Most advices already passed should get you on your way, but I'd like to add one thing, as you explicitely mention Excel.

As I am confronted with an increasing customer demand of results being presented in Excel, I have found that generating the Excel should be the final step.

Don't see Excel as your (main) source for data manipulation. Read input from many sources (Excel, LibreOffice, CSV, JSON, YAML, XML, MS-SQL databases, PostgreSQL databases, Oracle databases, whatever ...) then use your business logic in perl, optionally store intermediate results in a local database like PostgreSQL or SQLite so small fixes can be done quickly without parsing/fetching all the sources again, and then generate the required output. Another big advantage of storing your work data in a local database is that the generation of the required outpout is much easier to test/change.

If Excel is requested, I prefer to export every sheet in CSV, and then use the tools available in Text::CSV_XS' examples, like csv2xlsx to convert and merge the data into an Excel file.