Distribution App Development

Companies building third-party logistics platforms, wholesale distribution systems, and B2B fulfilment operations need engineering teams who understand the specific data architecture challenges of multi-warehouse inventory, order wave planning, carrier rate management, and the route optimisation logic that determines whether last-mile delivery is profitable. Scrums.com provides dedicated software engineering teams for distribution app development, deploying production-ready systems with append-only goods receipt ledgers, configurable pick-pack-ship workflows, carrier integration adapters, and the proof-of-delivery infrastructure that closes the fulfilment loop.

Warehouse Management, Inbound Receiving, and Putaway Architecture

Distribution platforms manage the full inbound flow from purchase order through goods receipt to putaway, and each step must be traceable as the basis for inventory accuracy.

Inbound receiving uses an append-only goods_receipt_events table: purchase_order_id, supplier_id, received_at, receiving_dock_id, and line items capturing variant_id, expected_quantity, received_quantity, and condition (ACCEPTABLE | DAMAGED | REJECTED). Discrepancies between expected and received quantities write discrepancy_records linked to the goods receipt. The inventory ledger is updated only after receipt confirmation: a RECEIVED event writes inventory_transactions rows for accepted quantities; DAMAGED quantities write to a separate quarantine_inventory table pending quality disposition.

Putaway management assigns received stock to storage locations based on a putaway_rules config table: rules specify preferred zones by product category, whether to consolidate with existing stock of the same SKU, and maximum pallet height per zone. A putaway_tasks table assigns work to warehouse operatives: variant_id, from_location (receiving dock), to_location (storage bin), quantity, status (PENDING | IN_PROGRESS | COMPLETED | EXCEPTION), and completed_at. Task completion writes the inventory_transactions row that moves stock from receiving to the bin location: inventory is not considered in storage until the putaway task reaches COMPLETED.

The warehouse_locations table models the physical structure: warehouse_id, zone_id, aisle, bay, level, bin, location_type (BULK | PICK_FACE | STAGING | RECEIVING_DOCK | DISPATCH_DOCK), and capacity constraints. Location capacity is enforced by checking current occupancy (aggregated from inventory_transactions) against the configured limit before any new putaway_tasks row is written.

Order Fulfilment, Pick-Pack-Ship Workflow, and Carrier Integration

Distribution order fulfilment connects inbound customer orders to the physical pick, pack, and ship workflow inside the warehouse, with inventory reservations ensuring no two fulfilment waves conflict over the same stock.

Outbound orders arrive through an order_intake_events table (from EDI, API, or manual entry) and are normalised into an orders table with a state machine: RECEIVED / ALLOCATED / WAVE_ASSIGNED / PICKING / PICKED / PACKING / PACKED / DISPATCHED / DELIVERED | CANCELLED | RETURNED. Each transition writes to order_status_events. Inventory allocation uses SELECT...FOR UPDATE to lock the relevant warehouse_inventory rows and write allocation_records linking order lines to specific bins and lot numbers. Allocation holds against quantity_reserved; inventory does not leave the warehouse_inventory table until dispatch.

Wave planning groups orders into fulfilment waves based on configurable criteria: cut-off time, carrier collection window, order priority tier, and shared pick paths to minimise operative travel. A wave_plans table records wave_id, planned order IDs, warehouse_id, target_dispatch_at, and status (PLANNED | RELEASING | IN_PROGRESS | COMPLETE). Releasing a wave generates pick_tasks rows sequenced by bin location to minimise walking distance. Task sequencing is driven by a pick_sequence_config table rather than hard-coded aisle order.

Carrier integration uses a carrier_adapter layer that normalises FedEx, UPS, DHL, Royal Mail, and customer-specific parcel contracts into a rate_quotes schema. Rate shopping compares carrier responses at label generation time and selects the cheapest option meeting the required service level. Carrier API calls write carrier_api_events rows for every request and response. Label generation writes a shipment_labels row; tracking number assignment writes to shipment_tracking_numbers. Both are immutable after creation.

Distribution apps like these are built and delivered by dedicated engineering teams through our mobile app development service.

