissions, the workbook and extracted data may be enough to change a view. Adding a field absent from the extracted data requires going back to an earlier stage.
If the data was prepared in Excel, Tableau Prep or another tool before Tableau loaded it, those preparation steps are needed too. The prepared table alone may not reveal which rows were excluded, how records were classified or which units were converted.
Receiving a .twbx file also calls for checking its contents. This format can package a workbook with related data, but that does not establish that preparation flows or earlier source material are included. Tableau documentation: Packaged Workbooks
3. Do not let analysis stop when an answer is coming into reach
One finding leads to another question. Once you see a difference, you want to locate it more precisely. Once you narrow it down, you want to test whether it persists under other conditions. Through this process, the evidence for a decision becomes more specific.
The initial views may not let you investigate that far. If you cannot add the necessary fields or change the aggregation, the analysis stops partway through. Earlier findings remain, but you cannot test them further or turn them into a basis for a decision.
Locating production files, reconstructing preparation steps and reproducing the existing state all take time away from investigating the question. Repeating this whenever responsibility changes makes it difficult to build on previous analysis.
You do not need to implement every possible future feature at the start. You do need to be able to make changes in response to new questions. Obtain the source data and preparation steps so that your team, or the next person responsible, can continue the analysis.
The purpose of an analytical tool is to investigate what you need to know and establish grounds for a decision. Can that work continue after the views have been completed? This needs to be checked both when commissioning the work and when accepting the handover.
4. Enable the recipient to trace the figures
Source data and preparation steps are also needed when a displayed figure raises a question.
Was the value already different in the source? Did preparation drop a row? Did Tableau’s aggregation rules produce the difference? Intermediate data and instructions allow comparisons at each stage.
Without them, the recipient may struggle to verify the creator’s explanation. Whether or not anyone intended that outcome, the result is a gap in what can be checked.
You cannot know in advance what an analysis will reveal. You can, however, check whether source data has been processed according to agreed rules. Not yet knowing the answer to an analytical question is a different matter from being unable to verify the processing.
Part 1, “When source figures are missing or do not match,” explains practical comparisons. Performing those checks requires material that lets you trace how the figures were produced.
5. Agree what counts as completion when commissioning the work
The creator and the user may not consider the same point to be the end of the job. Producing the specified views may complete the creator’s assignment, while the user’s analysis begins there.
If continued updates or changes are needed, state those requirements at the outset. Instead of assuming they are included, agree the data and instructions to be delivered, the operations possible after handover, and how acceptance will be checked.
For example, specify that your team will update next year’s data, that an internal colleague will change classifications, or that another developer may take over later. A concrete account of future use makes the necessary handover materials easier to identify.
| What to hand over | What to include |
|---|---|
| Workbook and data needed for display | An editable set of files with the adopted version identified |
| Source and prepared data used in production | Versions, years and scope, and the relationships between files |
| Preparation flows, code and manual instructions | Classification tables and settings needed to reproduce the preparation |
| Connection and update instructions | File locations, connection destinations, required software and permissions, and execution order |
| Checking methods and results | Comparison rules, checking code, verified scope and unresolved issues |
For databases and other sources that cannot be handed over as files, establish the recipient’s ability to connect and update. If the process works only in its creator’s personal environment, clarify how it will operate when responsibility changes. Also establish whether recipients can modify the materials and pass them to a subsequent maintainer.
A request could include wording such as:
After delivery, we intend to update the data and change aggregation rules as needed. Please include the source and prepared data used in production, preparation flows or code, and update instructions alongside the workbook. For external data sources, specify the connection requirements and permissions needed for our continued use. During handover, we will confirm that we can follow the instructions to perform an update in our environment.
6. Perform an update in the recipient’s environment
A handover check should go beyond a file inventory. The person taking responsibility should follow the instructions in the receiving environment. A process that works on the creator’s computer may depend on files or permissions absent elsewhere.
Use working copies of the source data and workbook. Change a test value or add a row, for example, and check that it passes through preparation and extracted data to the displayed result. For external databases, use a suitable test environment or dataset rather than modifying production values for the test.
If classification changes are part of the handover, try that operation too. Add any undocumented steps or settings known only to the creator to the instructions.
This single exercise does not verify every calculation or interaction. It establishes whether the agreed updates and changes can be carried out by the recipient. Use the data comparisons and screen checks described in Part 1 to verify the figures.
At Rokugeisha, we believe tools for continuing analysis should be handed over in a state that lets the next person update them, change them and trace the figures. We consider the data and procedures behind the finished views to be part of the tool.