I don't have any insider information, but it doesn't seem crazy to think that those discount stickers aren't in fact the same. At the very least they have the actual discounted price encoded them, it wouldn't be a big leap for the label printer to register a new SKU for "Semi-skimmed Milk, discounted to £0.30" so they can still track that item's stock.
I have just checked the barcodes, and all stickers are identical. Same barcode. Same text. Same everything. They are preprinted, not printed by a machine in the shop too.
Does it need it's own SKU? It is scanned when discounted.
The important information is "We did too much of this cheese sandwich today" and correlate that with stuff like store location, the weather, public holidays, travel, etc. That's the valuable data. Then when double-discounted there's a further scan "We really did too many cheese sandwiches today".
I'm not sure where the value-add in projecting future demand is with a mass of low-value-add information. Not giving it a unique SKU seems like a saving to me. If anything, doing random sampling might be worth it, but in measuring flow of what comes in and what goes out, it's going out on discount or trash in the end. The goal is to not overproduce, not waste time tracking overproduced/allocated stuff.
