Product classifier check
Shows how fully the required attributes of a product table (group, category, season, size, color, fibre content) are filled, how many duplicates there are and which attributes fall below your target. Without a single classification, analytics does not add up.
At a glance
- What question does it answer?
- Is the product table filled in well enough to build analytics on it.
- What data is needed?
- The number of items, the number of duplicates and the target fill rate.
- How to read the result?
- In a table of 1,200 items the average fill rate is 93.4%. Three of seven attributes are below the 98% target: subcategory (94.2%), color (91.7%) and fibre content (70%). There are 554 empty cells and 1.5% duplicates: readiness 91.9%. Example on the default values: Average fill rate — 93.4%; Weakest attribute — 70% · Fibre content; Attributes below the target — 3 of 7; Empty cells in total — 554.
Why and how to reason
Analytics only adds up from tables where the same thing is named the same way. If half the items lack a season or color, reports by season and by color mislead, and duplicate codes double stock and sales. So the check computes fill rate: the share of items where an attribute is filled, for each required attribute and on average. In the example of 1,200 items: group 100%, category 99.7%, subcategory 94.2%, season 98.3%, size 100%, color 91.7%, composition 70%. Average fill rate 93.4%; the target is 98%, and three attributes are below it (subcategory, color, composition). Empty cells 554: the weakest column, material composition, gives 360. Duplicates 18 of 1,200 are 1.5%. The main value of the check is that it shows where to fix data first, not everything at once.
When to use it
It fits before introducing analytics, purchase planning or launching on a marketplace where composition and size are mandatory. Assumption: filled does not mean correct: the check counts empty cells, not errors in values. It does not assess quality of names and photos. Material composition is especially important for marking and certification: empty values there are not only an analytics nuisance but a risk.
How to read the result
List attributes below target and assign owners: the weakest first (composition 70%), then the rest. The sum of empty cells gives the amount of work: 554 cells is hours, not weeks. Remove duplicates together with their stock and sales, or reports will be distorted the other way. Decision: set rules (no composition, no entry to the warehouse) and check completeness regularly, not once. It is worth fixing a dictionary of values and a rule for new items (no card without mandatory attributes), and then cleaning is a one-off job and not an endless one.
Common mistakes
Typical trap: counting a cell with value “none” or a dash as filled; setting a 100% target and giving up at the first difficulty; not agreeing on a single value dictionary (beige and beige-ish are two values); cleaning once and not maintaining. Look for duplicates by barcode and by name in different ways. Do not count size variants of one model as duplicates: they are different SKUs.
What data is needed?
- The number of items, the number of duplicates and the target fill rate.
- For each attribute: how many items have it filled.
Inputs
| Field | Unit | Value in the example | Meaning |
|---|---|---|---|
| SKUs in the table | pcs | 1,200 | An export from the accounting system or a spreadsheet. |
| Duplicates: one item under two codes | pcs | 18 | Find by name + size + color. Empty: zero. |
| Target fill rate of an attribute | % | 98 | Your own target: there is no single fill-rate standard in Russia, 98 here is an example. Set a stricter target for attributes that reports are built on. |
How it is calculated
| Step | Formula |
|---|---|
| Fill rate | items filled ÷ total items |
| Duplicates | duplicates ÷ total items |
Worked numbers
Default values from the app; in the interactive version you enter your own.
- Average fill rate
- 93.4 %
- Weakest attribute
- 70 % · Fibre content
- Attributes below the target
- 3 of 7
- Empty cells in total
- 554
- Duplicates in the table
- 1.5 %Shown apart from fill rate: there is no combined index, because completeness and duplicates break analytics in different ways.
Attribute fill
| Attribute | Filled | Fill rate | Missing | Target |
|---|---|---|---|---|
| Group | 1,200 | 100 % | 0 | yes |
| Category | 1,196 | 99.7 % | 4 | yes |
| Subcategory | 1,130 | 94.2 % | 70 | no |
| Season and collection | 1,180 | 98.3 % | 20 | yes |
| Size | 1,200 | 100 % | 0 | yes |
| Color | 1,100 | 91.7 % | 100 | no |
| Fibre content | 840 | 70 % | 360 | no |
“Subcategory”: 94.2 % filled against a target of 98 %; 70 items are missing.
“Color”: 91.7 % filled against a target of 98 %; 100 items are missing.
“Fibre content”: 70 % filled against a target of 98 %; 360 items are missing.
Duplicates distort sales and stock: merge the codes before analysis.
A working structure: group → category → subcategory → season → model → size and colour. Align it with labelling codes and national classifiers; check the exact reference against an official source.
Nuances and limits
- Completeness is counted here, not the correctness of the values.
- Duplicates with different codes distort sales and stock: merge them before analysis.
- Align the structure with labeling codes and national classifiers; check the exact reference against an official source.