I think your speed problems result from linear searches.

You may want to organise your data in a nested structure.

See perldsc And perlref

Like (draft)

$h_record = { 
         date     => '2018-01-01', 
         product  => 'apple', 
         price    => 200, 
         quantity => 50000,
     };
$product{apple}{'2018-01-01'} = $h_record; 
$date{'2018-01-01'}{apple}    = $h_record;
[download]

• Like this you can easily lookup products and dates and pass records to subroutines.
• Another hint is to cache and not repeat already done calculations.
• Last but not least keep an eye on memory consumption, as soon as the system needs to swap heavily, you'll be lost in a black hole. (Only processing chunks may help here)

Please note (like others mentioned) that this approach can't compete with an SQL DB in terms of efficiency.

Those hashes need lots a memory where a SQL is managing an index on columns of the same table. (Think of all the hashes you'd need to easily look up year, month and day separately)

Also reading files to build the data structure will cause you overhead where SQL is persistent.