Pick accuracy over shipped order lines, bucketed by `ship_month` (`toStartOfMonth(ship_date)`), `site`, `client_segment`, and `temp_class`. `accurate_lines` counts lines picked accurately; `total_lines` counts all lines. `pick_accuracy` is exposed as a FORMULA measure (sum(accurate_lines) / nullif(sum(total_lines), 0)) so it remains correct across any window — never divide the two count columns client-side. The `site` dimension is REGION-LEVEL: it is `site_region` rolled into the `site` label, because `mart_fulfillment_daily`'s grain does not carry `site_id`. A true `site_id`-level pick-accuracy breakdown is not available in P0 without widening the mart's grain — do not present these region-level numbers as site-level.
- highDividing the two count columns client-side
`accurate_lines` and `total_lines` are counts. Request the `pick_accuracy` formula measure — the metric layer divides with `nullif` window-correctly. Dividing the summed columns yourself across multiple buckets is an average-of-averages, biased.
- mediumReading total_lines as an order count
`total_lines` counts order LINES, not orders — an order has many lines. Do not read it as `order_count` (that lives on the `fulfillment` metric); the two are different grains of counting.
- mediumReading the `site` dimension as site_id-level
The `site` dimension is REGION-LEVEL in P0 (it is `site_region`); the source rollup does not carry `site_id`. Do not attribute a region-level accuracy number to an individual warehouse.
- “How does pick accuracy trend by temp class?”
- “Which region has the best pick accuracy this quarter?”
- “Pick accuracy for life_sciences clients over the last 6 months.”
- fulfillment
Fed by the same `mart_fulfillment_daily` rollup — pairs picking quality (line accuracy) with delivery performance (OTIF / on-time / in-full) for the same period.