Reliable Remote Staffing
Since 2004
Client success story · CS32
Validating true interchangeability across a Fortune 500 multi-category catalogue
When an item is out of stock, a marketplace offers the shopper something else. That substitute is a recommendation the retailer is making on its own authority, and when it is wrong — a different size, an incompatible model, a single unit offered in place of a pack — the cost is not a failed sale but a shopper who stops believing the suggestions.
A global e-commerce leader running a catalogue across apparel, electronics, accessories, personal care, toys and household goods asked for its substitute pairings to be checked. StaffCloud audited over a million product pairs and recorded, for each one, whether the second item genuinely replaces the first.
Problem
The client supplied pairs — an original item and a proposed substitute — and the question for each was whether a customer would accept the second in place of the first. That is a judgement, and judgements made by forty people across twenty categories drift unless the criteria are fixed first.
Clothing, electronics, accessories, toys, beauty and pet supplies do not share a test for interchangeability. Apparel has to match on size and colour; electronics on specification; accessories on device compatibility; bundles on unit structure. One rule applied across all of them would be wrong in most.
Many pairs looked identical and were not: a slight variation in dimensions, a different compatible device model, another material, a different unit count in the pack. These are the cases where a fast audit produces a confident wrong answer.
Some products carried too little information or too few images to decide either way. Guessing on those would have put noise into the dataset, so they needed a category of their own rather than a forced call.
Solution
Before auditing began, the decision was reduced to a checklist every auditor applied in the same order: product type, functional purpose, visual similarity, size and dimensions, material and design, technical specifications and compatibility, and unit count and packaging.
Price, ratings, reviews and availability were deliberately excluded. They move daily and they describe the offer rather than the product — a cheaper item is not thereby a substitute, and an out-of-stock one does not stop being one.
Each audit worked from the product page itself: images, title and description, technical specifications, compatibility information, and size and dimension details. The subtle mismatches — the pack of four against the single, the case cut for a different handset — are only visible at that level.
Every pair was closed as Valid Substitute, Not a Substitute, or Missing Data. Non-substitute decisions carried a reason code naming the attribute that failed — size, colour, compatibility, material or function.
The reason code is what makes the dataset usable afterwards. A flat rejection tells the client the pairing was wrong; a coded one tells them which rule their matching logic is getting wrong, and how often.
Auditors were trained against the written instructions and worked examples. Random samples were pulled and reviewed a second time, and complex or uncertain pairs were escalated rather than decided alone.
With forty people on one dataset, consistency between them is the quality measure that matters — forty individually careful auditors applying forty slightly different thresholds would still produce an unusable result.
Impact
Substitution is a recommendation, and a catalogue makes it at the scale of its own size — which is why the accuracy of the rule matters more than the accuracy of any single pair. The work here was less about judging a million products than about fixing one definition of ‘substitute’ and holding forty people to it, then recording why each rejection failed so the client could correct the matching upstream.
Describe it in a sentence. We will tell you what we would do about it, and whether it is something we should be doing at all.
Tell us the workload
We will map the roles, the hours and the cost, and come back within one working day.
Or submit an enquiry — the full form, with a service line