A logistics company stores shipment tracking documents in an Azure Cosmos DB for NoSQL container. Each document contains a large nested 'rawSensorPayload' object with dozens of properties that are never used in query filters, ORDER BY, or JOIN clauses — they are only returned when a document is read by id. The write workload is heavy, and the team notices high RU charges on every insert and update. Queries only ever filter on 'shipmentId', 'status', and 'region'. What is the most effective change to reduce write RU consumption while preserving query performance?