Glossary

• Job — a single unit of work that runs in the background and appears in the Job Queue.
• Job Queue — the shared list of active and recent jobs, with progress and per-job actions.
• Utility — one of the five built-in tiles. Each can be turned on or off individually.
• Output folder — the root location where utilities save files, each into its own subfolder.
• Companion file — the removals or rejects file written beside an output, listing whatever the run dropped or refused.
• Recipe — a named, saved sequence of steps that you run by hand on files you pick.
• Pipeline — the ordered steps a recipe runs on each file, for example clean, then convert to Excel.
• Dialect — how a delimited text file is laid out: the delimiter, the quote and escape characters, the encoding, whether there is a header row. DataTray shows you the dialect it detected instead of guessing silently.
• Data profiling — reading a file column by column to report its types, missing values, distinct values and outliers. Lives in Data Explorer → Columns.
• Quality score — the 0–100 headline, made of five weighted dimensions (completeness, validity, uniqueness, consistency, structure) with the weights shown. Lives in Data Explorer → Columns.
• Data dictionary — a one-row-per-column description of a file: name, type, nulls, distinct values, examples and inferred role. Exported from Data Explorer → Columns.
• DuckDB — the local, in-process engine that lets DataTray open files far larger than Excel can, entirely on your device. It ships inside the app; there is nothing to install and it has no server.
• Parquet — a compressed, columnar file format used for large datasets. A "Parquet file" is often really a folder of parts, and DataTray reads either.
• NDJSON — newline-delimited JSON: one JSON object per line. Also seen as .jsonl.
• Schema drift — the difference between two files' column sets: columns added, removed, retyped or renamed.
• Key column — a column that identifies a row, like a customer id. Naming one is what lets Compare Files report changed cells instead of just added and removed rows.
• Normalized matching — comparing values after ignoring case, punctuation and extra whitespace, so Acme Corp. and ACME Corp match.
• Fuzzy matching — comparing values by similarity score rather than equality, so near-misses match.
• Blocking — grouping rows before fuzzy comparison so only plausible candidates are ever compared. What keeps fuzzy matching tractable on a large file.
• Anonymisation — transforming data so a person can no longer be identified from it. In DataTray, only drop and redact do this.
• Pseudonymisation — replacing a value with a stand-in that can still be linked back with additional information. Mask, hash and tokenize are pseudonymisation, not anonymisation.
• Salt — extra text mixed into a hash so the same value hashes differently between runs. DataTray prints the salt in the report, so it is a record, not a secret.
• The bandwidth exchange — the sharing of your device's idle internet bandwidth that funds the free app; it never shares your files or personal data.
• Tray — the small icons near the Windows clock. DataTray runs here and reopens when you click its icon.
• Channel — the update track you follow: Stable (tested) or Beta (earlier, more risk).

---

DataTray — a free, local-first data utility belt for Windows.