GLOSSARY
Twenty-seven words
that come up in every conversation.
Product data has a language of its own, and that language is full of words that mean something slightly different on every platform. Here is what we mean by them, in plain sentences, so a conversation does not have to be about definitions.
THE FIELDS
Where a value lives.
Attribute
A property of a product stored separately: material, size, colour, standard. The difference with a sentence in the description is that a computer can read an attribute and cannot read a sentence.
Normalised value
The same value, always written the same way. Anthracite, charcoal and dark grey become one colour family. Without normalising you have filled fields and still no usable filter.
Taxonomy
A list of fixed values you pick from, instead of free text. Filters and feeds run on taxonomies. In WooCommerce that is called a global attribute, in Magento an attribute with options.
Multi-value attribute
A field that can carry more than one value at a time. Indispensable for fitment: one part fits dozens of versions, and those are not dozens of products.
Variant axis
The attribute a product’s variants split on, usually size or colour. Your stock hangs on it, so we stay off it. See Fashion.
Exclusion list
The categories, brands and fields we are not allowed to touch. You draw it up in week one and can add to it at any moment after that.
SKU
A product as you sell it, with its own product page. A jumper in four colours and six sizes is one SKU with twenty-four variants. Our meter counts SKUs.
WHERE IT GOES
The places where a filled field makes a difference.
Layered navigation
The filters beside or above a category page, also called faceted navigation. They only read taxonomies. A value sitting as free text on the page does not exist for these filters.
Product feed
The file that carries your catalogue to a channel: Google Shopping, a marketplace, a comparison site. What is not in a field does not travel with the feed.
Merchant Center
Where Google judges your feed. Missing fields produce warnings, wrong barcodes produce disapprovals, and repeated disapprovals can suspend the whole account. Which is why an invented GTIN is more expensive than an empty one.
Structured data
The schema.org markup in the source of a product page, letting a search engine read what is on the page instead of guessing. We put it in place per product, built out of the fields that exist.
Site search
The search function of your own shop. It usually searches titles and attributes, not the long description. Visitors who search buy more often, so they are the first to hit an empty field.
Alt text
The text description on an image. Needed for accessibility, counted in image search, and almost always empty or filled with a file name.
WHERE IT COMES FROM
The sources, and what they are worth.
Datasheet
The manufacturer’s technical sheet, usually a PDF. For us NOT a source: we read no pdfs. If a value stands only there, the field stays empty with us.
Supplier feed
The file you receive from a supplier. There is often more in it than ends up in the shop. For us NOT a source: we read only what stands in your shop itself. Put it in the shop as you do now, and we fill it in on the next pass.
Image recognition
Deriving a value from the product photo. We do NOT do this: there is no image recognition in Skuply. We do write alt text about a photo, built out of your own title.
GTIN and EAN
The barcode of a product. We never write it: we have no source to look it up in and never derive it. So it stays empty, and empty beats wrong.
MPN
The manufacturer part number. Like a GTIN it is a reference and not a property, so it can be copied and never invented.
OE number
The number of the original part a replacement refers to. In the parts sector the single most important search term there is. See Technical parts.
HOW WE MEASURE
The words from our own report.
Completeness score
The share of fields that are filled, across the attributes that apply to your categories. One number, always with the denominator and the measurement date, because without those two a percentage says nothing.
Confidence
The number between 0 and 1 saying how direct the source of a value is. A colour stated literally in the title scores 0.97. The same colour out of a sentence in the description 0.90, because there a word is more often about something else.
Threshold
The line below which a value does not go live. Set to 0.85 by default. Raising it means less live and a longer review list. Lower than 0.70 is not something we do.
Review list
Everything that stayed below the threshold, with the source and the proposal attached. Meant to be decided in ten seconds a line, not to be read. See The dashboard.
Holdout
Two thousand products we set aside at the start and deliberately never touch. They move with the season and not with us, so the difference shows what the work produced.
Baseline measurement
The measurement before anything happens, per category and per attribute. Without a baseline every later improvement is an opinion.
Namespace
Our own field space inside your shop, where every value sits with source, timestamp, confidence and confidence. From there we mirror into the field your theme shows. Which is why rolling back is one action.
Provenance
Where a value came from, in four kinds: derived, matched, generated or yours. That fourth is always the largest group. See Safety.
PER PLATFORM
The same intention, three different words.
WHAT A FILTERABLE FIELD IS CALLED
global attribute (pa_)
WooCommerce. A real taxonomy. Its counterpart is a custom attribute: plain text, invisible to the filter.
metafield with definition
Shopify. Without a definition a metafield is loose data: invisible in the admin and unusable in channels.
attribute with options
Magento. A dropdown or multiselect in the right attribute set, switched on for layered navigation. A text field does not filter.
bulk operation
Shopify. Thousands of products in one job instead of thousands of separate calls, so your rate limit stays clear.
REST and GraphQL
The two ways we talk to a shop. Both with one key and without anything landing inside your installation.