If, for any post_date, there are more that 3 transactions between post_date-29 and post_date (inclusive) then fee=1. BUT, while your data are grouped by cust_id, they are not sorted by date within cust_id. So, to avoid sorting, you can use a double dow. I.e. have 2 do loops reading each cust_id. The first loop increments a subset of a data array (every date from post_date through post_date+29). The second loop checks the result for any given post_date, looking for values over 3:
data have;
input Cust_ID TransactionID POST_DATE :mmddyy10.;
format post_date date9.;
*if cust_id=987;
datalines;
123 321 2/28/2018
123 322 2/10/2018
123 323 2/2/2018
123 324 2/1/2018
123 325 11/27/2017
456 326 2/1/2018
987 327 3/5/2018
987 328 2/5/2018
987 329 1/5/2018
987 330 12/5/2017
987 331 11/6/2017
654 332 11/13/2017
654 333 2/28/2018
654 334 2/28/2018
654 335 2/26/2018
654 336 2/26/2018
654 337 2/23/2018
654 338 2/21/2018
654 339 2/21/2018
654 340 2/20/2018
654 341 2/20/2018
run;
%let lower_bound=%sysfunc(inputn(01jan2017,date9.));
%let upper_bound=%sysfunc(inputn(31jan2019,date9.));
data want (drop=_:);
array _count{&lower_bound:&upper_bound} ;
do until (last.cust_id);
set have;
by cust_id notsorted;
do _d=post_date to post_date+29;
_count{_d}=sum(_count{_d},1);
end;
end;
do until (last.cust_id);
set have;
by cust_id notsorted;
fee=(_count{post_date}>3);
output;
end;
run;
Notice the _count array is index by date. And the lower bound of the index is set to the numerical equivalent of sas data value 01jan2017, and the upper bound to the equivalent of 31jan2019. Just use a lower bound date as the minimum expected date, and the upper bound at least 29 days after the highest expected date.
