Here is a Data Step Approach to the solution. Requires data to be pre-sorted by Customer / grouped as shown in your sample Data Sets. Added Customer 2 for checking the Data Step.   Using Tab1, we get the Month as Number from Date1 and compute perdaySales as Sales by the number of days of the month.   Using Tab2, we again get the Month as Number from Date2 and find the leftdays as the difference between number days of the Month and Date2.   In the final Data Step: In the first Do-Loop, we store the perdaySales in an array (K) indexed by Month (here 1 to 4) . In the second Do-Loop, if the leftdays is 7 or more, it is a simple lookup for the month in the array to get perdaySales and multiplication by 7. If leftdays is less than 7, we take the perdaySales for the current month and multiply it by leftdays PLUS take the perdaySales for the NEXT month and multiply by (7 - leftdays).   The array can be sized using the macro variable (&size) for changing number of months in Tab1. Hence, the Data Step is dynamic.   data tab1; input Customer Date1 :monyy7. Sales; eom = intnx('month', date1,0,'end'); numdays = day(eom); mon = month(date1); perdaySales = Sales / numdays; datalines; 1 Jan2018 1110 1 Feb2018 1245 1 Mar2018 1320 1 Apr2018 1100 2 Jan2018 1110 2 Feb2018 1245 2 Mar2018 1320 2 Apr2018 1100 ; run; proc print data = tab1; format date1 monyy7.; run; data tab2; input Customer Date2 date10.; leftdays = day(intnx('month',date2, 0, 'end')) - day(intnx('day', Date2, 0, 'end')) + 1; mon = month(date2); datalines; 1 01Jan2018 1 08Jan2018 1 15Jan2018 1 22Jan2018 1 29Jan2018 1 05Feb2018 1 12Feb2018 1 19Feb2018 1 26Feb2018 1 05Mar2018 2 01Jan2018 2 08Jan2018 2 15Jan2018 2 22Jan2018 2 29Jan2018 2 05Feb2018 2 12Feb2018 2 19Feb2018 2 26Feb2018 2 05Mar2018 ; run; proc print data = tab2; format date2 date10.; run; %let size = 4; data want; array k[&size] _temporary_; do until(last.Customer); set tab1; by Customer notsorted; k[mon] = perdaySales; end; do until(last.Customer); set tab2; by Customer notsorted; if leftdays >= 7 then weeksales = k[mon] * 7; else weeksales = k[mon] * leftdays + k[mon+1] * (7 - leftdays); output; end; keep Customer Date2 weeksales; run; proc print data = want; format date2 date10.; run; The Ouput is shown. The first over-lapping interval, the weeksales is slightly different here and for second interval it is same. Obs Customer Date2 weeksales 1 1 01JAN2018 250.645 2 1 08JAN2018 250.645 3 1 15JAN2018 250.645 4 1 22JAN2018 250.645 5 1 29JAN2018 285.276 6 1 05FEB2018 311.250 7 1 12FEB2018 311.250 8 1 19FEB2018 311.250 9 1 26FEB2018 303.715 10 1 05MAR2018 298.065 11 2 01JAN2018 250.645 12 2 08JAN2018 250.645 13 2 15JAN2018 250.645 14 2 22JAN2018 250.645 15 2 29JAN2018 285.276 16 2 05FEB2018 311.250 17 2 12FEB2018 311.250 18 2 19FEB2018 311.250 19 2 26FEB2018 303.715 20 2 05MAR2018 298.065 ... View more

Hi @roushankumar,   The segmentation violation error you observed in PROC FORECAST when specifying INTERVAL=WEEK and SEASONS=MONTH appears to be due to the fact that weeks do not nest within months (or years).  In other words, some weeks in the input data set span two different months.  For this very reason, modeling seasonal variation in weekly data can be particularly challenging.   Because PROC FORECAST is considered to be obsolete (see the Overview section of the FORECAST procedure documentation for details), you might want to consider some other approaches for modeling and forecasting your weekly data.  PROC UCM might be your best alternative.  For example, you could define monthly dummy variables and then include them in either the MODEL statement as simple regressors or in the RANDOMREG statement if you want the coefficients of these regressors to vary over time.  A similar alternative is to examine your weekly data and define dummy variables for the weeks in your series that are higher/lower than average from year to year.  You can then specify those dummy variables--rather than monthly dummy variables--in your model.  This alternative is helpful when you have seasonal variation that does not occur in a specific month from year to year, such as an Easter effect, which might occur in March or April.  Another alternative is to use the BLOCKSEASON statement in PROC UCM to define blocks of 4 contiguous weeks (to approximate 1 month) with 13 seasons per year.  For example, for weekly data: BLOCKSEASON BLOCKSIZE=4 NBLOCKS=13;  produces a seasonal cycle of 4*13=52 weeks.   I hope this information and these suggestions are helpful!  Links to the aforementioned sections of the documentation are noted below. DW   PROC FORECAST - Overview http://go.documentation.sas.com/?docsetId=etsug&docsetTarget=etsug_forecast_overview.htm&docsetVersion=14.3&locale=en   PROC UCM - RANDOMREG statement http://go.documentation.sas.com/?docsetId=etsug&docsetTarget=etsug_ucm_syntax18.htm&docsetVersion=14.3&locale=en   PROC UCM - BLOCKSEASON statement http://go.documentation.sas.com/?docsetId=etsug&docsetTarget=etsug_ucm_syntax04.htm&docsetVersion=14.3&locale=en ... View more

Thanks! Most sas procedures have a 'by' command to create models by a specific class of variable. Does proc hpneural have anything similar? Going around in loops is taking a lot of time. ... View more