I know you "found" the code on the internetz and did not write it yourself. Maybe it is time to find better example sites than, as this is bad code: do not mix diamond read with parse on CVS, as it will break on embedded line separators.

Of course I have no idea what version of Text::CSV_XS you are using, but this should be simplified to something like:

use strict;
use warnings;   # <-- always use strict *and* warnings
use Spreadsheet::WriteExcel;
use Text::CSV_XS;

# Check for valid number of arguments <-- @ARGV in scalar context give
+s the number of entries in the list
@ARGV == 2 or die "Usage: csv2xls csvfile.txt newfile.xls\n";

# Open the Comma Separated Variable file
open my $fh, "<", $ARGV[0] or die "$ARGV[0]: $!"; # <-- use lexical fi
+le handles and 3-arg open

# Create a new Excel workbook
my $workbook  = Spreadsheet::WriteExcel->new ($ARGV[1]);
my $worksheet = $workbook->add_worksheet ();

# Create a new CSV parsing object
my $csv = Text::CSV_XS->new ({ binary => 1, auto_diag => 1 }); # <--- 
+use diagnostics and binary

# Row and column are zero indexed
my $row = 0;

while (my $row = $csv->getline ($fh)) { # <-- safe and fast parsing
    # due to auto_diag there is no reason for error-checking
    my $col = 0;
    $worksheet->write ($row, $col++, $_) for @{$r};
    $row++;
    }
[download]

Now, doesn't that read much easier? Also easier to extend and to maintain.

Note that this is a basic simplified example from your code. The tools mentioned by the others also take note of dates and special formats if you want and have examples of how to deal with errors. (and csv2xlsx can merge many CSV files into a single Excel file.

Have you seen csv2xls/csv2xlsx?