Same formula generating inconsistent results

Interesting. I enjoyed puzzling this one out.

Try this formula in E9

=MATCH(TRUE,IF(MMULT(IF(ROW(D9:D$16)>=TRANSPOSE(ROW(D9:D$16))=TRUE,1,0),D9:D$16)>=C9-J$5,TRUE,FALSE),0)-1

I noticed that you made no allowance for starting stock. That is why you see J5 in my formula.

I can't help thinking this would be simplified if you kept your B3:BM5 data in a normalised data table

In your case columns for

Date
Product
Category ' i.e. Opening Stock, Inbound, Sales, Forecast, Weeks to fulfil Order
Value

Then have a report matrix with formulae that produce the three statistics from the data table above.
i.e. rows for Date and columns for Product, S/Ratio, Unfilled Orders and L/Time

Maybe try at E9

=IF(C9>SUM(D10:D$17),"PO",IFNA(MATCH(C9-0.1,MMULT(--(ROW(D10:D$17)>TRANSPOSE(ROW(D10:D$17))),N(+D10:D$17))),0))

Many thanks to each for the prompt responses and solutions.

Never had a case where the same formula in the same Dataset generated inconsistent results, so this has been a first all the way.

mehmetcik, apologies if I have misunderstood the point, but the initial Opening Stock in J5 is in my calculation for the Opening stock in week 29?

Richard, appreciate the transposing Table suggestion, but can't see any advantage to the End User? I provided one SKU across eight weeks. The 'real' file has sixty SKUs covering three years, so that would be over nine thousand rows.
As it stands, 'freezing' Cols A - I and using the Week Number to pull the relevant week across, with the SKUs under each other, the planners can look down each column and see an "overview" of each element across the range at one glance. Not sure it would be as effective looking across sixty columns in the same row?

Bo_Ry, It works perfectly, but value your explanation of how looking for errors in matching the weekly Inbounds with the Unfilled Order value "-10%" solves the problem? And what is the "N" at the end if a match is found?

And thanks again, everyone.

Ochimus