Route Optimisation, Last-Mile Delivery, and Proof of Delivery

Last-mile delivery planning uses OR-Tools VRPTW (Vehicle Routing Problem with Time Windows) to compute optimal stop sequences across the delivery fleet. Input parameters (vehicle capacities, driver shift start and end times, customer delivery windows, and depot location) are read from configuration tables rather than hard-coded, allowing operations managers to adjust fleet and schedule parameters without engineering involvement. Solved route sequences are written to a route_assignments table as a point-in-time snapshot; when a replan is triggered by a new order or vehicle unavailability, a new snapshot row is inserted and the previous sequence is retained for audit.

Dynamic re-routing updates routes mid-shift when delivery failures, traffic events, or customer cancellations change the optimal sequence. A route_events append-only table records REPLAN_TRIGGERED, STOP_ADDED, STOP_REMOVED, and SEQUENCE_CHANGED events with the reason code and operator reference. The active route sequence for any vehicle is always derived from the latest route_assignments row: the route_events table provides the audit trail of why the sequence changed but is never the source of truth for what the driver should execute next.

Driver mobile apps communicate with the backend via a delivery_events API: ARRIVED_AT_STOP, DELIVERY_ATTEMPTED, DELIVERED, FAILED_DELIVERY (with a reason code: NOT_AT_HOME | ACCESS_DENIED | REFUSED | DAMAGED), and RETURNED_TO_DEPOT. Each event is appended to a delivery_status_events table. Proof of delivery (POD) captures signature_image_url, recipient_name, photo_url, and geolocation at the moment of the DELIVERED event. POD records are immutable after creation; a separate pod_disputes table handles customer disputes without modifying the original POD.

Failed delivery management follows a configurable re-attempt policy stored in a delivery_attempt_config table: maximum attempts per order, re-attempt interval, and escalation action (RETURN_TO_DEPOT | HOLD_AT_LOCKER | NOTIFY_CUSTOMER). After the maximum attempts are exhausted, a return_to_depot_events row is created and the order transitions to a FAILED terminal state: the inventory is returned to the warehouse_inventory ledger via a RETURN inventory_transaction.

Multi-Warehouse Operations, Inter-Depot Transfers, and Distribution Analytics

Multi-warehouse distribution platforms must route orders to the optimal fulfilment location without creating inventory inconsistencies between sites. Order routing uses a fulfilment_routing_config table that defines rules by priority: route to the warehouse closest to the delivery address with available stock; if out-of-stock at the closest site, route to the next preferred warehouse. Routing decisions are logged in order_routing_events rows (the decision, the rule that triggered it, and the warehouse assigned), creating an auditable record of why each order was routed where it was.

Inter-depot transfers move stock between warehouse locations without passing through a customer order. A transfer is modelled as two inventory_transaction rows: a TRANSFER_OUT from the source location and a TRANSFER_IN at the destination, linked by a transfer_id. Both rows are written atomically within a database transaction so that stock is never simultaneously absent from both locations. Transfer state is tracked in a transfer_requests table (REQUESTED / APPROVED / IN_TRANSIT / RECEIVED | CANCELLED) with each transition appended to transfer_events. Transit inventory (stock that has left the source but not yet reached the destination) is held in a TRANSFER_IN_TRANSIT transaction type that does not contribute to available inventory at either location until the TRANSFER_IN event is confirmed.

Distribution analytics materialised views aggregate from the append-only ledger without querying transactional tables directly. Standard views include throughput_by_warehouse_by_day (orders received, picked, packed, dispatched, delivered), carrier_performance_by_lane (on-time delivery rate, average transit days, failed delivery rate per carrier and route), inventory_turn_by_sku, and return_rate_by_product_category. Because these views are recomputable from source rows, correcting a source event and refreshing the view regenerates accurate performance metrics without manual data patching.

SLA monitoring uses a sla_configs table keyed by customer account and order priority tier, defining maximum pick time, pack time, carrier collection deadline, and delivery window. A sla_breach_events table records any breach: order_id, the SLA metric that breached, the configured threshold, the actual value, and detected_at. SLA breach events are immutable; they are never deleted even if the underlying delivery is subsequently resolved, preserving the evidence required for customer penalty calculations and carrier performance reviews.