Which products deserve more attention — and why?
01 / THE APPROACH
Keep product and variation identity clear
A parent product can contain variations with very different performance. Track product and variation identifiers consistently and retain historical order-line information even when a product has been deleted. A missing current catalogue record should not erase an old sale. Separate quantity sold, orders containing the item, and unique identified customers: they measure different behaviour.
02 / THE APPROACH
Compare sales velocity on an equal basis
Compare units per day over a recent window with units per day over an earlier baseline. Raw totals from a 30-day window and a 90-day window are not directly comparable. Where enough history exists, a same-period-last-year comparison can add seasonal context. Stock availability, price changes, promotions, and merchandising can all affect the interpretation.
03 / THE APPROACH
Measure refunds and downstream customer quality
Calculate refund amounts and, where reliable line-level data exists, refunded quantities. An order-level refund cannot automatically be assigned to every product. Separately examine customers whose first order contained the item: their repeat rate and fixed-window revenue may tell a different story from immediate product sales. Apply volume thresholds before surfacing a product as an opportunity.
Before you act on the analysis
- Preserve variation and historical product references.
- Compare daily rates across clearly defined windows.
- Distinguish product-level refunds from unallocated refunds.
- Use sufficient order volume and check seasonal or stock context.
WHAT WE’RE BUILDING
From the question to a useful next step.
Retail Science Co is developing product metrics, recent sales velocity, growth and decline observations, refund analysis, and acquisition-product quality comparisons. The goal is a focused investigation list rather than another large product table.
Join the early-access waitlist