What to infer if data augmentation does not improve F1 score?

I am classifying text on a highly imbalanced data set, hence data augmentation is necessary.

However augmentation has not led to a higher macro averaged F1 score.

What should be my step moving forward?

Should I try a different augmentation method? How do I increase the macro averaged F1 score?

augmentation data datascience f1