Keep in mind ways to standardise, harmonise the data with other data.
for example convert all variable names to lowercase, and enforce underscores rather than dots in names
Into a spreadsheet take a list of all files, tables, queries, worksheets (using RODBC sqlTables to extract this list from databases or spreadsheets)
assign a new variable based on the observational types, these will act as an umbrella to collect the LPU.
Then I’d try to carve these umbrella groups into separate LPU ie: Jellybean_History _by_site, Jellybean_Interval_Table, Days_Since_Jellybean) and Jellybean_Response.
And then start a script in whatever language (like R),
Structure the script into sections, the first section would be Jellybean, then first subsection Jellybean_History, then another subsection Jellybean_Response.
Then move on to the next section/subsection.
Then scroll up and down between sections adding in exploratory code (maps, graphs, cross-tabs) and comments about what you find, think, decide, change in main data files (if appropriate).
Then shuffle these around like moving blocks of a jigsaw puzzle until happy enough to share with someone else.
Do a presentation of the work so far and make notes immediately afterward about things you thought of during the presentation