The following question concerns a phase of development where I’m working with a new data set to figure out its features...

Concerning higher level Mathematica constructs such as TimeSeries, EventSeries, Associations and DataSets, my initial tendency is to convert raw imports of my data almost immediately into these higher level constructs. However, during my exploration phase of figuring out features of my data, I quickly realize I am frequently invoking //Normal — so as to invoke Mathematica functions that are not tuned for the higher level constructs. And then that leads me to worry about runtime efficiency (or worse) that I am committing too early to converting my raw data into these higher level constructs.

Lately, I’m thinking to myself that I ought to re-think my Mathematica notebook development into a “exploration” phase were I use only basic Mathematica constructs . And later, when I am confident I understand my data set features, that the second section of Mathematica notebooks is where I’m committing to using the higher level constructs knowing I won’t need to devolve them with //Normal.

In summary : If there is indeed negative runtime consequences by invoking //Normal too frequently, what is a good practice of creating a cookie-crumb of initial data analysis to discover essential features of unknown data sets, followed by a full-fledged jump into the higher level Mathematica constructs as the data processing pipeline takes shape?