How about this: Create an index on Cust_no on the file Then create an extra file of unique customers numbers Read the file with the unique customers numbers and for each one go to the original file using the KEY option in the SET statement to retrieve and concatenate the Prod and Acc_flag variables. The _iorc_=0 and _error_=0 at the end of the step are to prevent error messages in the log. data original(index=(Cust_no)); input Cust_no Prod $ ACC_flag $ Bal $ ; datalines; 111    a1    qaz    10 111    b1    wsx    11 222    c1    wsx    233 333    d1    grw    251 444    e1    dc    25 444    f1    vefw    1231 444    g1    wefr    36 555    h1    bdfgh    615 666    i1    qe    767 777    j1    erg    31 777    k1    eth    472 777    l1    ret    251 777    m1    dfwef    3761 888    p1    ffef    36 run; proc sort data=original(keep=cust_no) out=customers nodupkey; by Cust_no; run; data new; set customers; length Prod_con Acc_con $50; do until (_iorc_ ne 0);    set original key = Cust_no;    if _iorc_ = 0 then do;      Prod_con = CATX('|',Prod_Con,Prod);      Acc_con = CATX('|',Acc_con,Acc_flag);    end; end; output; _iorc_ = 0; _error_=0; run; proc print; run; ... View more