A view of raw catalog strings versus true normalized base parts, with manufacturer mix and heuristic product-type grouping from the normalized parts dataset.
The dataset contains 70 source rows and 70 distinct raw part strings, but normalization reduces the true base-part count to 67. The difference comes from collapsing formatting variants that point to the same underlying device.
Manufacturer counts are shown from source rows, while product types are grouped into broader families using keywords and product-name heuristics. That avoids over-fragmentation from inconsistent free-text descriptions.
| Normalized base part | Raw variants collapsed | Source description |
|---|