You can also split the detection of WINDSOR RIVER from the extraction on the values. Here is my way ...

#!usr/bin/perl
use strict;
use warnings;
use List::Util qw(sum);

my @windsordigits;
while (my $x = <DATA>){
  next unless $x =~ /WINDSOR\sRIVERSIDE/;
  push @windsordigits, +(split /,/, $x)[-1];
}
die "No 'Windsor digits found in input\n" unless @windsordigits;
my $average = sum(@windsordigits) / @windsordigits;
print "Average is: $average\n";

__DATA__
CA006139520,"WINDSOR RIVERSIDE, ON CA",2018-01-02,10
CA006139520,"WINDSOR RIVERSIDE, ON CA",2018-01-02,20
CA006139520,"WINDSOR RIVERSIDE, ON CA",2018-01-02,14
CA006139520,"WINDSOR DRIVE, ON CA",2018-01-02,10
[download]

I use a regex to match, then if we matched in the regex, split the line on the , character and take the last element or the resulting list using the [-1] offset.

Update

Cheers,
R.