Product data quality

What product data quality means

Product data quality describes how complete, correct and consistent item information is. It is not a design question but a commercial one. A missing dimension leads to a return, an unclear statement to a customer enquiry, a wrong colour to a complaint. Every one of those cases costs money that the costing of the item usually does not plan for.

The criteria quality can be measured by

• Completeness: are all attributes filled that the customer needs for their decision.
• Correctness: do the values match the product. A wrong weight is more expensive than a missing one.
• Consistency: are dimensions held in the same unit and spelling, and are statements comparable across items.
• Unambiguity: is it clear which detail is meant, and are there no contradictory fields.
• Currency: is the item still described the way it is delivered.

How errors arise

The most common reason is three sources for the same detail. The supplier sends data in their format, the description is written by hand at the retailer, and at some point someone corrects a field directly in the shop. After that there are two truths and nobody knows which applies.

The second reason is mandatory fields without validation. A system with thirty attributes of which none is required collects gaps. The third reason is missing ownership: when anyone may write and nobody checks, quality falls slowly and permanently.

What can be automated

Four checks are machine feasible and find a large part of the errors. First, locating missing values in attributes that are decisive for the customer, such as dimensions, material and colour. Second, plausibility: a weight of 500 kilograms for a coffee cup is an error, not an extreme value. Third, the unit: if dimensions and weights are not maintained in the same unit, sorting and filtering are impossible. Fourth, duplicates: duplicate item numbers and duplicate images in the assortment.

A further lever is checking against the supplier's catalogue. Anyone reading the manufacturer data regularly finds deviations before customers do.

What it costs when nothing happens

Poor product data acts in three places. Search engines rate pages with complete details higher, and filters that are not filled make items invisible. Customers do not find what they are looking for and buy elsewhere. And customer service spends time on questions the page could answer. The last item is usually the most expensive, because it repeats with every enquiry.

Frequent questions

• Where do I start? With the attributes named in the most frequent customer questions.
• Who owns quality? One named person, even when several maintain the data.
• Is a tool enough? Without rules and ownership a tool only finds what has already gone wrong.