Hi Nam, it actually depends what those changes in the LDM are. In some cases you might very well stay safe. Evolving the LDM as your project evolves and grows is pretty standard thing in GoodData. Thr LDM and insights actually uses internal IDs for all objects so even if you rearrange things a bit, it might not affect your reporting if you do it in the right way.
• Obviously if you for example try to drop some attributes/facts which you are using in the insighta, this would require to also drop the insights.
• On the other hand, adding new attributes or datasets should be absolutely safe.
• For moving attributes between datasets - this depends on how you do it - what is the semantic. For example it should be completely OK to move one attribute from a dataset to another dataset that is "to the left" of the original one (i.e. take "Customer" attribute from your "Order" dataset and move it to new "Customer" dataset which is connected to the "Order").
In general - if you do not drop something, you should be safe in a sense that no metric or insight should dissappear. The worst case is some of them might become not computable if you make the changes structurally not compatible to the calculations (i.e. if you would move "customer" to dataset not connected to "order", an insight with "order amount" and "customer" in it would no longer work.
Also note that some changes might require you to reload the data and make sure you maintain the correct mapping to the source data.