Indicuma
RUSend a sample

Indicuma / Cases / Attribute extraction

Attribute extraction

From inconsistent dairy titles to a reviewable attribute catalog.

A typical data-enrichment case: extract analytical fields from raw SKU names without hiding uncertain interpretations.

CaseFMCG Product Data Enrichment

A dairy catalog often already contains the information needed for analysis — but the information is locked inside inconsistent free-text titles.

Demonstration case. This page describes a typical working method and deliverable format. It does not claim a disclosed client result or measured commercial outcome.

Input

Excel / CSV with source SKU titles

Focus

Brand, type, pack, size and claims

Output

Structured file with review flags

Situation

A product team receives dairy assortments from operational or market-data files. Product characteristics appear in titles, but with varying abbreviation, order and unit notation.

Typical input rows

Milk VKUSNOTEEVO pasteur. Special 3.2% jug 930g
VKUSNOTEEVO Milk UHT 3.2% 950g carton
YOGHURT DANONE THERMOST. STRAWB. 3.3% CUP 170G

Problem to solve

Free text does not support reliable filtering or matching. A user cannot safely analyse processing types, compare pack sizes or validate similar products unless the title is decomposed into standard attributes and unresolved content is made visible.

Approach

  1. Normalize source representation. Align case, spacing, separators, unit notation and known safe abbreviations while retaining the source title.
  2. Extract formal attributes. Apply controlled logic for size, volume, fat percentage, multipack patterns and explicit packaging terms.
  3. Standardise semantic attributes. Use reviewed dictionaries for brands, product types, technologies, variants and claims.
  4. Flag uncertainty. Keep unparsed remainder, conflicts and interpretations that cannot be verified from the title alone.

Illustrative output

Raw titleCategoryBrandProcessingPack / SizeStatus
Milk VKUSNOTEEVO pasteur. Special 3.2% jug 930gMilkVkusnoteevoPasteurizedJug / 930 gHigh-confidence parse
VKUSNOTEEVO Milk UHT 3.2% 950g cartonMilkVkusnoteevoUHTCarton / 950 gDistinct SKU signal
What the deliverable provides

A structured spreadsheet retaining source titles, parsed fields, normalized values, unparsed fragments and a queue for controlled review.

Business use

  • Prepare consistent product attributes for BI or PIM intake.
  • Compare assortment, formats and sizes using the same definitions.
  • Expose terms requiring dictionary extension before scaling automation.
  • Provide a safer basis for later retailer-to-master matching.

Need to turn product titles into usable attributes?

Send a sample Excel or CSV file. Indicuma will identify which fields can be extracted safely and where a review rule is needed.

Send a sample →