Category assignment logic

Each item code in the Tally Movement Analysis reports was matched to a category in four passes:

1. Exact match (whitespace/dash/case-insensitive) against a master mapping built from all three supplied Google Sheets — the "Item to category mapping" tab, the "2025 calculation process" tab, and the "Website item id to category mapping" sheet (~2,900 unique item codes combined).
2. For codes not found, keyword rules mirroring the business's own "Type" shorthand convention (e.g. MEDAL→Medals, FRAME→Frames, FIBER→Fiber Trophy, trailing N-suffix codes→Mementos, BRANDING/LANYARD→Services/Branding).
3. An exact match against 107 item codes Rahil manually labeled from the original uncategorized list.
4. 37 fuzzy rules derived from those 107 labels, propagated to every other unlabeled code sharing the same prefix, numeric model-family, or keyword.

Category contribution definition

"Category contribution" = each category's share of total Outward (sales/dispatch) Value for the period. Inward (purchase) value/qty is shown alongside for reference but the % contribution ranking is based on Outward Value, consistent with the "Sum of Gross Value" pivot approach used in the client's own 2025 calculation process tab.

Manual labels + fuzzy propagation

Rahil reviewed the original 1,239-item "Uncategorized (mapping gap)" export and directly labeled 107 codes across four categories: Memento, Trophy (generic), Corporate Trophy, and Fiber Trophy. Each rule below generalizes one label to every other code sharing its prefix (e.g. all "CP*" codes), numeric model-family (e.g. all "4288-*" variants), or an embedded keyword (e.g. any code containing "CRIC" or "IPL") — the same logic the business already uses internally for its "Type" shorthand column.

<aside> 💡

Applied to the remaining 1,132 unlabeled codes, these 37 rules resolved 272 additional item codes (₹22.2L / ~31% of the then-remaining unlabeled value). Full per-item resolution is in the companion Excel workbook's "Fuzzy Mapping Rules" and item-detail tabs.

</aside>

Data quality — remaining mapping coverage gap

<aside> ⚠️

After all four passes, 16.2% of July 2026's outward value (₹433,290) and 13.0% of the 2024–2026 outward value (₹4,914,478) remain in "Uncategorized (mapping gap)" — down from 37.7% and 31.9% before Rahil's manual labels.

These are codes that don't share a prefix, model-family, or keyword with any labeled example (mostly one-off model codes and a long tail of low-value items). They were not guessed into a specific category to avoid misrepresenting the split.

Recommendation: extending the manual-label list (or the source mapping sheets directly) would close most of what remains — the companion CSV of remaining uncategorized codes can be re-sent for another labeling pass if useful.

</aside>

Trend methodology limitation

The 2024–2026 source file reports period totals per item, not a monthly time series. The combined trend analysis is therefore a comparison of July 2026 against the 28-month average and share — a valid indicator of relative momentum, but it cannot show month-by-month seasonality within the 28-month window. A true monthly trend would require re-running the Tally Movement Analysis report once per month.

Companion files

A full PDF report and an Excel workbook (with item-level detail, filters, native charts, and a dedicated "Fuzzy Mapping Rules" tab) were also generated and shared directly with Rahil outside of Notion.