Skip to main content
Back to case studies
E-commerce clientClient project

E-commerce · Web Scraping

B2B IT Hardware Catalogue with Translated Descriptions

Built by Van Data Team for E-commerce client

A scraper that captured the catalogue into a CSV of 77,588 rows and 14 fields across 319 brands, with product descriptions pushed through a per-locale translation pass.

SKUs across 319 brands
72,685
fields per SKU, from EAN to images
14

E-commerce72,685 SKUs × 14 fields

B2B product catalogue: coverage and data quality

Products
72,685
14 fields per SKU
Brands
319
distinct manufacturers
With an image
41%
30,071 of 72,685 SKUs
With a description
24%
17,798 SKUs carry copy
Built for this portfolio from the delivered data

Measured outcomes

  1. Outcome 01

    72,685 SKUs across 319 brands, 14 fields each

  2. Outcome 02

    Product descriptions pushed through a per-locale translation pass

  3. Outcome 03

    41% of SKUs with an image and 24% with a description, measured per brand

What the data says

  1. Ten brands are 71% of the catalogue; the tail is 309 deep.Lenovo alone is 21% (15,180 SKUs). By brand 30 the cumulative share reaches 93%.

  2. Media coverage is the catalogue's weakest field, not the scraper's.42,614 SKUs (59%) carry no image at all and only 5,223 carry more than one.

  3. Description coverage varies sharply by brand.CoreParts ships copy on 100% of its SKUs while Sony manages 1%, a gap worth flagging to the client.

What to decide

The calls this data supports, each with its evidence and the number to watch.

  1. Negotiate and merchandise around the ten brands that make 71% of the range.

    BecauseTen brands are 71% of the catalogue; the tail is 309 deep.

    Watch Share of SKUs in the top ten brands

  2. Fix product images first: 59% of SKUs have none.

    BecauseMedia coverage is the catalogue's weakest field, not the scraper's.

    Watch SKUs with at least one image

  3. Ask the brands with the thinnest copy, Sony first, for descriptions.

    BecauseDescription coverage varies sharply by brand.

    Watch Description coverage by brand

Challenge pressure

The client needed the full public IT hardware catalogue of EET Group, a B2B distributor, as structured product data, with descriptions translated for each locale.

The situation

A distributor catalogue is a long tail of manufacturers and part numbers. To be reused, each product needs its identifiers (EAN barcode and manufacturer part number), brand, title and condition as separate fields, and its descriptions ready for each locale. EET Group is the source here, not the client.

What we built

The scraper captured each product's EAN, manufacturer part number (MPN), brand, title, condition, short description, description table and up to six images, then sent the descriptions through a per-locale translation pass. The delivered CSV holds 77,588 rows; when it was analysed for the dashboard, 72,685 parsed cleanly and the 4,903 malformed rows were counted and skipped.

What changed

The client got the catalogue as 72,685 product rows across 319 brands, with image and description coverage measurable per brand. The dashboard shows the catalogue's shape: ten brands make up 71% of SKUs, Lenovo alone 21% (15,180 SKUs), and by the 30th brand the cumulative share reaches 93%. 41% of SKUs have an image and 24% carry a description.

Constraints and decisions

Completeness describes the source catalogue, not scraper loss: 42,614 SKUs (59%) have no image at all on the source and only 5,223 have more than one, and description coverage runs from 100% of SKUs for CoreParts to 1% for Sony, a gap worth flagging to the client. Source-site identifiers and image URLs are removed from the rows shown on this site.

Technology stack

CSVEAN and MPN keysPer-locale translation

Share this

Need a similar system?

If this case study maps to your current workflow, the next useful step is a scoped project conversation with constraints, timeline, and integration needs.

Related case studies

Academic research

SSRN Research Papers and Authors Dataset

Academic research clientClient project

239,037

papers delivered, 43 fields each

Recruiting

Cybersecurity Job Listings from Dice.com

Recruiting data clientClient project

500

Dice.com listings, 23 fields each

Social and brand

LinkedIn Company Posts and Engagement Dataset

Social media data clientClient project

15,192

posts from 12 company pages