Parameters

How mismatched names and values in feeds become one consistent parameter table — and in what order it happens.

On this page

The order it does not pay to skip
How you tell whether it is done
Automation

This is the work a PIM exists for. Three vendors send the same property under three names, spelled three ways:

Vendor	Field name	Value
A	`Barva`	`černá`
B	`Colour`	`Black`
C	`farba-produktu`	`BK`

A customer in the shop wants one Colour filter with one black value. The road from the first to the second has four steps, and they have to happen in this order.

The order it does not pay to skip

Define the parameter
A target has to exist first. The Colour parameter carries a type, a unit, example values and context for the AI.
→ Parameter definitions
Map the names
Barva, Colour and farba-produktu are all matched to Colour. Without this step the system does not know they are the same thing.
→ Parameter mappings
Map the values
černá, Black and BK collapse into black. Only now does the shop filter work.
→ Value mappings
Fill in what is missing
What a vendor never sent is added by hand, or extracted by the AI from the description and the photos.
→ Filling in values

How you tell whether it is done

Instrument	What it shows	Where
Normalisation coverage	The share of values that are mapped.	The Dashboard and System overview.
Parameter template	Which parameters a product in a category should carry.	Templates — completeness is measured against them.
The "missing parameters" filter	The specific products that are short of something.	Advanced filters in the product list.

Automation

Once normalisation is done, it can maintain itself:

Templates — what a product in a category should carry.
Rules — conditional actions over values.
Bulk actions — a one-off change across many products.
Product flows — the order all of it runs in after an import.