Loyalty glossary · 6. Retail and grocery (20)
Sku Level Data
Sku level data is the record of member purchases at individual product level, capturing what each member bought and when, which powers accrual rules, activity based qualification, and actuarial models.
Sku level data is the raw transaction log at its most granular: one line per item purchased, not one line per shopping trip. It is the input that makes accrual rules precise, activity based qualification honest, and actuarial models defensible.
Most retail loyalty programmes still run on basket totals. They record that a member spent 45 dollars in a visit, but they cannot tell which of the 14 items in that basket were eligible for points. The result is a flat earn rate per dollar, which is simple but wrong for any product that should earn differently.
The trap is treating sku level data as a nice to have. Activity based qualification without it becomes a crude spend threshold. A programme that promises a bonus tier after three purchases of any organic product cannot verify the claim if its data shows only a produce aisle total of 20 dollars, not three separate qualifying skus.
Set accrual at 2 points per dollar on eligible skus and 0 points on excluded skus. A member spends 40 dollars on eligible items and 10 dollars on excluded items in one month. The correct accrual is 80 points, not 100 points if the programme used the total 50 dollar spend. Across 12 months, the error compounds to 240 points per member.
Actuarial models need sku level data to set breakage assumptions by product category. A model that treats all points as equally likely to be redeemed will overstate the liability for points earned on low redemption products like private label staples and understate it for high redemption products like fuel discounts.
The position is simple: a loyalty programme that cannot see sku level data is not measuring what it promises. It is guessing, and the guess is almost always wrong in the direction that flatters the P&L until a redemption spike proves otherwise.