That's exactly the case where Data::Diver is helpful!

This is my first exposure to Data::Diver, and I'm impressed and happy to put this in my bag of various scripts:

$ ./1.diver.pl 
$VAR1 = {
          'product_id' => 'ABC123',
          'unit_qty' => '1',
          'price_excl_vat' => '1000'
        };
$ cat 1.diver.pl
[download]

#!/usr/bin/perl
use 5.016;
use warnings;
use Data::Diver qw{ Dive };
use Data::Dumper;

my $external_structure = {
  'name' => 'Test product',
  'id'   => 'ABC123',
  'data' => {
    'enabled'     => '1',
    'internal_id' => '',
    'urls'        => {
      'main_url' => ''
    },
    'price'     => '1000',
    'unit'      => 'ST',
    'unit_info' => {
      'qty'  => '1',
      'gtin' => '1234567890'
    }
  }
};

my %format = (
  product_id     => ['id'],
  price_excl_vat => [qw[ data price ]],
  unit_qty       => [qw[ data unit_info qty ]]
);    # (1)

my $internal_structure = {};
for my $key ( keys %format ) {
  $internal_structure->{$key} =
    Dive( $external_structure, @{ $format{$key} } );
}

print Dumper $internal_structure;

__END__
$
[download]

Note that the second element of the path is unit_info, not just unit.

Well noted. This points out the problem of whether to "fix" the original post. If OP does, then both of our posts might need "fixing" too. My preference is to leave the warts on the original post.