+=head1 PARAMETERS
+
+=head2 TIME PERIOD
+
+C<start_month>, C<start_year>, C<end_month>, and C<end_year> specify the date
+range to be included in the report. The start and end months are included.
+Each month's values are summed from midnight on the first of the month to
+23:59:59 on the last day of the month.
+
+=head2 REPORT ITEMS
+
+=over 4
+
+=item items: An arrayref of observables to calculate for each month. See
+L<FS::Report::Table> for a list of observables and their parameters.
+
+=item params: An arrayref, parallel to C<items>, of arrayrefs of parameters
+(in paired name/value form) to be passed to the observables.
+
+=item cross_params: Cross-product parameters. This must be an arrayref of
+arrayrefs of parameters (paired name/value form). This creates an additional
+"axis" (orthogonal to the time and C<items> axes) in which the item is
+calculated once with each set of parameters in C<cross_params>. These
+parameters are merged with those in C<params>. Instead of being nested two
+levels, C<data> will be nested three levels, with the third level
+corresponding to this arrayref.
+
+=back
+
+=head2 FILTERING
+
+=over 4
+
+=item agentnum: Limit to customers with this agent.
+
+=item refnum: Limit to customers with this advertising source.
+
+=item remove_empty: Set this to a true value to hide rows that contain
+only zeroes. The C<indices> array in the returned data will list the item
+indices that are actually present in the output so that you know what they
+are. Ignored if C<cross_params> is in effect.
+
+=back
+
+=head2 PASS-THROUGH
+
+C<item_labels>, C<colors>, and C<links> may be specified as arrayrefs
+parallel to C<items>. Those values will be returned in C<data>, with any
+hidden rows (due to C<remove_empty>) filtered out, which is the only
+reason to do this. Now that we have C<indices> it's probably better to
+use that.
+
+=head1 RETURNED DATA
+
+The C<data> method runs the report and returns a hashref of the following: