**Short answer:** the dataset has almost no real categorical variety apart from the stock identifier. Company and ticker are a 1:1 pair, and the other categorical columns hold a single constant value. **Company and ticker (`company_name`, `stock_name`)** - Each has 152 distinct values (the card estimated ~149). The top-value counts are identical in both columns, so each company maps to exactly one ticker. - The most common values are BE Semiconductor (BESI.AS), Soitec (SOI.PA) and ASM International (ASM.AS), with 3,475 rows each (0.82% of all 422,729 rows). - Next is Melexis NV (MELE.BR) with 3,473 rows. Elmos Semiconductor, Infineon and SÜSS MicroTec follow with 3,448 each (0.816%), then centrotherm international and Aixtron with 3,445 each. - Weebit Nano, BrainChip and IQE plc have 3,439, 3,439 and 3,431 rows. - The US-listed names in the top 20 (NVIDIA, Applied Materials, KLA, Analog Devices, Marvell, Wolfspeed and others) all have 3,415 rows, or 0.808%. - The shares are nearly uniform, between about 0.81% and 0.82%. That is expected because the grain is one row per stock per trading day. Frequency here reflects how long each ticker has been listed, not its popularity or importance. The top-20 lists I pulled show only the highest-count companies. They do not show the bottom of the 152, so I can't say how low the smallest counts go. - The first dates start on 2012-01-02 for most European listings and 2012-01-03 for the US and Australian ones. The last dates are 2025-08-01, except the two ASX stocks (Weebit Nano, BrainChip), which run to 2025-08-04. **Constant metadata columns** - `_batch_id` (batch_d9e1fae7c9b5), `_source_system` (api) and `_source_file` (the single `semi_conductor_se.csv` path) each have one distinct value covering 100% of rows. They carry no analytical information. **Data-quality note** - One company name contains an HTML entity: "Alpha & Omega Semiconductor" (AOSL). It should read "Alpha & Omega", and it could break joins or name matching against other sources.