Is it all really about data?

Of course not. But in the majority of the cases of failing standard tests, it worked out that way for me. Based on this experience, my best practice in getting a failing standard test fixed is the following sequence. Try fixing the error by adjusting:

1. The shared fixture; if that does not do the job.
2. The fresh fixture; if not.
3. The test code, often the verification part, in this case you hit upon a test code bug; if still not working.
4. The app code, you found a real bug!