Data quality (sometimes less obvious, such as loyalty card abuse, scam accounts),
Business process changed but downstream data applications not aware of the change (say method of forecast changed so like for like comparison became useless without adjusting) Models hard to extend, a report model is built for a specific purpose, possible doesn't contain the low level keys needed to link to a new attribute.
If the above is combined with insufficent documentation (another issue), then this makes analytic work harder to maintain.
After analytical results are released, how to regularly monitor how users are using it (this is essential to get better ROI from analytical work); also for the same report, user may request different views or further customise some metrics - these should be somehow evaluated during design.
Data Inconsistency, how to reconcile between various source systems and manage the difference.
Single Customer View- how to obtain a closer-to-truth view of business data.